Limits of Legality

This page intentionally left blank

Limits of Legality
The Ethics of Lawless Judging

Jeffrey Brand-Ballard

2010
Oxford University Press, Inc., publishes works that further
Oxford University’s objective of excellence
in research, scholarship, and education.

Oxford New York

Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam

Copyright © 2010 by Oxford University Press, Inc.

Published by Oxford University Press, Inc.

198 Madison Avenue, New York, New York 10016
www.oup.com
Oxford is a registered trademark of Oxford University Press.
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise,
without the prior permission of Oxford University Press.

Library of Congress Cataloging-in-Publication Data

Brand-Ballard, Jeffrey.
Limits of legality: the ethics of lawless judging / Jeffrey Brand-Ballard.
p. cm.
ISBN 978-0-19-534229-1 (alk. paper)
1. Judicial ethics—United States. 2. Judicial corruption—United States.
3. Rule of law—United States. I. Title.
KK8779.B73 2010
174′.3—dc22 2010007430

9 8 7 6 5 4 3 2 1

Printed in the United States of America

on acid-free paper
For my magnificent parents, Gabriella and Donald
Only this lifetime, but enough love for many
This page intentionally left blank
Acknowledgments

This book was conceived and written during my junior years in the Depart-
ment of Philosophy at George Washington University, where I have worked
with many fine people. I am especially grateful to my full-time colleagues:
Peter Caws, R. Paul Churchill, David DeGrazia, Michèle Friend, William
Griffith, Eric Saidel, Gail Weiss, and Tadeusz Zawidzki. In one way or an-
other each of them has supported me—in all my endeavors, at every oppor-
tunity. Several have made specific contributions to this book by attending
colloquia and responding insightfully to my arguments. For these contribu-
tions I thank Paul, David, Michèle, Bill, Eric, and my former colleagues Ilya
Farber and John Rudisill.
Paul, Bill, and David have served successively as chairs of the depart-
ment since I joined it. I thank Paul for welcoming me into the department
with open arms and for keeping them wide open ever since. Bill reached
out to me before I had even arrived. For seven years he has inspired me
with his intellect, sound judgment, work ethic, sense of fair play, and
devotion to our calling. David’s example of productivity, moral courage,
and high scholarly standards would make him an intimidating colleague
were he not also such a marvelous person. I doubt I could have published
this book without his wisdom.
I tested arguments from part II at various professional conferences. These
included three annual meetings of the American Philosophical Association:
Central Division (Chicago, 2005) and Pacific Division (Pasadena, 2004; San
Francisco, 2005). Other venues for these papers included the Association
for Practical and Professional Ethics (Cincinnati, 2004; San Antonio, 2005),
the Midwest Political Science Association (Chicago, 2004), and the Austra-
lian Society for Legal Philosophy (Melbourne, 2004). I received useful
feedback at these events, particularly from Jim Evans, Jeff Goldsworthy,
and David Lefkowitz. I am especially grateful to my APA commentators:
Larry Alexander, Alan H. Goldman, and Kenneth Einar Himma. Ken’s
combination of astute observations and gushing were just what I needed to
hear. His kindness gave me confidence in the project. So did Larry and
Alan, both of whom I now proudly call friends. By agreeing to serve as
commentators, these two senior scholars drew attention to an unknown
philosopher who was criticizing their work. They gently showed me where
I was wrong and graciously conceded the minor points that I had right.
Despite disagreeing with me, they have encouraged my efforts ever since.
viii Acknowledgments

I am also grateful for several speaking invitations. I was honored to

deliver some material from part I as the 2007 Constitution Day Lecture
at the College of Wooster (Ohio). I thank John Rudisill for suggesting my
name and for his hospitality. In 2008, I discussed role morality with the
Joint Bioethics Colloquium of the National Institutes of Health. Thanks
to Dr. Ezekiel Emanuel, chair of the Department of Bioethics, for inviting
me to lead the session.
Also in 2008, I was invited to present a version of chapter 6 at a con-
ference celebrating twenty years of the Brazilian constitution, held at the
Federal University of Rio de Janeiro. This invitation became a speaking
tour that included Fundação Getulio Vargas Law School, the School of
Federal Magistrates, and Pontific Catholic University Law School (all in
São Paulo), the University of São Paulo Law School, the Rio Branco Insti-
tute, and the Department of Philosophy at the University of Brasilia. Pre-
senting my ideas to Brazilian audiences, whose language I do not speak
and whose legal system I did not understand, was an unforgettable expe-
rience. I owe it all to the initiative and generosity of Noël Struchiner, who
took an interest in my research, and to a U.S. Speaker and Specialist Grant
from the Department of State, Bureau of International Information Pro-
grams. Thanks to Karla Carneiro, Marcos Hirata, and Carla Waehneldt for
welcoming me to their cities.
The philosophy and political science faculties at Vassar College inspired
my love of theory and my curiosity about public affairs but I owe any-
thing I know about practical philosophy to my teachers at the Univer-
sity of Michigan. They included Allan Gibbard, Donald Regan, J. David
Velleman, and especially my dissertation supervisors: Elizabeth Anderson,
Stephen Darwall, and Peter Railton. Without the tutelage of these re-
markable philosophers, I would not even know how to read the texts that
have engaged me since leaving Ann Arbor. I also learned from many grad-
uate students at Michigan, notably Craig Duncan, Christina Frohock,
Nadeem Hussain, Tim Schoettle, Nishi Shah, Chris Roberson, and Michael
Weber.
For my legal education I thank the law faculties at Michigan and Yale
and the lawyers who supervised me at the scrappy Chicago firm known,
when I last worked there, as Abrahamson Vorachek & Mikva.
My editor at Oxford University Press, Peter Ohlin, has always treated
me like a valued author rather than a lapse in judgment. His expertise has
been most useful. I am grateful to several anonymous reviewers, including
two readers of the book proposal for Oxford and one reviewer of the final
manuscript. Their comments greatly improved the book, although some
of their criticisms were too difficult for me to answer.
The George Washington University provided research funds in the
form of a University Facilitating Fellowship and a Junior Scholar Incen-
tive Award. Section 11.7 contains material from my doctoral dissertation,
completed with fellowship support from the Andrew W. Mellon Founda-
tion and the Horace H. Rackham School of Graduate Studies. Thanks to
Acknowledgments ix

the Doris Ulmann Galleries of Berea College for providing access to

the cover painting on affordable terms, and to Tina McCalment and Lisa
Kriner for working fast. Special gratitude goes to my friends from the
Department of Philosophy and Religious Studies at the University of
Wisconsin—Eau Claire, including the late Richard deGrood, for their
faith in me and for fond memories.
I owe the most to my family: my parents, Donald and Gabriella Brand,
to whom I dedicate the book; my beloved brother and friend for life,
Thomas; my stepmother, Catherine; my stepfather, Douglas Peary; and
Rebecca Rosenbaum. My daughter, Maresca, entertained her little broth-
er, Quentin, at critical times with few complaints. Quentin often agreed
to sit on my lap while I typed.
My wife, Laura, reviewed several drafts under great time pressure. She
has my eternal love, gratitude, and admiration—for this and for every-
thing else.
I found a helpful and enthusiastic audience in the Department of Phi-
losophy at the Virginia Polytechnic Institute and State University in
November 2006, a few months before a tragedy occurred on that campus.
A portion of the royalties from the sale of this book will be donated to the
Virginia Tech Hokie Spirit Memorial Fund: http://www.vt.edu/fund/.
I, alone, am responsible for all remaining defects and errors in this book.
No one else should be assumed to agree with any of it.
This page intentionally left blank
Contents

PART I

1 Introduction, 3
2 Practical Reasons and Judicial Use of Force, 19
3 Deviating from Legal Standards, 35
4 The Legal Duties of Judges, 56
5 The Normative Classification of Legal Results, 74
6 Reasons to Deviate, 92
7 Adherence Rules, 111
8 Obeying Adherence Rules, 123
9 The Judicial Oath, 142
10 Legal Duty and Political Obligation, 157

PART II

11 Systemic Effects, 181

12 Agent-Relative Principles, 202
13 Optimal Adherence Rules, 212
14 Guidance Rules, 233
15 Treating Like Cases Alike, 253
16 Implementation, 270
17 Theoretical Implications, 292
18 Conclusion, 308

List of Authorities, 315

Bibliography, 319
Index, 339
This page intentionally left blank
Part I
This page intentionally left blank
1

Introduction

1.1 WHEN JUDGES DISREGARD THE LAW

Mrs. Russell has a prescription from her doctor to use marijuana for
medical purposes. She grows the plant for personal use in her California
home. She is charged with a crime under the federal Controlled Substances
Act.1 Her case comes before a federal district judge. Because California
has legalized medical marijuana, her attorney moves for dismissal of the
indictment, claiming that Congress has no authority to preempt California’s
law.2 However, the U.S. Supreme Court has upheld the application of the
act to marijuana possession in California in a very similar case.3 All
informed lawyers are confident that the judge will deny the motion and
proceed to trial: the law on this issue is settled.
The judge knows of the Supreme Court decision. He knows that as a
trial judge he is expected to follow it. However, he believes that medical
marijuana should be legalized. Therefore, he does something surprising:
he ignores the law and dismisses the indictment.
What should we say about this judge? One thing is clear: his decision
constitutes legal error.4 It could be reversed on appeal.5 Many would also
insist that the judge was wrong to make that decision. His conduct seems
unprofessional at the least, perhaps incompetent. But did he act immor-
ally or unethically?6 Is his decision one that he should not have made, all
things considered? If his action is morally wrong, how wrong is it? Worse

1. This is a hypothetical case, facts based loosely on the unreported case appealed in U.S.
v. Alden, 141 Fed. Appx. 562 (9th Cir. 2005). Attorney General Eric H. Holder, Jr. has since
announced that the federal government will no longer raid legitimate growers of medical
marijuana. David Johnston and Neil A. Lewis, “Ending Raids of Dispensers of Marijuana for
Patients,” New York Times, Mar. 19, 2009, p. A20.
2. Under the “negative Commerce Clause.” U.S. Const., art. I, §8, cl. 18; Oregon Waste
Systems, Inc. v. Department of Environmental Quality, 511 U.S. 93, 95 (1994).
3. Gonzales v. Raich, 545 U.S. 1 (2005).
4. On legal error see, e.g., Harry T. Edwards, “To Err Is Human, But Not Always Harm-
less: When Should Legal Error Be Tolerated?” New York University Law Review 70 (1995):
1167–1213.
5. The prosecution may appeal the dismissal of charges as opposed to an acquittal.
6. I use unethical and immoral interchangeably, as do many philosophers.

3
4 Limits of Legality: Part I

than shoplifting a magazine? Not as bad as pressuring an employee for

sexual favors? Should we publicly denounce the judge for his decision?
Should he regret it? If he were your son, would you be disappointed in
him? These are normative questions, not descriptive or empirical ones.
They are questions about role morality—the moral rights and duties asso-
ciated with a particular social role.7 Asking these questions about judges
can teach us about our conception of the rule of law. Our answers may
have implications for our practice of praising and blaming judges, and for
our standards of judicial selection, oversight, and discipline.
Someone who believes that the federal ban on marijuana, medical or
otherwise, serves important social purposes might readily accuse the judge
in Mrs. Russell’s case of unethical behavior. His decision could weaken the
deterrent effect of the Controlled Substances Act and increase marijuana
usage. Someone who thinks the state should ban marijuana might con-
clude that the judge acted unethically, regardless of whether his decision
was legally correct.
Of course, many people think the federal act should be revised, as
California’s laws have been, to legalize medical uses of marijuana.8 Many
proponents of legalization would be tempted to praise judges who refuse
to enforce the act. Some proponents would yield to the temptation.
Others would resist it. If a legalization proponent accepts a strong princi-
ple of judicial fidelity to law, then even he would agree, perhaps reluc-
tantly, that the judge was wrong to dismiss the charges against Mrs. Russell.
In Anglo-American legal systems, fidelity entails that inferior courts must
follow superior court rulings on matters of law. It also entails many other
basic norms of adjudication: trial judges must take all admissible evidence
into account; judges must apply recognized sources of law, such as consti-
tutions, legislation, and common-law rules; judges must obey writs of
mandamus; courts should pay some deference to their own precedents;
et cetera. The conviction that judges are “bound by the law” is very com-
mon among lawyers, judges, legal scholars, and members of the general
public. In the words of one academic lawyer, “Jurists and legal scholars
often seem to regard the judicial obligation to follow statutes as too obvi-
ous to require justification.”9 One of the most severe accusations that one

7. See, e.g., Arthur Isak Applbaum, Ethics for Adversaries: The Morality of Roles in
Public and Professional Life (Princeton, N.J.: Princeton University Press, 1999); Michael O.
Hardimon, “Role Obligations,” Journal of Philosophy 91 (1994): 333–63; David Luban,
Lawyers and Justice: An Ethical Study (Princeton, N.J.: Princeton University Press, 1988), ch.
6; Alan H. Goldman, The Moral Foundations of Professional Ethics (Totowa, N.J.: Rowman
and Littlefield, 1980).
8. Of course, many support the legalization of marijuana, full stop.
9. Steven D. Smith, “Why Should Courts Obey the Law?” Georgetown Law Journal 77
(1988): 113–64, p. 114. See also M. B. E. Smith, “May Judges Ever Nullify the Law?” Notre
Dame Law Review 74 (1999): 1657–71, p. 1658 (“[The view that] judges are always legally
and morally bound to follow the law is undoubtedly the conventional wisdom among legal
scholars and laypeople alike”).
Introduction 5

can make against a public official, especially a judge, is that he has not
upheld the law in his official capacity. Nevertheless, judges sometimes
depart from the law,10 disregarding either admissible evidence presented
or controlling sources of law such as statutes, constitutions, and precedent.11
When a decision is reversed on appeal, for example, either the lower court
or the reversing court has made some kind of legal mistake.12
Judges reach legally incorrect results for various reasons. Some reasons,
such as negligence and incompetence, are not especially interesting. In the
interesting cases judges willfully disregard the law, although they rarely
admit it.13 Corrupt judges seek to profit financially from their decisions, to
reward friends or family members, or to settle personal scores. Acting for
these reasons violates codes of judicial conduct.14 It is also morally wrong
and blameworthy. Judges are ethically and legally bound to decline bribes.
They are bound—ethically, at least—to recuse themselves from cases pre-
senting conflicts of interest. This book largely ignores judges who deviate
from the law either inadvertently or for self-serving reasons. I address
cases in which judges disapprove of the legally required results as a matter
of morality or public policy.15 They face what Robert Cover calls “moral-
formal” dilemmas: conflicts between the law and their moral convictions.16
Sometimes they decide to misrepresent or disregard the law—a practice
known as judicial nullification.17

10. “[J]udges . . . sometimes bend or break the rules for the sake of other values, such
violations being in fact rather common because detection and sanctioning are difficult.”
Richard A. Posner, “What Do Judges and Justices Maximize? (The Same Thing Everybody
Else Does),” Supreme Court Economic Review 3 (1993): 1–41. See also M. B. E. Smith, “Do
Appellate Courts Regularly Cheat?” Criminal Justice Ethics 16 (1997): 11–20 (arguing that
they do). Only someone who denies that the law has any determinate content could deny
that judges sometimes get it wrong. See §5.5.1.
11. An analogous phenomenon in administrative law is called nonacquiesence—when
agencies refuse to follow federal appellate courts. See Samuel Estreicher and Richard
L. Revesz, “Nonacquiescence by Federal Administrative Agencies,” Yale Law Journal 98
(1989): 679–772.
12. Unless the law has changed in the interim at the hands of the legislature, a higher
court, or the reversing court itself.
13. But see examples in §3.12.
14. See chapter 4.
15. Various sources of law—public and private—figure in my examples throughout this
book. Common law, statutory law, and constitutional law differ in profound ways. There is
obviously a world of difference between a lawsuit for breach of contract with $500 in con-
troversy and a constitutional challenge to the death penalty before the Supreme Court of
the United States, although both are cases in courts of law. My arguments operate at a level
of abstraction that encompasses both. In practice, different sources of law have different
functions, so courts should treat them differently.
16. Robert M. Cover, Justice Accused: Antislavery and the Judicial Process (New Haven,
Conn.: Yale University Press, 1975), pp. 197–225.
17. See, e.g., Smith, “May Judges Ever Nullify the Law?”; Jack B. Weinstein, Comments
on Jury Nullification: Proceedings of the Fifty-Third Judicial Conference of the District of
Columbia Circuit, 145 F.R.D. 149, 170 (1993) (commenting that “trial judges are, I suppose,
6 Limits of Legality: Part I

Politicians and commentators often accuse judges of misunderstand-

ing or ignoring the laws they have sworn to uphold.18 Such accusations
are especially common in high-profile constitutional cases. Few are
better known than the 2000 case of Bush v. Gore, in which the U.S.
Supreme Court overruled the Florida Supreme Court and halted the
presidential recount under the Equal Protection Clause of the Four-
teenth Amendment.19 Many lawyers condemn that majority opinion as
lawless.20
Another recent example is Newdow v. U.S. Congress, in which the
United States Court of Appeals for the Ninth Circuit ruled that inviting
public school students to recite the Pledge of Allegiance violates the
Establishment Clause of the First Amendment, as the Pledge contains
the words “under God.”21 Senator Tom Daschle (D-S.D.) derided the
decision as “just nuts,”22 a sentiment also expressed by many of his fellow
legislators.

nullifying the guidelines”); Smith, “Why Should Courts Obey the Law?” p. 113; Kent
Greenawalt, Conflicts of Law and Morality (New York: Oxford University Press, 1987), pp.
367–68; Samuel Estreicher, “Judicial Nullification: Guido Calabresi’s Uncommon Common
Law for a Statutory Age,” New York University Law Review 57 (1982): 1126–73. Compare
jury nullification (jurors disregarding their instructions).
18. See, e.g., Neil S. Siegel, “Why President Bush Should Not Take the 5th; Judges Who
Ignore Law are Possible Court Candidates,” Houston Chronicle, Jun. 17, 2005, p. B11; Thomas
Sowell, “Real Judicial Crisis Is Judges Who Ignore the Law,” The Post and Courier (Charleston,
S.C.), Jan. 14, 1998, p. A11; Orrin Hatch and Sam Brownback, “‘Extreme’ Judicial Activism,”
Washington Times, Feb. 10, 2005, p. A19 (condemning U.S. District Judge Gary Lancaster for
“ignor[ing] the law in favor of [his] own agenda”); Stephen B. Bright, “Let’s Try Brian Nichols
Properly the First Time,” Atlanta Journal-Constitution, Nov. 7, 2007, p. 19A (“Like it or not,
agree or disagree, trial judges must follow the law”); Rachel Graves, “The Terri Schiavo Case;
Schiavo Dies, but Debate Lives; DeLay Insists Judges Must ‘Answer for their Behavior,’”
Houston Chronicle, Apr. 1, 2005, p. A1 (in 1997, Rep. Tom DeLay publicly contemplated an
impeachment drive against judges whose decisions “ignored existing laws”).
19. Bush v. Gore, 531 U.S. 98 (2000).
20. See, e.g., Ward Farnsworth, “‘To Do a Great Right, Do a Little Wrong’: A
User’s Guide to Judicial Lawlessness,” Minnesota Law Review 86 (2001): 227–66; Alan
M. Dershowitz, Supreme Injustice: How the High Court Hijacked Election 2000 (Oxford:
Oxford University Press, 2001); Cass R. Sunstein and Richard A. Epstein, eds., The Vote:
Bush, Gore, and the Supreme Court (Chicago: University of Chicago Press, 2001); Michael J.
Klarman, “Bush v. Gore through the Lens of Constitutional History,” California Law Review
89 (2001): 1721–65; “Symposium: Bush v. Gore,” University of Chicago Law Review 68
(2001): 613–791; Bruce Ackerman, “Anatomy of a Constitutional Coup,” London Review of
Books, Feb. 8, 2001, p. 3; Ronald Dworkin, “A Badly Flawed Election,” New York Review of
Books, Feb. 8, 2001, p. 1; Jack M. Balkin, “Bush v. Gore and the Boundary between Law and
Politics,” Yale Law Journal 110 (2001): 1407–58.
21. Newdow v. U.S. Congress, 292 F.2d 597 (9th Cir. 2002), amended by Newdow v. U.S.
Congress, 328 F.3d 466 (9th Cir. 2003), reversed by Elk Grove Unified School District v. New-
dow 542 U.S. 1 (2004).
22. Carl Hulse, “Lawmakers Vow to Fight Judges’ Ruling on the Pledge,” New York
Times, June 27, 2002, p. A20.
Introduction 7

In the 2005 case of Kelo v. City of New London, the U.S. Supreme Court
upheld municipal authority to demolish private residences for the pur-
pose of largely private redevelopment projects.23 Legal scholar Richard A.
Epstein accused the Court of “refusing to look closely at past precedent
and constitutional logic.”24 Senator F. James Sensenbrenner, Jr. (R-Wis.),
called the ruling “the Dred Scott decision of the 21st century.”25
In addition to criticism, judges accused of disregarding the law are in-
creasingly threatened with reprisals: requests for resignation and threats
of impeachment.26 In 1996, United States District Judge Harold Baer, Jr.,
suppressed the results of a vehicle search that yielded eighty pounds of
cocaine and heroin. The alleged buyers had run off when approached by
police officers. Supreme Court precedent at the time held that such eva-
sive behavior generates the “reasonable suspicion” needed for a search
under the Fourth Amendment.27 However, Judge Baer asserted that run-
ning from police does not necessarily generate reasonable suspicion in
neighborhoods where police are widely distrusted.28 The judge was
accused of either misunderstanding the law or knowingly ignoring it. Over
200 members of the U.S. House of Representatives called for his resigna-
tion.29 The Clinton White House stated that it disagreed with Judge Baer’s
ruling and had not ruled out asking him to resign on account of it.30 Sen-
ate Majority Leader Robert Dole (R-Kan.) proposed impeachment.31 The
judge reversed himself.32
I shall not address whether any of these court decisions was, in fact, a
legal mistake. Reaching that issue would require a full legal analysis of
each case. These are just examples of decisions that reasonable people
might consider lawless. No one denies that judges sometimes fail to apply

23. Kelo v. City of New London, 545 U.S. 469 (2005).

24. Richard A. Epstein, “Blind Justices: The Scandal of Kelo v. New London,” Wall Street
Journal, July 3, 2005.
25. David Lightman, “Lawmakers Stand up to Court; Bipartisan Coalition Hopes to
Dilute Impact of Eminent Domain Ruling,” Hartford Courant, July 1, 2005, p. A1. Dred Scott
is the notorious ruling that returned an escaped slave to his master. Dred Scott v. Sandford,
60 U.S. 393 (1857).
26. Elected judges may, of course, lose reelection. In recent years there have also been a
handful of disciplinary proceedings against judges for departing from the law. See chapter 4.
27. United States v. Sokolow, 490 U.S. 1, 8–9 (1989); Florida v. Rodriguez, 469 U.S. 1, 6
(1984) (per curiam); United States v. Brignoni-Ponce, 422 U.S. 873, 885 (1975).
28. United States v. Bayless, 913 F. Supp. 232 (S.D.N.Y. 1996). In a subsequent case the
Supreme Court confirmed that sudden flight creates reasonable suspicion even in high-
crime neighborhoods. Illinois v. Wardlow, 528 U.S. 119 (2000). The fact that the Court even
granted certiorari on this issue might, however, suggest that Judge Baer’s decision was not as
obviously lawless as his critics claimed.
29. Jon O. Newman, “The Judge Baer Controversy,” Judicature 80 (1997): 156–64, p. 156.
30. Allison Mitchell, “Clinton Pressing Judge to Relent,” New York Times, Mar. 22, 1996,
p. A1.
31. Katharine Q. Seelye, “A Get-Tough Message at California’s Death Row,” New York
Times, Mar. 24, 1996, p. 29.
32. United States v. Bayless, 921 F. Supp. 211 (S.D.N.Y. 1996).
8 Limits of Legality: Part I

the law correctly. My question is this: when, if ever, is it ethically permis-

sible for judges knowingly to promote their own values and policy prefer-
ences at the expense of the law?
Many commentators consider decisions that contradict the law to be
ethically impermissible by definition. For them “what the law requires” is
synonymous with “how the judge ought to decide.” Ronald Dworkin sug-
gests that people do not generally accept this as a definitional truth:
Our lawyers and citizens recognize a difference between the question what the
law is and the question whether judges or any other official or citizen should
enforce or obey the law. They regard these as separate questions, not only when
they have in mind foreign, wicked legal systems . . . but even in considering
how citizens and officials in our own communities should behave. The opinion
that our judges should sometimes ignore the law and try to replace it with bet-
ter law is far from a stranger to law school classrooms and even political
debates.33

Dworkin may be correct that people see two distinct questions, as a

conceptual matter, but many believe that judges are nevertheless obli-
gated to enforce the law even when the results are morally objectionable.

1.2 CLEAR CASES AND UNCLEAR CASES

Lawyers have different opinions regarding how much latitude judges

should have to avoid morally objectionable results. We must consider
cases in which the law is clear as well as those in which it is unclear.
Regarding cases in which the law is clear, the most restrictive position is
that judges must apply the law correctly, regardless of how unjust the
legally required result may be. The least restrictive position is that judges
have no moral obligation to reach legally mandated results, even when the
law is clear.
Regarding legally unclear cases, the most restrictive position is that
judges must endeavor to apply the law correctly, even when it is not clear
what the law requires. According to this position, judges must never allow
extralegal considerations to influence their decisions, even when the law is
unclear.34 The least restrictive position always allows judges to consult
extralegal considerations when deciding unclear cases.35
Many writers defend unrestrictive positions with respect to unclear
cases. Disregarding clear legal mandates is much more controversial, but
it has advocates: the natural law tradition is often cited for the principle

33. Ronald Dworkin, Law’s Empire (Cambridge, Mass.: Harvard University Press, 1986),
p. 109.
34. See, e.g., Stephen J. Burton, Judging in Good Faith (Cambridge: Cambridge Univer-
sity Press, 1992), pp. 243–44.
35. See, e.g., H. L. A. Hart, The Concept of Law, 2nd ed. (Oxford: Oxford University
Press, 1994), p. 153.
Introduction 9

that unjust laws do not “bind in conscience.”36 A weaker version holds that
courts are not morally obligated to enforce extremely unjust laws: those
that require returning escaped slaves to their masters, stoning women for
adultery, confiscating property because of the owner’s religious beliefs, or
executing the innocent without due process of law. Even those who reject
other precepts of natural law theory might agree that judges are morally
permitted to disregard such laws.
In the past century, much has been written about judges who uphold
the laws of extremely unjust systems. Robert Cover’s classic, Justice
Accused, takes on nineteen-century judges whose decisions facilitated
slavery in the United States.37 Cases from that era continue to generate
important scholarship.38 David Dyzenhaus’s Hard Cases in Wicked Legal
Systems addresses judges in apartheid South Africa.39 Equally relevant are
discussions of adjudication under fascism, beginning with Gustav Rad-
bruch’s accusation that the acceptance of legal positivism by German of-
ficials encouraged them to comply blindly with the worst Nazi laws.40
This issue was taken up by H. L. A. Hart and Lon Fuller in their famous
Harvard Law Review debate.41
The aforementioned literature has drama aplenty. However, the resi-
dents of modern democratic republics could easily conclude that these
debates have little relevance to their own legal systems, in which the law
does not promote such horrors as genocide, slavery, racial apartheid, or
religious persecution. Some will suspect that questions about the ethics of
lawless judging become interesting only under conditions of extreme injus-
tice. They may assume that under happier conditions, lawless judging is
self-evidently unethical, end of story. Few commentators assert that judges
are morally permitted to disregard laws that require only moderate, rather
than extreme, injustices. This includes commentators who would liberate
judges to disregard extremely unjust laws and to base their decisions on
extralegal considerations when the law is unclear. As Dworkin notes,

36. This principle does not entail the further claim that unjust legislation does not even
constitute genuine law. I need not consider whether canonical natural lawyers believed the
further claim. For discussion see Mark C. Murphy, “Natural Law Jurisprudence,” Legal Theory
9 (2003): 241–67, pp. 244–45.
37. Cover, Justice Accused.
38. See, e.g., Mark A. Graber, Dred Scott and the Problem of Constitutional Evil
(Cambridge: Cambridge University Press, 2006).
39. David Dyzenhaus, Hard Cases in Wicked Legal Systems: South African Law in the
Perspective of Legal Philosophy (Oxford: Clarendon Press, 1991).
40. Gustav Radbruch, “Gesetzliches Unrecht und übergesetzliches Recht,” Süddeutsche
Juristen-Zeitung 1 (1946): 105–8. Translated by Bonnie Litschewski Paulson and Stanley
L. Paulson, “Statutory Lawlessness and Supra-Statutory Law,” Oxford Journal of Legal Studies
26 (2006): 1–11.
41. H. L. A. Hart, “Positivism and the Separation of Law and Morals,” Harvard Law
Review 71 (1958): 593–629; Lon L. Fuller, “Positivism and Fidelity to Law: A Reply to
Professor Hart,” Harvard Law Review 71 (1958): 630–72.
10 Limits of Legality: Part I

We disagree about the exact force law has in certain special circumstances,
when there are strong competing considerations of justice. We disagree, perhaps,
about what the judges in Massachusetts who were asked to enforce the Fugi-
tive Slave Act before the American Civil War should have done. But we share
a general, unspecified opinion about the force of law when such special consid-
erations of justice are not present, when people disagree about the justice or
wisdom of legislation, for example, but no one really thinks the law wicked or
its authors tyrants. Our different convictions about the force of law unite in
such cases. We think the law should be obeyed and enforced . . . .42

I agree with Dworkin that “we” think this but I shall argue that “we” are
mistaken. I think interesting questions about the ethics of lawless judging
arise even in reasonably just legal systems. I shall argue that judges are
sometimes ethically permitted to deviate from the law in order to avoid
results that are only moderately unjust. Writers have not developed general
theories of judicial lawlessness that apply to reasonably just systems.43

1.3 ARGUMENTS FOR FIDELITY

Consider again the conclusion that the judge acted wrongly in Mrs. Russell’s
marijuana possession case. This conclusion could be supported with state-
ments such as the following:
1. The Supreme Court has already upheld the application of the
Controlled Substances Act to cases like this. The justices may have
been wrong to do so, but they outrank district judges.
2. Perhaps the act should be revised, but that function belongs to
Congress, not to judges.
3. We have separation of powers in the United States. Legislators
legislate, and judges adjudicate.
4. Judges are supposed to apply the law, not make it.
5. The judge’s job is to follow the law. He is paid to do it.
6. The judge swore an oath to uphold the law whether he likes it or not.
7. Everyone expects the judge to follow the law. We are all counting
on him.
8. A judge is supposed to act as a professional. The ethics of his
profession—judicial ethics—require him to apply the law.

42. Dworkin, Law’s Empire, p. 111. I do not mean to imply that Dworkin, himself,
believes that judges should ever appeal to extralegal considerations.
43. There is, however, one substantial body of literature on modern judges facing laws
to which they are conscientiously opposed. This literature discusses judges, notably Roman
Catholics, whose religious convictions conflict with the law. Some of this literature is useful
for my purposes but it focuses on when a judge should recuse himself rather than applying
laws to which he has religious objections. The literature on religious judges rarely mentions
departing from the law as an acceptable option. See §6.11.
Introduction 11

9. If judges refuse to apply laws of which they disapprove, then we

will have anarchy.
10. By failing to apply the law, the judge damages the rule of law—one
of our basic political values.
To most readers, especially lawyers, these ten observations will seem
like common sense, hardly worth stating. In this book I shall contend that
the corresponding arguments are not nearly as conclusive as they are
thought to be. It would be misleading, but not literally inaccurate, to state
my thesis as follows: judges in courts of law are not ethically obligated to
apply the law correctly when they decide legal issues. This statement
would be misleading because it could imply that judges are ethically
licensed to misapply the law as often as they wish, and for any reason
whatsoever: laziness, personal profit, dislike for a party, et cetera. As stated,
my thesis seems to license irresponsible or corrupt judging. It seems to
advocate the destruction of the rule of law and regression to a “govern-
ment of men” (sic) rather than laws. It appears to commend the annihila-
tion of law itself, perhaps even a world without order.
These are not my intentions. Once the terms of my thesis are carefully
defined it has none of these implications. Even properly understood, how-
ever, it remains an unorthodox thesis with some surprising lessons.

1.4 MODIFYING PRECEDENT

I hope you are reassured that my thesis will not be utterly outrageous. It
is, however, unusual. It should not be confused with some commonplace
claims about legal precedent. Courts often modify their own precedents
and those of courts at the same level in the judiciary hierarchy—horizontal
precedents—including doctrines of the common law and prior interpreta-
tions of statutes and constitutions.44 Mainstream legal scholars agree that
judges are legally permitted to do so.45 Understanding the scope of my
thesis requires distinguishing between the law simpliciter and particular
legal standards. If a judge has legal authority to modify a horizontal prec-
edent before applying it, then he does not deviate from the law simpliciter
in so doing. Nor does modifying horizontal precedent necessarily consti-
tute a failure of judicial craft, much less an ethical failure. In fact, such
modifications are often seen as excellent judging and as ethically sound.
Commentators merely disagree about the conditions under which judges
should modify horizontal precedent.46

44. Lewis A. Kornhauser, “Adjudication by a Resource-Constrained Team: Hierarchy

and Precedent in a Judicial System,” Southern California Law Review 68 (1995): 1605–29, p.
1608 (distinguishing horizontal and vertical precedents).
45. See many scholars cited in Larry Alexander, “Constrained by Precedent,” Southern
California Law Review 63 (1989): 1–64.
46. A few writers believe that judges should never modify precedents. Although I dis-
agree with them, my objective in this book is not to challenge them directly. I am addressing
12 Limits of Legality: Part I

This book may, incidentally, reinforce the common view that judges
are morally permitted to modify horizontal precedent, but my central
topic and thesis are much broader. I am interested in decisions in which
courts depart from the law simpliciter—cases of deviation from the legally
correct result as determined by all applicable legal standards. These are
cases in which the judge either lacks legal authority to modify the appli-
cable standards or else declines to exercise that authority, if he has it,
while still departing from the law. The trial judge who dismisses the
charges against Mrs. Russell contradicts appellate precedent—vertical
precedent—without distinguishing the case.
The conventional wisdom is that such nonacquiescence is wrong, but
even state courts occasionally disregard vertical precedent47 or entertain
the possibility of doing so. In 2005, Justice Tom Parker of the Alabama
Supreme Court wrote an extraordinary open letter to his fellow justices.48
He urged them to defy a recent decision of the United States Supreme
Court that banned the death penalty for murderers who were younger
than eighteen when they committed the crime.49 Parker’s extremely un-
usual choice to openly advocate defiance was condemned even by some
who agreed with him about the death penality.50 Decisions that deviate
from vertical precedent are almost universally regarded as unlawful.51 My
question is, when are they also unethical?

1.5 WHY DEFEND JUDICIAL DEVIATION FROM THE LAW?

Instead of questioning the judicial duty to follow the law, the standard
response of lawyers who disapprove of a result is to argue that the judge
misunderstood or misapplied the law. The language in judicial opinions often
implies that the law left the judge no choice but to reach a certain result.52
The critic claims that the law was less clear or less rigid than the judge

the mainstream. See, e.g., Jeremy Bentham, A Fragment on Government, eds. J. H. Burns and
H. L. A. Hart (Cambridge: Cambridge University Press, 1988), pp. 19–20. Even some writers
who believe that courts must treat precedents as rules agree that they may sometimes mod-
ify them. Alexander, “Constrained by Precedent,” pp. 51–52.
47. One writer has recently urged us to notice that “[e]very so often . . . state courts
actively disregard binding Supreme Court precedent,” notwithstanding the conventional
wisdom that such deviation is utterly inappropriate. Frederic M. Bloom, “State Courts
Unbound,” Cornell Law Review 93 (2008): 501–54, p. 503.
48. Tom Parker, “Alabama Justices Surrender to Judicial Activism,” Birmingham (Ala.)
News, Jan. 1, 2005, p. 4B.
49. Ibid. The decision criticized is Roper v. Simmons, 543 U.S. 551 (2005).
50. Phillip Rawls, “Parker ‘Attack’ Irks Fellow State Justice,” Huntsville (Ala.) Times, Jan.
16, 2006, p. 2B.
51. Bloom, “State Courts Unbound,” pp. 502–3.
52. In the words of Judge Wald, “while judges still typically write as if they were abso-
lutely certain about the rightness and soundness of their analysis and decisions, everyone
(including the judges) knows that’s not necessarily the case.” Patricia M. Wald, “The Rhetoric
Introduction 13

represented it to be. She argues that in such unclear cases judges are legally
and ethically permitted to invoke extralegal considerations—considerations
of morality, fairness, justice, and public policy. In this case, of course, these
considerations support her favored result. She insists that it was permissi-
ble, both legally and ethically, for the judge to reach that result.
Many people appear to believe that the law always, or almost always,
allows the court to reach results of which they approve, either because it
clearly favors those results or because it is unclear enough to permit them.
They will think it unnecessary and unnecessarily difficult to defend, as I
do, the moral right of judges to disregard the law. I have no quarrel with
writers who take the more cautious approach, defending the permissibil-
ity of invoking extralegal considerations in legally unclear cases without
insisting that judges have the right to disregard clear legal mandates. How-
ever, I choose to defend the more difficult position—that judges some-
times have the moral right, and moral reasons, to disregard clear legal
mandates, and not only when the law is extremely unjust. I have several
reasons for defending this unusual position. My first reason is simply that
I think there is a plausible argument for my conclusion. This argument is
quite interesting whether or not it serves any practical purpose.
My second reason is that the more difficult position is also the more
logically fundamental. If judges have the right to disregard clear legal
mandates, then it stands to reason that they also have the right to invoke
extralegal considerations when the law is unclear. Defending departures
from clear laws is like running a marathon. After that, defending the use
of extralegal considerations in unclear cases is like running a mile.
My final reason for defending this unorthodox position is that I think the
writers who want judges to avoid unjust results put themselves at a disad-
vantage when they accept, uncritically, the premise that judges must follow
the law when it is clear. Many of the arguments for a duty to follow clear
laws can be adapted to support the claim that, whenever possible, sources
of legal authority should be read as containing clear mandates. By conceding
a judicial duty to follow clear laws, one actually makes it harder to defend
the right of judges to invoke extralegal considerations in unclear cases.
The argument goes differently if we accept the challenge of defending
the moral right of judges to deviate even from the clearest of laws. If judges
have this right, then we need not argue that the law is unclear in order to
maintain that judges are sometimes permitted to reach results that appear to
contradict the law. The law could be entirely clear and yet good results could

of Results and the Results of Rhetoric: Judicial Writings,” University of Chicago Law Review
62 (1995): 1371–419, p. 1417. Justice Cardozo observes that judges who innovate tend “to
disguise the innovation even from themselves, and to announce in all sincerity that it was all
as it had been before.” Benjamin N. Cardozo, “Jurisprudence,” in Selected Writings of Benjamin
Nathan Cardozo, ed. Margaret E. Hall (New York: Fallon, 1947), p. 37. See generally, Dan
Simon, “The Double-Consciousness of Judging: The Problematic Legacy of Cardozo,” Oregon
Law Review 79 (2000): 1033–80.
14 Limits of Legality: Part I

still be reached. The permissibility of reaching good results would no longer

be held hostage to debates about how much latitude the law affords.53

1.6 OVERVIEW

Part I of this book sets up a problem. I argue that judges have moral rea-
sons—albeit defeasible ones—to deviate from the law whenever it requires
them to reach results that are otherwise objectionable on moral or policy
grounds. Next I ask, are judges nevertheless morally obligated to apply the
law correctly in such cases and, if so, why? I consider several affirmative
answers that depend upon what I call the undermining principle, which
holds that if the law requires a public official to perform a certain action,
then any moral reasons that she would otherwise have not to perform
that action are undermined. I consider every argument I know for the
undermining principle. Concluding that none is sound, I reject it, although
I cannot refute it. Then I consider several important arguments for the
conclusion that judges have moral reasons to reach legally correct results:
arguments from the judicial oath and political obligation. I find that these
arguments fail to support the conclusion that judges have even a defeasi-
ble moral reason to reach legally correct results in all cases.
Most readers, however, will be confident that judges are morally obligated
to apply the law correctly in at least some cases in which it requires morally
objectionable results. Such readers should share my motivation to find a
better argument for this conclusion. In part II, I present a different argument
for the principle that judges have moral reasons to apply the law correctly.
In at least some cases in which the law requires objectionable results, these
reasons may be strong enough to outweigh reasons to deviate from the law.
Although my argument in part II incorporates many familiar elements,
it is unusual and likely to be controversial. It has several implications that
distinguish it from the arguments rejected in part I. The rejected argu-
ments claim that judges have agent-relative reasons to reach legally correct
results.54 These are reasons that apply differently to different agents. They
contrast with agent-neutral reasons, which apply to all agents.55 I argue that

53. See section 5.5.1.

54. Also known as agent-centered reasons. Stephen L. Darwall, “Agent-Centered Restric-
tions from the Inside Out,” Philosophical Studies 50 (1986): 291–319.
55. Formally defined, an agent-relative reason is “one that cannot be fully specified
without pronominal back-reference to the person for whom it is a reason,” whereas an
agent-neutral reason is “one that can be fully specified without such an indexical device.”
Philip Pettit, “Universality without Utilitarianism,” Mind 72 (1987): 74–82, p. 75. Classic
treatments of agent-relativity include Samuel Scheffler, The Rejection of Consequentialism,
revised ed. (Oxford: Oxford University Press, 1994); Thomas Nagel, The View from Nowhere
(Oxford: Oxford University Press, 1986); Derek Parfit, Reasons and Persons (Oxford:
Clarendon Press, 1984). See the excellent discussion in Michael Ridge, “Reasons for Action:
Agent-Neutral vs. Agent-Relative,” http://plato.stanford.edu/entries/reasons-agent/.
Introduction 15

judges have agent-neutral reasons to adhere to the law. Unlike the reasons
considered in part I, these reasons apply even when the law requires objec-
tionable results. However, I argue that these reasons can be outweighed by
reasons to deviate from the law. My argument puts pressure on the popular
view that judges in reasonably just legal systems have all-things-considered
reasons to adhere to the law in all cases. My argument is consistent with
that claim, if certain empirical conditions are satisfied, but I do not believe
that they are satisfied today in realistic legal systems such as the United
States. I also argue that, because judges have only agent-neutral reasons to
adhere when the law requires objectionable results, certain factors that are
universally considered irrelevant are, in fact, relevant to the overall permis-
sibility of deviation from the law in particular cases. These factors concern,
for example, the overall health of the rule of law in the system.
My study has both interpretive and revisionist dimensions. I aim to
interpret accurately some central features of the practice of adjudication
as it actually occurs in modern legal systems. However, I also urge sub-
stantial revision of some hoary ideas about adjudication that remain pop-
ular with lawyers and theorists alike. I conclude that deviation from the
law may be justified more often than others have claimed. At the least, I
hope to persuade readers that my theory has as many virtues as more
traditional theories.
Those who like labels might say that my theory combines elements of
legal formalism, philosophical anarchism, sophisticated consequentialism,
and natural law theories of adjudication.

1.7 RELATED DISCOURSES

My project lies at the intersection of several theoretical discourses, two of

which merit special mention. First is the ongoing debate concerning the
“problem of political obligation.” Dozens of books and articles on this
topic have appeared since the early 1970s.56 However, most focus on the
citizen’s duty to obey the law, often in the context of civil disobedience,
conscientious objection, or political revolution. Less often have they
addressed the public official’s right to enforce the law, his professional
duty to enforce it, or his possible right to disregard it. By contrast, authors
who concentrate on adjudication have rarely examined it through the lens
of political obligation.
The book also intersects with scholarship on rule-based adjudication.57
Most of this work answers one or more of the following questions. First,

56. For an excellent survey see William A. Edmundson, “State of the Art: The Duty to
Obey the Law,” Legal Theory 10 (2004): 215–59.
57. See, e.g., Alan H. Goldman, Practical Rules: When We Need Them and When We
Don’t (Cambridge: Cambridge University Press, 2002); Larry Alexander and Emily Sherwin,
The Rule of Rules (Durham, N.C.: Duke University Press, 2001); Frederick Schauer, Playing
16 Limits of Legality: Part I

how should judges interpret legal standards that are unclear or indetermi-
nate? Second, how should judges decide cases to which no valid legal
standard applies (the question of judicial discretion)? Third, under what
conditions should judges overrule existing legal standards? Although my
arguments have implications for each of these questions, I concentrate on
a fourth: when, if ever, may judges deviate from a legal standard without
overruling it or making a legally justified exception? More precisely, how
should they decide cases in which unjust results are dictated by determi-
nate legal standards that should not be overruled? Even the laws of rea-
sonably just societies inflict injustices. As writers on equity have recognized
since Aristotle, even a law that is itself just can mandate unjust results in
particular cases. I ask, inter alia, if judges are morally permitted to deviate
from the law in such cases, which arise in all realistic legal systems.58

1.8 JUDICIAL ACTIVISM AND THE POLITICS

OF LEGAL REASONING

My project will strike some readers as a labyrinthine defense of judicial

activism. I may seem coy in avoiding this terminology so I should explain
my choice up front. If judicial activism means “legally unauthorized depar-
ture from the law as it is properly understood,” then I do, indeed, present
an ethical defense of limited judicial activism. That confession will prompt
some readers to assume that this book is yet another attempt to defend
politically progressive judges and disparage conservatives. However,

by the Rules: A Philosophical Examination of Rule-Based Decision-Making in Law and in Life

(Oxford: Oxford University Press, 1991).
58. My project is also analogous to some well-known studies of legal ethics. Lawyers
often justify otherwise immoral actions undertaken in the course of zealous advocacy by
appealing to the “adversary system excuse.” They use the excuse to defend their choice to
advocate zealously as morally permissible, if not required. The excuse has been vigorously
challenged by David Luban, William Simon, and others. Sitting judges, of course, do not
have clients. It is the hallmark of the judicial role that the judge does not represent any party
and remains impartial between them all. But one might claim that judges are obligated
“zealously to represent the law,” even when the law requires morally objectionable results.
One could claim that, much as lawyers are permitted, perhaps even required, to take certain
otherwise immoral actions that serve their clients’ interests, judges are required to take
certain otherwise immoral actions when the law requires it. This is a judicial version of the
adversary system excuse, which we might call the “judicial system requirement.” I shall not
suggest that legal ethicists have watertight arguments against the adversary system excuse,
but these arguments have compelled replies and remain serious contenders in the literature.
This book similarly examines, and presents ethical challenges to, the judicial system require-
ment. See Luban, Lawyers and Justice; William H. Simon, The Practice of Justice: A Theory of
Lawyers’ Ethics (Cambridge, Mass.: Harvard University Press, 1998); Gerald J. Postema,
“Moral Responsibility in Professional Ethics,” New York University Law Review 55 (1980):
63–89, p. 78; Richard A. Wasserstrom, “Lawyers as Professionals: Some Moral Issues,” Hu-
man Rights 5 (1975): 2–15.
Introduction 17

I believe that my theses are compatible with virtually any position on

substantive issues of public policy, from far left, to moderate, to far right.
I advance a distinctive understanding of how judges are ethically permit-
ted to promote sound policy objectives, whatever these may be. Later I
will say more about how my claims relate to various definitions of judicial
activism.59
My project speaks directly to the relationship between politics and
legal reasoning. Barry Friedman offers the following observation, discuss-
ing the state of constitutional theory in 2004:
When the ideological valence of Supreme Court decisions shifts, constitutional
theorizing about judicial review tends to shift as well. Over the last century or
more there have been two general positions taken about judicial review: that it
is a blight in a democratic system that must be curtailed, and that it is a valued
part of U.S. government essential to the protection of constitutional liberty.
One is a critique, the other a justification. Progressives and conservatives have
advanced both positions (in various permutations) at different times, depend-
ing upon which position seemed most apt to present circumstances, given their
political views. . . . [P]rogressive and conservative positions are shifting again at
this very moment. . . . Progressives at the turn of the twenty-first century are
echoing criticisms offered by progressives one hundred years earlier, though
progressives took a more positive position toward judicial review during the
Warren Court.60

Friedman discusses opinions about the legitimacy of judicial review,

but opinions about adjudication in general are equally susceptible to shifts
based on the political composition of the judiciary. Between 1937 and
1980, federal courts rendered many decisions that pleased political pro-
gressives. Conservatives accused the courts of deviating from the law: giv-
ing insufficient weight to precedent, misreading legal sources such as
statutes and constitutions, et cetera. Since then, the federal courts have
started rendering more decisions that please conservatives.61 Liberals have
increasingly taken on the role of accusing the courts of deviating from the

59. See §3.11.

60. Barry Friedman, “The Cycles of Constitutional Theory,” Law and Contemporary
Problems 67 (2004): 149–74, pp. 149–50.
61. A few examples from the U.S. Supreme Court include United States v. Lopez, 514
U.S. 549 (1995) (holding that Gun-Free School Zones Act exceeds federal commerce
power); United States v. Morrison, 529 U.S. 598 (2000) (holding that Violence Against
Women Act exceeds federal commerce power); Boy Scouts of America v. Dale, 530 U.S. 640
(2000) (holding that prohibition on discrimination on basis of sexual orientation in New
Jersey public accommodations law, as applied to Boy Scouts of America, violates right of
expressive association under First Amendment); Bush v. Gore (holding that Florida’s manual
vote recount process in 2000 presidential election violated Equal Protection Clause of Four-
teenth Amendment); District of Columbia v. Heller, 128 S. Ct. 2783 (2008) (holding that
District’s total ban on handgun ownership violates respondent’s individual Second Amend-
ment right to bear arms); Gonzales v. Carhart, 550 U.S. 124 (2007) (upholding federal
Partial-Birth Abortion Ban Act).
18 Limits of Legality: Part I

law. Liberals and conservatives now accuse each other of using double
standards—of condemning deviation only when it thwarts their politi-
cally favored ends. Politically opinionated legal scholars display great inge-
nuity in trying to establish that they are consistent whereas their opponents
are not. Conservatives try to justify disregarding liberal constitutional
precedents by claiming that those cases themselves deviated from the
Constitution. Liberals construct theories of legal content and legal rea-
soning designed to show that these earlier decisions were, in fact, legally
correct.
Scholars in these debates disagree on many issues, but the debates also
presuppose a basic point of agreement. The disputants almost always
agree that, whatever the law requires, judges are obligated to apply it
correctly, full stop. Liberals and conservatives merely disagree about what
constitutes correct application of the law.62 Each faction assumes that it
can prevail by demonstrating that its opponent’s legal reasoning is mis-
taken. My thesis entails that the underlying premise itself is mistaken.
Judges are not necessarily obligated to apply the law correctly.
My thesis distinguishes this book from much legal scholarship. Legal
scholars often write as advocates for particular results, doctrines, and legal
arguments. They attempt to convince readers that particular cases were
decided correctly or incorrectly. They advocate for particular results in
pending cases. This book is different. I have opinions about cases just as
you do, but I rarely have occasion to mention my legal opinions in this
book and I never rely upon them. I shall not try to persuade you to
embrace “my” interpretations of the Due Process Clause or the Fair Labor
Standards Act. You could disagree with me on every controversial legal
question and still accept my arguments in this book. My conclusions could
be used to support results favored by liberal Democrats or those favored
by conservative Republicans. My position entails, for example, that a fed-
eral judge is not necessarily obligated to uphold a ban on late-term abor-
tions even if it has been proven constitutional.63 But my position also entails
that a judge is not necessarily obligated to invalidate such a ban even if it
has been proven unconstitutional. Such a position risks offending every-
one but I am prepared to take that chance.

62. “Strong opponents of abortion regarded Roe v. Wade as utterly lawless in much the
same way that Justice Ginsburg and other strong proponents of abortion rights now regard
Gonzales v. Carhart as utterly lawless.” Neil S. Siegel, “The Virtue of Judicial Statesmanship,”
Texas Law Review 86 (2008): 959–1032, p. 1027.
63. By “proven constitutional,” I do not just mean “held constitutional.” I mean “correctly
held constitutional.”
2

Practical Reasons and Judicial

Use of Force

2.1 PRIVATE REASONS, IMPARTIAL REASONS,

AND ROLE REASONS

This chapter introduces the concepts needed to develop my account of

judicial reasons. Practical reasons are my basic units of analysis. A practical
reason is a reason to act or to refrain from action,1 as opposed to a theo-
retical reason—a reason for belief.2 Throughout this book I shall write
as though practical reasons really exist.3 They have properties that we
can meaningfully discuss. They interact with one another. They explain
and justify intentional action. Most contemporary philosophers in ethics,
action theory, political philosophy, and legal philosophy make these
assumptions. I shall not defend them.4
Prudential reasons are easy to understand: they are reasons pertaining
to my own welfare. I have prudential reasons to eat when I am hungry,
sleep when I am tired, pursue my interests, make friends, avoid contagious
diseases, et cetera.5

1. Stephen J. Burton, “Law as Practical Reason,” Southern California Law Review 62

(1989): 747–93.
2. A note on terminology: possible readers of this book include philosophers, lawyers,
judges, and social scientists. Different fields use different jargon, so I sometimes explain basic
terms of art. I never mean to condescend to anyone.
3. This way of writing does not, necessarily, reflect a commitment to a fundamental
ontology that includes reasons. Some philosophers think any complete ontology will include
reasons at the basic level. Others think that reasons supervene on other, more basic facts.
Some even think that reasons will ultimately reduce to those more basic facts, without
remainder. As I incline to a naturalist ontology, I think reasons supervene on natural facts,
but I shall not defend that idea in this book. Most philosophers appear to believe that some
progress can be made in the theory of practical reasoning without settling fundamental
ontological issues. This book proceeds in that spirit.
4. But see, e.g., J. David Velleman, The Possibility of Practical Reason (Oxford: Oxford
University Press, 2000); Christine M. Korsgaard, “Skepticism about Practical Reason,” Jour-
nal of Philosophy 83 (1986): 5–25.
5. I like mushrooms, so I have preferential reasons to eat them. I dislike bananas, so I have
preferential reasons not to eat them. I could have preferential reasons to eat bananas; never-
theless, as when I want to climb a mountain, my body needs energy, and I have only bananas.

19
20 Limits of Legality: Part I

It is often said that special reasons—reasons of partiality—apply to

family members, friends, compatriots, and so forth.6 Perhaps the fact that
a girl who needs shoes is my daughter gives me a reason to buy her shoes.
I cannot have this kind of reason to buy your daughter shoes. Perhaps the
fact that George H. W. Bush was a U.S. citizen gave him a reason to join
the U.S. Navy. He could not have had that kind of reason to join the
French armed forces.7
Reasons of preference, prudence, and partiality collectively comprise the
class of private reasons. I have many reasons in addition to my private rea-
sons. When a man with a facial disfigurement passes me on the street, I have
moral reasons not to shout an insult at him. I have moral reasons to dial the
emergency line on my cell phone if I see him having a heart attack. I have
legal reasons not to shoplift a package of gum and legal reasons to cover my
icy front steps with sand. These legal and moral reasons are impartial reasons:
they apply to anyone in my situation if they apply to me.
Public officials are usually expected to disregard private reasons when
deliberating in their official capacities. A public official is supposed to act
only for reasons that would apply equally to anyone who occupied her
office. Some of these are reasons that apply to her in virtue of her office.
These are her role reasons.8 Police officers have role reasons to arrest sus-
pects. Public defenders have role reasons to defend them. Trial judges
have role reasons to preside over trials. Some of these role reasons are
legal reasons; others are moral reasons.
This book is about role reasons and impartial reasons, not private rea-
sons. My overarching question is, what happens to someone’s impartial,
moral reasons when she becomes a judge? Judges retain many of their
moral reasons, of course: they should not shoplift or shout cruel insults,
they should help strangers in distress, et cetera. I ask which moral reasons
judges retain, especially when moral reasons conflict with role reasons or
legal reasons not associated with the judicial role.

Some philosophers also believe that one’s welfare can be affected by events that are
unrelated to the satisfaction of one’s desires. They say that my welfare improves with my
physical health, even if an unhealthy lifestyle would satisfy more of my desires in the long
run. They insist that I have prudential reasons to live healthily even if I have no preferential
reasons to do so. The distinction between preferential and prudential reasons will play no
role in this book, but it is worth noting. Preferential reasons are nonmoral but agent-relative.
Prudential reasons are also agent-relative. They are arguably moral, if one has duties to
oneself. See chapter 12 on agent-relativity.
6. See, e.g., Susan Wolf, “Morality and Partiality,” Philosophical Perspectives 6 (1992):
243–59.
7. One might, however, have other reasons to buy shoes for someone else’s daughter or
to defend a foreign country.
8. See Arthur Isak Applbaum, Ethics for Adversaries: The Morality of Roles in Public and
Professional Life (Princeton, N.J.: Princeton University Press, 1999); David Luban, Lawyers
and Justice: An Ethical Study (Princeton, N.J.: Princeton University Press, 1988), chs. 6–7.
Practical Reasons and Judicial Use of Force 21

2.2 REASONS INTERACTING

Before beginning to answer these questions I must draw several additional

distinctions within the theory of practical reason. The most important is
the distinction between pro tanto reasons and all-things-considered rea-
sons.9 A consideration serves as a reason when it favors a certain action.10
A pro tanto reason to perform an action favors that action but can be over-
ridden by a stronger pro tanto reason. An agent has an all-things-considered
reason to do whatever the balance of pro tanto reasons favors. If I promise
to meet you for lunch at noon, then I acquire a pro tanto reason to meet
you. But if my young daughter breaks her leg shortly before noon, then I
have a stronger pro tanto reason to take her to the hospital and miss our
lunch appointment. We would say that I have an all-things-considered
reason to take my daughter to the hospital and a reason, perhaps, to apol-
ogize to you.
Morally relevant factors can interact in various ways at the ontological
level. A factor can enable another factor to favor a certain action without
itself favoring the action. It can intensify (or, conversely, attenuate) another
factor, causing the latter to favor an action more (or less) strongly.11 One
reason can override another without depriving the latter of its force, or it
can undermine12 the other, depriving it of all force. At the epistemological
level one reason can exclude another from consideration. Such an exclu-
sionary reason is a reason for an agent to disregard another reason.13

2.3 OBJECTIVE/SUBJECTIVE AND APPRAISAL/GUIDANCE

REASONS

A fact constitutes an objective reason for an agent if and only if it would

motivate a reasonable and fully informed agent. A fact constitutes a sub-
jective reason for a less than fully informed agent if and only if it would
motivate a reasonable and equally informed agent.14 It is warranted to
blame an agent for failing to act on her subjective reasons but not
for failing to act on objective reasons. Subjective reasons diverge from

9. Many have followed W. D. Ross, who called them “prima facie reasons,” but “pro tanto”
has become the preferred terminology. Shelly Kagan, The Limits of Morality (Oxford: Claren-
don Press, 1989), p. 17; W. D. Ross, The Right and the Good (Oxford: Clarendon Press, 1930).
All-things-considered reasons are also known as conclusive, decisive, or dispositive reasons.
10. I take the term from Jonathan Dancy, Ethics without Principles (Oxford: Oxford
University Press, 2004), p. 29.
11. Ibid., pp. 41–42.
12. Dancy calls this “disabling.” Ibid., p. 41.
13. Joseph Raz, Practical Reason and Norms, 2nd ed. (New York: Oxford University
Press, 1990), pp. 35–48.
14. Peter Railton, “Alienation, Consequentialism, and the Demands of Morality,”
Philosophy and Public Affairs 13 (1984): 134–71, p. 152.
22 Limits of Legality: Part I

objective reasons in cases of blameless wrongdoing and blameworthy

actions that happen to be morally permitted.15 I discuss both objective
and subjective reasons in this book, but I always try to specify which is
which.
A fact constitutes an appraisal reason for an agent to attempt to Φ if
and only if she has objective reason to succeed in her attempt. A fact con-
stitutes a guidance reason for an agent to attempt to Φ if and only if, in
attempting to Φ, she will successfully do whatever she has objective rea-
son to do, whether or not she succeeds in Φ-ing.16
The following scenario illustrates both objective/subjective and appraisal/
guidance distinctions. A Marine sniper has an objective, appraisal reason
to kill B.G., a terrorist who is about to explode himself in a crowd. The
sniper has been assured by his superiors that the man in his rifle scope is
B.G. However, the sniper knows from experience that if he aims at the
target’s head, then he misses, whereas if he aims at the space above the
target’s head, then he hits. If the man in his scope is B.G., then the sniper
has an objective guidance reason to fire at the space above the head of the
man in his scope. Given what he believes, he has a subjective guidance
reason to fire at the space above the head of the man in his rifle scope,
whoever that may be. If the man is, in fact, B.G., then the sniper has an
objective guidance reason to fire at the space above the head of the man
in his rifle scope.
Although I shall address what judges have subjective guidance reasons
to do in chapter 14, in the earlier chapters I shall mostly address their
objective appraisal reasons.

2.4 NATURAL REASONS AND DUTIES

A duty in my lexicon is a type of reason. It can objective or subjective, pro

tanto or all-things-considered. We have legal duties and moral duties.17 A
duty need not be owed to any identifiable person and it need not have a
correlative right, although many duties are owed to someone who has a
right that the duty be discharged.18

15. Stephen R. Perry, “Second-Order Reasons, Uncertainty, and Legal Theory,” Southern
California Law Review 62 (1989): 913–94, p. 922.
16. Similar distinctions appear in Cynthia A. Stark, “Decision Procedures, Standards of
Rightness and Impartiality,” Noûs 31 (1997): 478–95; David O. Brink, Moral Realism and the
Foundations of Ethics (New York: Cambridge University Press, 1989), pp. 216–17; R. E. Bales,
“Act Utilitarianism: Account of Right-making Characteristics or Decision-Making Proce-
dure,” American Philosophical Quarterly 8 (1971): 257–65.
17. Perhaps we have other types of duties, such as duties of etiquette.
18. See David Lyons, “The Correlativity of Rights and Duties,” Noûs 4 (1970): 45–55.
My arguments are consistent with the position that all moral reasons are moral duties, which
precludes the possibility of supererogatory actions: those that one has a moral reason, but no
Practical Reasons and Judicial Use of Force 23

I shall now examine the rights and duties of individuals in three groups:
people in a state of nature, residents of jurisdictions with established legal
systems, and public officials, particularly judges. My arguments assume
that all human beings have certain moral rights and reasons just by virtue
of being human. These are often called natural rights and reasons19 either
because they are natural to us or because they apply even to people in a
state of nature. Some natural rights and reasons are more controversial
than others. Although I assume that such rights and reasons exist, I try to
make minimal assumptions about exactly which natural rights and rea-
sons we have. I try to indicate when I make an especially controversial
assumption.
Individuals in a state of nature have several rights—to life and liberty,
at least. They have negative duties such as the duty not to kill other
people and positive duties such as the duty to care for dependent chil-
dren, to keep promises, et cetera. The details are not important at this
point. The duties and rights that concern me are duties of nonmaleficence,
samaritan rights, and rights of justice. These duties and rights are elements
of common morality and are recognized by most major moral theories,
including deontology, contractualism, indirect consequentialism, and virtue
ethics.20
First consider duties of nonmaleficence, preeminently the duty not to
use physical force against other human beings21 without justification.
Physical force includes, inter alia, the infliction of pain, injury, physical
constraint, or confinement. Also forbidden, absent justification, are using
coercion, inflicting emotional distress, and depriving someone of the use
of his property. One may also have a duty not to subject anyone to exces-
sive risks of pain, physical injury, emotional distress, or deprivation of
property.22 One has a duty not to threaten to do any of these things, a
duty not to instruct others to do any of these things, and a duty not to
assist others to do any of these things.23

moral duty, to perform. But my arguments are also consistent with the possibility of super-
erogatory actions. See David Heyd, Supererogation (Cambridge: Cambridge University Press,
1982). Similarly, my arguments are consistent with the view that all legal reasons are legal
duties, but also with the view that some legal reasons are not legal duties, leaving room for
the possibility of “legal supererogation.”
19. See, e.g., Susan Hurley, Natural Reasons (Oxford: Oxford University Press, 1989).
20. See, e.g., Marcia Baron, Philip Pettit, and Michael A. Slote, Three Methods of Ethics
(Oxford: Wiley-Blackwell, 1997); T. M. Scanlon, What We Owe to Each Other (Cambridge:
Belknap, 1998). The many theoretical disagreements concern the basis of these duties and
rights, how they apply in particular cases, and how to resolve conflicts between them.
21. I believe we have duties to nonhuman animals as well, but I shall not enter that
debate.
22. See §13.4.2.
23. At least when there is a nonnegligible possibility that one’s instructions will be
obeyed.
24 Limits of Legality: Part I

Individuals also have samaritan rights: rights to help strangers. We do not

often think in terms of “rights” to help strangers or to refrain from harming
them, but these are implied by our samaritan duties which are widely rec-
ognized. One has a right to discharge one’s moral duties. If one has no moral
right to rescue strangers in distress, then one has no moral duty to do so. But
one has a moral duty to rescue strangers in distress when one can easily do
so at negligible cost to oneself. Therefore, one has the right to protect
strangers from all kinds of misfortunes, at least when these are undeserved,
and provided that one does not thereby violate anyone’s moral rights. One
also has the right to instruct others to perform protective acts and to assist
others in doing so.24 The fact that one has duties of nonmaleficence and
samaritan rights entails that others have duties not to use force or coercion
to prevent one from fulfilling these duties or exercising these rights.25

2.5 SELF-HELP IN A STATE OF NATURE

Finally, individuals have several rights of justice. These are rights to use
force or coercion to achieve certain goals of justice. Rights of justice are
limited exemptions from certain general duties of nonmaleficence.26 There
is some controversy about the goals to which our rights of justice apply,
so I shall begin with the least controversial claim: individuals in a state of
nature are morally permitted to use reasonable force for the purpose of
defending themselves or others, at least against culpable aggressors who
pose imminent threats of bodily harm or substantial loss of property. The
fact that force is necessary constitutes a pro tanto reason that overrides or
undermines one’s general duty of nonmaleficence.27
Individuals also have a natural right to use force for the purpose of
obtaining compensation or restitution for wrongful injuries and losses,
and for the purpose of enforcing agreements or contracts.28 Suppose that
in a state of nature Gus deliberately sprains Valerie’s arm without

24. A right to instruct does not entail that anyone has a duty to obey.
25. These are Hohfeldian claim-rights. Leif Wenar, “The Nature of Rights,” Philosophy
and Public Affairs 33 (2005): 223–52, p. 229. Wesley Newcomb Hohfeld, Fundamental Legal
Conceptions (New Haven, Conn.: Yale University Press, 1919).
26. Rights of justice are, at least, single privileges: the possessor has no duty to refrain
from using force to do justice. If one has no duties of justice, then rights of justice are also
paired privileges. Wenar, “The Nature of Rights,” p. 226.
27. See, e.g., Jeff McMahan, “Self-Defense and the Problem of the Innocent Attacker,”
Ethics 104 (1994): 252–90; Judith Jarvis Thomson, “Self-Defense,” Philosophy and Public
Affairs 20 (1991): 283–310.
28. This right exists in both Lockean and Kantian traditions, although Kant also believes
that individuals in a state of nature must make efforts to found a civil society. See Katrin
Flikschuh, “Reason, Right, and Revolution: Kant and Locke,” Philosophy and Public Affairs 36
Practical Reasons and Judicial Use of Force 25

justification. On any theory of corrective justice Gus owes Valerie some-

thing. Stipulate that he owes her at least fifty apples29 in restitution for
this injury. Valerie has a right of justice to take fifty apples from Gus, by
force if necessary. Gus’s neighbor, Jack, has no individual duty to inter-
vene, but he has a right of justice to take fifty apples from Gus, by force if
necessary, for the purpose of delivering them to Valerie. Jack also has a
right to instruct others to exercise their rights of justice and to assist others
in exercising the same. Similar conclusions apply regarding the enforce-
ment of agreements.

2.6 PUNISHMENT AS A SPECIAL CASE

Many people also believe in a natural right to punish wrongdoers, what

Locke calls a “natural executive right.”30 However, this belief is more con-
troversial than the proposition that our rights of justice extend to restitu-
tion and the enforcement of agreements, so I shall treat punishment as a
special case. My arguments do not depend upon a natural executive right,
although they have different implications in many cases if such a right
exists. Specifically, if and only if there is a natural executive right, then my
arguments could support the right of judges to punish wrongdoers who
have committed no crime under positive law.31
Suppose for now that there is a natural executive right. Gus deliber-
ately sprains Valerie’s arm without justification. How, if at all, may some-
one punish Gus? This may depend, of course, upon the correct theory of
punishment. Let us stipulate that according to the correct theory, what-
ever it may be, Gus should be punished. If the correct theory is retribu-
tive, for example, then we stipulate that Gus deserves at least one month
of physical confinement. If the correct theory is consequentialist, then we
stipulate that better consequences result from confining Gus for one
month (or from maintaining such a practice). The same holds, mutatis
mutandis, for other theories of punishment. If there is a natural executive
right, then Valerie has a right of justice to punish Gus by humanely confin-
ing him for one month. Gus’s neighbor, Jack, has the same right, assuming

(2008): 375–404; Jeremy Waldron, “Kant’s Legal Positivism,” Harvard Law Review 109
(1996): 1535–66, p. 1562; Robert Nozick, Anarchy, State, and Utopia (New York: Basic
Books, 1974), p. 10.
29. Or something of comparable value.
30. See, e.g., A. John Simmons, “Locke and the Right to Punish,” Philosophy and Public
Affairs 20 (1991): 311–49, p. 313; Daniel McDermott, “The Permissibility of Punish-
ment,” Law and Philosophy 20 (2001): 403–32; Daniel M. Farrell, “Punishment without the
State,” Noûs 22 (1988): 437–53; Warren Quinn, “The Right to Threaten and the Right to
Punish,” Philosophy and Public Affairs 14 (1985): 327–73; Nozick, Anarchy, State, and Utopia,
p. 10; John Locke, Second Treatise of Government, para. 89.
31. See §10.8.
26 Limits of Legality: Part I

that no one else has already exercised it. Jack also has a right to instruct
others to exercise their rights of justice and to assist others in this exercise.

2.7 SELF-HELP IN A LEGAL SYSTEM

Jack now leaves the state of nature for a jurisdiction with a functioning
legal system.32 He retains his duties of nonmaleficence and his samaritan
rights from the state of nature. Others retain their duties not to use force
to prevent him from fulfilling these duties or exercising these rights. He
also acquires legal rights and duties for the first time. For the most part his
legal duties are compatible with his natural duties of nonmaleficence and
his samaritan rights. However, the law imposes an almost complete prohi-
bition on nondefensive self-help. The law generally forbids individuals to
use force for the purpose of carrying out private efforts to do justice, at
least when the state is prepared to intervene.33 However, the law permits
defensive self-help when one is threatened with certain kinds of immi-
nent harm and the state is not prepared to intervene.34
When Jack leaves the state of nature his rights of justice change ac-
cordingly. Suppose, again, that Gus sprains Valerie’s arm. Jack retains his
moral right to exact restitution from batterers, but this right is under-
mined if the state is prepared to perform this function more effectively
than Jack would, as in a reasonably just and effective legal system. For
various familiar reasons it is preferable that justice be done through ac-
countable, public agencies when these are available, rather than through
private self-help. Therefore, the fact that the state is prepared to take fifty
apples from Gus and deliver them to Valerie undermines Jack’s right to do
so.35 Jack’s general duties of nonmaleficence then prevail, giving him an
all-things-considered moral duty not to take Gus’s apples.
Similarly, if the state is prepared to enforce agreements or to punish
wrongdoers, then Jack has no right to use force for these purposes either.
Nor does Jack retain the right to instruct other nonofficials to exercise
rights of justice, or the right to assist them in exercising rights of justice.
In general terms, the fact that the state is willing and able to achieve a

32. I shall not discuss societies that are intermediate between the state of nature and
modern legal systems, in which some legal institutions exist but enforcement relies substan-
tially upon self-help. See the discussion of “compulsory universal suretyship” in Joel Feinberg,
“Collective Responsibility,” in Doing and Deserving (Princeton, N.J.: Princeton University
Press, 1970), p. 238. See also William Ian Miller, Bloodtaking and Peacemaking: Feud, Law,
and Society in Saga Iceland (Chicago: University of Chicago Press, 1990).
33. Landlord-tenant law was one of the last areas in which self-help was permitted.
Edward H. Rabin, “Symposium: The Revolution in Residential Landlord-Tenant Law: Causes
and Consequences,” Cornell Law Review 69 (1984): 517–84, p. 539.
34. See, e.g., Model Penal Code §3.04 (Use of Force in Self-Protection).
35. Unless Jack himself happens to be the presiding judge in Gus’s case. See §2.8.
Practical Reasons and Judicial Use of Force 27

certain goal of justice undermines the rights of justice, possessed by pri-

vate individuals, to use force to achieve that goal.
This reasoning even applies to defensive force. If the state is prepared
to defend Jack and others from aggression, then his moral right to use
defensive force is undermined. However, in practice the state is often
unable to defend individuals against imminent and unexpected attacks.
Jack is morally permitted to use even deadly force when necessary for
defending himself or others from attack, if the state is not prepared to
perform the defensive actions.36

2.8 JUDGES

Unlike most of us, some public officials use physical force and threats of
force against persons and property in the course of performing their pro-
fessional duties. Police officers forcibly enter private residences. They
physically assault, subdue, restrain, transport, confine, and detain suspects,
sometimes injuring or killing them in the process. Courtroom bailiffs
shackle defendants and lead them to jail. County sheriffs forcibly confis-
cate property. Marshals escort convicts to prison. Corrections officers
force inmates into cells—assaulting, injuring, and sometimes killing those
who attempt escape. Executioners carry out death sentences.
It is obvious that bailiffs, corrections officers, and other “hands-on”
officials use force on the job. It is less obvious but no less true that judg-
es regularly initiate the use of force in the course of their professional
duties. Of course judges engage in many professional activities that
involve no obvious use of force. They appear in court and oversee legal
proceedings. They maintain order in the courtroom, rule on motions,
study the law, et cetera. Trial judges examine evidence presented, hear
testimony, and instruct jurors. But many judicial activities, including
some just mentioned, actually involve the use of force or threats of force.
Judges levy fines, award monetary damages, grant injunctions, award
custody, et cetera. They sentence criminals to community service, prison,
or death.
The fact that judges initiate the use of force by their decisions would
be obvious if judges did the “dirty work” themselves. In modern legal
systems, however, the fact that judges use force is systematically obscured.37
Judges do not use their own hands to restrain, escort, transport, assault,

36. George P. Fletcher, “Domination in the Theory of Justification and Excuse,” Univer-
sity of Pittsburgh Law Review 57 (1996): 553–78, p. 570; Thomson, “Self-Defense,” p. 289.
37. Robert Cover identifies “three related responsibility-mitigation mechanisms” that
antebellum antislavery judges used to justify to themselves and others their decisions apply-
ing the law of slavery. These included “(1) Elevation of the formal stakes (sometimes com-
bined with minimization of the moral stakes). (2) Retreat to a mechanistic formalism. (3)
28 Limits of Legality: Part I

injure, execute, or confiscate property. They leave these tasks to subordi-

nate officials.38 In Robert Cover’s memorable words,
Legal interpretation takes place in a field of pain and death. . . . Legal interpre-
tive acts signal and occasion the imposition of violence upon others: A judge
articulates her understanding of a text, and as a result, somebody loses his free-
dom, his property, his children, even his life. . . . When interpreters have fin-
ished their work, they frequently leave behind victims whose lives have been
torn apart by these organized, social practices of violence.39

The fact that subordinates, rather than the judges themselves, use the
force can fool one into thinking that judges are not morally responsible
for its use. This is a mistake. A principal who gives instructions to an agent
is morally and legally responsible for actions taken by the agent that are
reasonably pursuant to his instructions.40 A supervisor who instructs a
subordinate employee to maim a competitor, for instance, is morally and
legally responsible for the injury.41
Likewise, when a judge awards civil damages to a plaintiff or sentences
a defendant to prison, his ruling includes an instruction to his subordi-
nates to use force—a command, in fact. He foresees that his command
will be obeyed, as it almost always is. Under these conditions the judge is
morally responsible for any actions taken by his subordinates that are
reasonably pursuant to his command, just as the supervisor is morally
responsible for the actions he orders. When a bailiff or prison guard carries

Ascription of responsibility elsewhere.” Robert M. Cover, Justice Accused: Antislavery and the
Judicial Process (New Haven, Conn.: Yale University Press, 1975), p. 199. Markus Dirk Dub-
ber notes that “[t]hese strategies have proved popular not only in early nineteenth century
North America, but in many countries and in many ages, whenever judges were eager to
deny the violence of their official function.” Markus Dirk Dubber, “The Pain of Punishment,”
Buffalo Law Review 44 (1996): 545–611, p. 582. Other writers who have emphasized the
fact that judges use violence include Hans Sherrer, “The Complicity of Judges in the Gener-
ation of Wrongful Convictions,” Northern Kentucky Law Review 30 (2003): 539–83; Patricia
M. Wald, “Violence under the Law: A Judge’s Perspective,” in Law’s Violence, ed. Austin Sarat
and Thomas R. Kearns (Ann Arbor: University of Michigan Press, 1992); Lynne Henderson,
“Authoritarianism and the Rule of Law,” Indiana Law Journal 66 (1991): 379–456, p. 405
(“judge as agent of state violence”); and John T. Noonan, Jr., Persons and Masks of the Law:
Cardozo, Holmes, Jefferson, Holmes, and Wythe as Makers of the Masks (New York: Farrar,
Straus & Giroux, 1976).
38. Raz classifies courts as “norm-applying institutions,” as distinct from “norm-enforcing
institutions” such as police departments, prisons, et cetera. Raz, Practical Reason and Norms,
pp. 132–34.
39. Robert M. Cover, “Violence and the Word,” Yale Law Journal 95 (1986): 1601–29,
p. 1601. See also Sherrer, “The Complicity of Judges in the Generation of Wrongful
Convictions,” p. 558.
40. Melvin Aron Eisenberg, An Introduction to Agency, Partnerships, and LLCs, 3rd ed.
(New York: Foundation Press, 2000), p. 12.
41. Model Penal Code §5.02 (Criminal Solicitation); Dan B. Dobbs, The Law of Torts, vol.
2 (St. Paul, Minn.: West, 2001), p. 905 (respondeat superior tort liability is the general rule
for employers).
Practical Reasons and Judicial Use of Force 29

out a judge’s orders, as the judge intended, the judge shares moral respon-
sibility for these actions with the compliant official.42 Hereafter, for the
sake of brevity, I shall often refer to a judge’s use of force with the under-
standing that the judge never actually uses her own hands.
It is rarely noticed that when a judge takes office and begins hearing
cases, she steps onto a scene in which threats of force have already been
made by others and new threats will be made, regardless of what she does.
Even before a judge has been assigned to his case, the defendant already
lives under heightened threats of force and a substantial likelihood of
being subjected to additional force or threats thereof. There is a substan-
tial possibility that the judge who ultimately decides the case, whoever
that turns out to be, will rule against the defendant. Therefore, every eli-
gible judge poses a potential threat to the defendant ex ante. These poten-
tial threats from all eligible judges survive even after one judge or panel
has been assigned to hear the case but before a decision has been ren-
dered. The first judge assigned does not always decide the case. Before she
decides it she may resign from the bench, be removed from office, self-
recuse, become incapacitated, die, or be disqualified from deciding the
case. But other judges are always waiting in the wings to be assigned as
replacements. So some judge will decide the case whether or not this judge
does. This is so even if the decision amounts to a dismissal of the case
pursuant to a plea bargain or civil settlement. Until a judge decides
the case the possible replacement judges pose potential threats to the
defendant.43
These background facts partly determine how the decision should be
characterized. By deciding the case, regardless of how she decides it, the
judge blocks potential threats to the defendant posed by the other judges
who could have decided it. This is so in both civil and criminal cases. If the
judge rules against the defendant, however, then the same action whereby
she blocks potential threats to the defendant also creates a new threat to
him, or we might say that she transforms a potential threat into an actual
threat. The threat is usually tacit. The judge’s order rarely mentions
explicitly what will happen if a civil defendant refuses to pay damages or
a convict tries to escape the bailiff while leaving the courtroom. But the
judge is, in fact, instructing other officials to threaten the defendant with
force if he does not comply with the order. An adverse ruling usually
requires the defendant to perform actions that he would prefer not to
perform or to refrain from actions that he would prefer to perform. If he
disobeys, then the state is very likely to compel compliance by force. The
defeated defendant is expected to understand this.

42. The judge may even be more responsible for these actions than is the subordinate
official because the judge gave the order. However, I need not make this claim.
43. Often, of course, one cannot predict with any confidence which threats will actually
be made, but that is just to say that they are all potential threats.
30 Limits of Legality: Part I

When she rules against a defendant the judge also withdraws or blocks
certain existing threats from other public officials. Because of the judge’s
decision these officials can now punish the (criminal) defendant or use
force to extract damages from the (civil) defendant, without themselves
being threatened with force by other officials.44 Whereas, when a judge
rules for a defendant she declines to block these threats of force, which
serve to protect the defendant. In criminal cases a prison or death sentence
may also block threats that the defendant poses to his future victims.
So a judge can choose to refrain from subjecting the defendant to
threats of force and to protect him from threats of force by other officials
(including judges) who might enforce the law against him. Or she can
withdraw threats of force from subordinate officials and instruct them to
use force against the defendant. Whatever the judge decides, she either
creates, withdraws, or blocks threats, or she does all three. Judges are in the
force business. In the next sections I shall ask what legal and moral author-
ity judges have to block and create threats of force.

2.9 JUDICIAL AUTHORITY

When someone becomes a judge she acquires new legal powers. The state
partially withdraws its legal prohibition on her use of force, conferring
upon her a limited legal power to use it.45 She receives partial legal autho-
rization to exercise her rights of justice. When a judge rules on a motion,
decides a case at law, or sentences a convict, she initiates the use or threat
of force. As the presiding judge, she has legal authorization to do so. Her
decision is called ultra vires if it exceeds the scope of her legal authority.46
Individuals ordinarily need moral justification, not just legal authoriza-
tion, when they use physical force or threats of force. As David Lyons
observes, “[t]he judicial process obtains against a certain background,
which includes the assumption that acts intended to deprive a person of
liberty, other valued goods, or even life itself require justification.”47

44. If a bailiff uses force without legal authorization, for example, then police officers
could arrest him.
45. “[A] judge has the legal right (power) to sentence a criminal to prison, meaning that
a judge has the ability to annul the criminal’s privileges of free movement.” Wenar, “The
Nature of Rights,” p. 231.
46. See, e.g., Boumediene v. U.S., 128 S. Ct. 2229, 2294 (2008) (Scalia, J., dissenting)
(Court acts “ultra vires” by conferring constitutional right to habeas corpus on alien enemies
detained abroad); William C. Duncan, “Goodridge and the Rule of Law Same-Sex Marriage
in Massachusetts [sic]: The Meaning and Implications of Goodridge v. Department of Public
Health,” Boston University Public Interest Law Journal 14 (2004): 42–55, p. 45 (Supreme
Judicial Court’s decision requiring same-sex marriage was “ultra vires”).
47. David Lyons, “Derivability, Defensibility, and the Justification of Judicial Decisions,” in
Moral Aspects of Legal Theory (Cambridge: Cambridge University Press, 1993), p. 134. See also
Grant Lamond, “The Coerciveness of Law,” Oxford Journal of Legal Studies 20 (2000): 39–62,
p. 40 (coercion is “ordinarily regarding as something which stands in need of justification”).
Practical Reasons and Judicial Use of Force 31

One could take various views of the moral permissibility of the use of
force by public officials. According to the naïve view a public official is
morally permitted to perform a certain action in a certain situation only if
a private party would be permitted to do so in that situation. When a
functioning legal system exists, private parties are rarely morally permit-
ted to use force, the main exception being when force is necessary to
defend oneself or others from imminent, unjustified aggression.48 Most of
the standard law enforcement and penal purposes for which officials use
force do not involve defending anyone from imminent aggression. There-
fore, if the naïve view were correct, then public officials would be morally
forbidden to use force for most of the standard purposes.
One can imagine the naïve view occurring to a child. The child under-
stands that it is “wrong” to hit other people. She sees one man strike
another with a short club. To her this looks like wrongful action. She calls
the assailant a “bad guy.” She wonders why no one intervenes. Assume,
however, that the assailant is actually a court bailiff striking an unruly
convict with a baton in order to lead him from the courtroom. In that case
the girl’s reaction is, indeed, naïve. As she matures she will learn that the
law authorizes certain individuals—called “bailiffs,” “police officers,” et
cetera—to use force in situations in which private parties are legally and
morally forbidden to use it. She will learn that this authorization is not
just legal, but moral.49 At least in a legal system with de facto legitimacy,
adults do not ordinarily become angry when public officials use legally
authorized force. They do not condemn these actions or try to stop them.
Although it would be wrong for a private party to use force in this (non-
defensive) situation, a legally authorized official is permitted to use it. If
one lives for some time under a legal system that one regards as legitimate,
then one will reject the naïve view in favor of what I shall call the habitu-
ated view of the morality of official uses of force within the system.50
Normal adults accept the habituated view. We believe that judges are
morally permitted to use force as authorized by law, at least in most
instances. The state is morally permitted to use force in at least the cir-
cumstances in which individuals in a state of nature are permitted to use
it: for defense, restitution, contract enforcement, and punishment of mala
in se offenses.51 Public officials are agents of the state who exercise these

48. See, e.g., Thomson, “Self-Defense.”

49. Children do not even draw this distinction until a certain age. Lawrence Kohlberg,
Essays on Moral Development: The Philosophy of Moral Development, vol. 1 (New York: Harper
& Row, 1981).
50. See Tom R. Tyler, Why People Obey the Law (New Haven, Conn.: Yale University
Press, 1990).
51. Remembering, again, that the natural executive right is controversial. Criminal
conduct is classified as malum in se if such conduct would be wrongful even if it were not
proscribed by law (e.g., in a state of nature). Bryan A. Garner, ed., Black’s Law Dictionary,
32 Limits of Legality: Part I

prerogatives on its behalf. Even philosophical anarchists can agree that

the state is morally permitted to use force to protect citizens from vio-
lence, to punish wrongdoers, et cetera. Anarchists merely deny that indi-
viduals have a moral duty to obey the law, as such, and that legitimate
states are possible.52
Shifting from the naïve view to the habituated view is a normal part of
moral development for residents of de facto legitimate legal systems. They
come to believe that the law can give public officials the moral authority
to perform actions that are legally and morally forbidden to the rest of us.
They correctly adopt this principle:
Moral Authority
Public officials are morally permitted to perform actions that would other-
wise be morally forbidden, if the law authorizes them to do so.

By authorizing officials to use force in certain cases, the law creates a

limited exception to the general moral prohibition on the use of nonde-
fensive force in civil society. If the moral authority principle is true, then
Jack’s moral rights change when he becomes a judge. Earlier, I claimed
that Jack’s residence in a substantially just and effective legal system par-
tially undermines his natural rights of justice.53 However, when Jack pre-
sides over Gus’s trial as a judge these rights are partially restored. Two
facts are necessary to this restoration. First, the jury has found Gus civilly
liable for spraining Valerie’s arm and has awarded Valerie monetary dam-
ages. Second, Jack is the presiding judge in Gus’s case, so the state is not
otherwise prepared to extract restitution from Gus if Jack does not do so.
Together, these two facts partially restore Jack’s natural right to do jus-
tice.54 Jack has a pro tanto moral reason to extract restitution from Gus.
The partial restoration of Jack’s rights of justice is significant because
Jack retains his natural pro tanto duty of nonmaleficence not to coerce
Gus. This duty persists even if Gus has received a fair trial and morally
ought to have restitution taken from him. However, Judge Jack’s rights of
justice undermine his duty of nonmaleficence in this case. In a highly cir-
cumscribed way, Judge Jack is returned to the state of nature with respect
to Gus.

8th ed. (St. Paul, Minn.: Thomson/West, 2004). Common examples of such conduct include
acts of violence against the person, such as homicide, rape, physical assault, injury, and con-
finement. Examples might also include theft, fraud, breach of agreement, inchoate offenses
such as conspiracy and attempt, et cetera.
52. See, e.g., A. John Simmons, Moral Principles and Political Obligations (Princeton,
N.J.: Princeton University Press, 1979), pp. 192, 196.
53. See §2.5.
54. The best way to describe this state of affairs, I think, is to say that these two facts
jointly undermine the reason that undermined Jack’s rights of justice: his residence in a
substantially just and effective legal system. Perhaps nothing turns on the accuracy of this
description.
Practical Reasons and Judicial Use of Force 33

I think this is why Judge Jack should feel no remorse about coercing
Gus. Jack’s act is not merely excusable, it is justified.55 This contrasts with
another scenario. I have been assuming that Gus ought to be coerced as a
moral matter. If the law under which Gus was found liable was an unjust
law, then some would deny that Jack was objectively morally permitted to
use force against Gus. Suppose Jack mistakenly believes that Gus deserves
to be coerced and also that he (Jack) has the moral right to coerce him. In
that case Jack should feel remorse when he learns the truth—that Gus did
not deserve coercion. Jack should apologize to Gus even if Jack’s beliefs
were reasonable at the time.56
Now imagine a different scenario. Damien and Pearl live in a politically
legitimate state. Damien negligently and unjustifiably damages Pearl’s
property. Under the law Pearl is entitled to an award of monetary dam-
ages. Such an award would, moreover, be entirely just. However, Pearl
does not sue Damien. Lewis, a third party, proceeds to conduct his own
private “trial” of Damien. Lewis “rules” for Pearl and forcibly extracts
damages from Damien. Lewis is not morally permitted to do so, even if he
follows the rules of civil procedure to the letter and reaches the same
result that a competent court would reach. If, however, Lewis becomes
the presiding judge in Pearl’s case, then he is morally permitted to rule in
her favor and to extract damages from Damien. I stated earlier that a
legitimate legal system provides reasons that undermine the moral rea-
sons to use force that residents would otherwise have. I now suggest that
the law restores those reasons to use force when it authorizes a judge to
use it in a certain case.57 If Judge Lewis has legal authorization to use force
against Damien, then he reacquires whatever reasons he would have had
to use it in a state of nature.58
We should reject the naïve view in favor of the habituated view. How-
ever, I believe that many people go too far when they abandon the naïve
view. They end up embracing a stronger principle:

55. See, e.g., Kent Greenawalt, “The Perplexing Borders of Justification and Excuse,”
Columbia Law Review 84 (1984): 1897–1927; J. L. Austin, “A Plea for Excuses,” Proceedings
of the Aristotelian Society 57 (1956–57): 1–30.
56. See Abigail Penzell, “Note: Apology in the Context of Wrongful Conviction: Why
the System Should Say It’s Sorry,” Cardozo Journal of Conflict Resolution 9 (2007): 145–61;
Bernard Williams, Moral Luck (Cambridge: Cambridge University Press, 1981), pp. 27–30.
57. I think the fact that someone becomes a judge undermines the reason that under-
mined his natural rights of justice when he “entered” civil society.
58. However, legal authorization does not override or undermine whatever other moral
reasons the judge might have not to use force. I am not morally permitted to pull your hair
for fun, even if the law permits me to do so.
34 Limits of Legality: Part I

Undermining
If the law requires a public official to use force in a given situation, then he
has no moral reason not to use it.59

The moral authority and undermining principles are easily conflated.

The latter entails the former, but not vice versa. In many cases the distinc-
tion makes no difference. The law often requires officials to use force
when it is right and good for them to use it. The difference between the
moral authority and undermining principles appears only when the law
requires officials to use force that is all-things-considered suboptimal. The
undermining principle entails that officials are morally permitted to use
force in such cases. Moral authority does not.
As you matured you made the transition from the naïve to the habitu-
ated view. If, when you did so, you came to accept the undermining prin-
ciple, not just moral authority, then I am going to ask you to unlearn some
of what you learned during that transition. As I do this you may feel as
though I am pushing you back into the naïve view—the view of a child
who does not yet understand that positive law can change what is morally
permitted and required. In fact, I reject the naïve view. I merely challenge
the undermining principle. I shall argue that judges retain their pro tanto
moral reasons to perform certain actions that the law forbids and to omit
certain actions that the law requires.60

59. A principle even stronger than undermining states the following: if the law autho-
rizes a public official to use force, then he has no moral reason not to use it. This principle
has many implausible implications. It entails, for example, that judges never have moral
reasons not to impose the maximum sentence allowed by law, even if they have discretion
to impose a lower sentence. The Montana Penal Code authorizes prison terms of up to forty
years for marijuana possession, but does not mandate such lengthy sentences. The stronger
principle entails that a Montana judge has no moral reason not to impose forty-year sen-
tences in such cases. For many readers, this implication of the stronger principle will suffice
to discredit it. If you find it appealing, however, you can still accept my arguments. Because
the stronger principle entails undermining, my arguments against undermining also reach
the stronger principle, which I shall not mention again. See Mont. Code Ann. § 45–9-103
(2005); Hutto v. Davis, 454 U.S. 370 (1982) (upholding forty-year prison sentence for dis-
tributing, and possessing with intent to distribute, less than nine ounces of marijuana).
60. See chapter 6.
3

Deviating from Legal Standards

3.1 THE JUDGE

The protagonist in this book is a judge deliberating in her chambers. She

is a generic character meant to represent a wide range of judges in the real
world. I will not tell you much more about her.1 I will not tell you any-
thing that would distinguish her from most other judges. I imagine her
sitting in the United States because this is the legal system that I know
best.2 She might be a state judge, a federal judge, an administrative law
judge, or what have you. She might sit at the trial level, on an intermedi-
ate appellate court, or on a high court. The case on which she is deliber-
ating could be a civil case, a criminal case, a constitutional case, a regulatory
matter, or some other type of legal proceeding. These distinctions can be
important in practice, but I try to make points that apply across all these
categories. I also try to notice when these distinctions make a difference
to my discussion.
I imagine my judge in the judicial branch, although some of my con-
clusions apply, mutatis mutandis, to other officials as well, including adju-
dicators in the executive branch. I assume that my judge sits alone on her
court rather than as a judge on a multimember panel (as when federal
appellate courts sit en banc). Collegial courts raise special issues that I
cannot discuss in this book.
My judge is far from perfect. Her mental powers, knowledge, and
moral judgment are all human, not superhuman. She could be deeply
religious, atheistic, or spiritually uncommitted. Her political views could
fall anywhere on the political spectrum represented by modern public
officials and intellectuals: she could be Ann Coulter or House Speaker
Nancy Pelosi, Justice Clarence Thomas or the late Justice Thurgood
Marshall. She has a conscience and a set of sincerely held values upon
which she is prepared to act when appropriate. She does not simply wish
to use her position to accumulate money or power. She wants to be a

1. My hypothetical judge changes gender frequently.

2. I suspect many of my arguments apply outside the United States, but I cannot write
with confidence about this.

35
36 Limits of Legality: Part I

good judge, whatever that means to her, and a good person, whatever that
means to her.

3.2 JUDICIAL INCENTIVES TO PERFORM

The first question to ask about my protagonist is what reasons she has
for and against performing her professional functions. Consider pruden-
tial reasons, which operate as incentives for substantially self-interested
judges. Judges have incentives, often financial, to take certain actions
that are forbidden to them: to accept bribes, to practice law on the side,
et cetera. However, judges who take these actions engage in judicial
misconduct, inviting censure, reprimand, impeachment, removal from
office, disbarment, and so forth.3 Judges wish to avoid these sanctions,
so they have strong counterincentives to reject bribes, to refrain from
practicing law, et cetera. Some would refrain from these actions from a
sense of professionalism or personal ethics even if no sanctions were
threatened.
Judges also have positive duties of office: for example, to appear in
court as scheduled, to maintain order in the courtroom, to rule on
motions by parties, to examine evidence presented, to hear testimony,
to study the law, to decide cases, to sentence convicts. At one time or
another, every judge has an incentive to shirk his duty in order to avoid
hard work. Some judges dislike their work and would prefer less of it.
Even judges who enjoy the job also enjoy leisure time and would like
more of it. Some judges would prefer playing golf to appearing in court.
However, failure to fulfill the positive duties of the office constitutes mis-
conduct and can bring sanctions.4 Again, the judge who wishes to simplify
his job or to avoid work altogether has counterincentives.
A judge who wishes to avoid hearing a certain case can, in theory,
recuse himself, but recusal is an ineffective means of reducing one’s work-
load. Given docket backlog, a judge who recuses himself will shortly
receive another case assignment. Moreover, recusing oneself can consti-
tute misconduct if done for an improper reason.5 A judge who occasion-
ally recuses himself for improper reasons is unlikely to face sanctions, but

3. See chapter 4.
4. “The judicial duties of a judge take precedence over all other activities.” Code of Con-
duct of United States Judges, Canon 3; “The judicial duties of a judge take precedence over all
the judge’s other activities.” Model Code of Judicial Conduct, Canon 3 (1990).
5. “A judge shall hear and decide matters assigned to the judge except those in which
disqualification is required.” Model Code of Judicial Conduct, Canon 3B(1) (1990); “A judge
should hear and decide matters assigned, unless disqualified. . . .” Code of Conduct of United
States Judges, Canon 3A(2). See also Model Code of Judicial Conduct, Rule 2.7 (1990) (“A
judge shall hear and decide matters assigned to the judge, except when disqualification is
required by Rule 2.11 or other law.”); 28 U.S.C. § 453 (“A judge has a duty to decide what-
ever cases come before him to the best of his ability.”)
Deviating from Legal Standards 37

a pattern of frequent recusal could bring sanctions if it appears that the

recusals are improper (e.g., he is deliberately avoiding difficult cases, using
conflict of interest as a pretext).6
Judges also want to maintain good professional reputations. Many want
to receive interesting and/or prestigious opportunities, both remunerative
and nonremunerative, in the future. Many wish to be promoted to higher
office, judicial or otherwise.7 A judge who recuses himself too frequently
or fails to appear in court may harm his professional reputation long
before sanctions are threatened. Therefore, he has another incentive to
hear his assigned cases. Insofar as a record of unwarranted recusal (or,
worse, absenteeism) reduces one’s attractiveness for interesting, lucrative,
and prestigious opportunities, judges have an incentive to hear and decide
the cases that are assigned to them.
Many judges also want other courts to cite and follow their decisions.8
Many want to influence the development of legal doctrine.9 Judges who
fail to show up or who recuse themselves too often have fewer opportu-
nities to exert influence.

3.3 ANSWERING LEGAL QUESTIONS

My protagonist appears on the bench. Once there, she tries to answer

legal questions on a rational basis.10 She does not decide randomly (unless,

6. It is worth noting, moreover, that the usual concern is under-recusal, not over-recusal.
See, e.g., Debra Lyn Bassett, “Recusal and the Supreme Court,” Hastings Law Journal 56
(2005): 657–98.
7. See Frederick Schauer, “Incentives, Reputation, and the Inglorious Determinants of
Judicial Behavior,” University of Cincinnati Law Review 68 (1999): 615–36; Thomas J. Miceli
and Metin M. Cogel, “Reputation and Judicial Decision-Making,” Journal of Economic Behav-
ior and Organization 23 (1994): 31–51; Richard A. Posner, “What Do Judges and Justices
Maximize? (The Same Thing Everybody Else Does),” Supreme Court Economic Review 3
(1993): 1–41; Mark A. Cohen, “The Motives of Judges: Empirical Evidence from Antitrust
Sentencing,” International Review of Law and Economics 12 (1992): 13–30.
8. William M. Landes and Richard A. Posner, “Legal Precedent: A Theoretical and
Empirical Analysis,” Journal of Law and Economics 19 (1976): 249–307, p. 273.
9. Jonathan R. Macey, “The Internal and External Costs and Benefits of Stare Decisis,”
Chicago-Kent Law Review 65 (1989): 93–112, pp. 111–12.
10. In wanting to make decisions on a rational basis, my judge does not insist that reason
and emotion are fundamentally opposed. Nor does she deny that emotions can provide
moral insight. See, e.g., Shaun Nichols, Sentimental Rules: On the Natural Foundations of
Moral Judgment (Oxford: Oxford University Press, 2004); Martha C. Nussbaum, “Emotion
in the Language of Judging,” St. John’s Law Review 70 (1996): 23–30; Antonio R. Damasio,
Descartes’ Error: Emotion, Reason, and the Human Brain (New York: Harper & Row, 1995);
Ronald De Sousa, The Rationality of Emotion (Cambridge, Mass.: MIT Press, 1987). She
simply recognizes that many emotions should not control judicial thinking. For example, she
does not treat parties with attractive faces more (or less) favorably than average-looking
ones. Nor is she influenced by the fact that the attorney arguing before her wrote an opinion
editorial in the local paper with which she disagrees. Real judges do, unfortunately, some-
times allow inappropriate emotions to influence them. My judge tries to avoid this.
38 Limits of Legality: Part I

perhaps, she determines that it is rational to do so).11 Nor does she simply
consider the facts presented and reach what she considers to be the opti-
mal result without regard for the law. Rather, before she reaches a decision,
she determines whether there are any applicable legal standards. Legal
systems contain many putative sources of law, including constitutions,
treaties, executive orders, statutes, ordinances, administrative regulations,
and reported judicial opinions. These sources purport to provide legal
standards for decision. The judge may ultimately decide to disregard these
standards, but she always wants to know what they are.
Readers who think it obvious that competent judges aim to apply the
law will be surprised that I find it necessary to mention that my judge
takes an interest in legal standards. I mention this stipulation because it
cannot be taken for granted after the twentieth century. The past century
witnessed many interrelated debates about the process of adjudication.12
One was a factual, psychological debate about how judges decide. The
American Legal Realists claimed that real judges decide cases for nonlegal
reasons, such as their personal opinions about fairness or good public pol-
icy, and then construct legal arguments, adverting to sources of law, for
use in their reported opinions. The most radical realist believes that legal
standards never influence judges.13
The concurrent debate was a normative debate about how judges ought
to decide. Again, the realists tended to support the use of nonlegal stan-
dards whereas their critics opposed it. My judge is still deciding whether
and when to take legal standards into account. She is willing to be per-
suaded to ignore them, but she could also be persuaded to take them into
account. It remains an open question for her. If you believe that judges
never take legal standards into account in any way, and that they could
not be persuaded to do so, then this book has little to offer you.

3.4 STANDARDS IN CONFLICT

If you are still reading, then I assume that you are prepared to accept the
possibility that at least some judges, some of the time, could be persuaded
to take legal standards into account, whether or not they currently do so.
If you accept this possibility, then you can share my interest in cases in
which valid, applicable legal standards conflict with other normative rea-
sons, such as justice or the general welfare.
For any legal question, one can imagine a legal standard that purports
to be relevant to an answer—one that applies to the question. Consider

11. See §15.4.

12. See generally, Brian Leiter, “Legal Realism and Legal Positivism Reconsidered,”
Ethics 111 (2001): 278–301.
13. See generally, William W. Fisher, III, Morton J. Horwitz, and Thomas Reed, eds.,
American Legal Realism (Oxford: Oxford University Press, 1993).
Deviating from Legal Standards 39

the following: “Is it legal to sell fireworks in the City of Houston, Texas?”
Imagine a municipal ordinance that states, “It shall be a misdemeanor for
anyone to sell fireworks in the City of Houston.” This rule applies to the
question posed. Compare another proposed ordinance that states, “It shall
be a misdemeanor for anyone to sell alcoholic beverages in the City of
Houston.” This rule does not apply to the question at all. The question
here is not whether either ordinance is legally valid in Houston but
whether it would apply to the question posed if it were valid.
For any legal question and any legal standard or set of standards, we can
refer to the degree to which the standard regulates the question. A question
is not legally regulated at all unless at least one legal standard applies to it.
At one extreme lie questions to which the specified standards dictate a
unique answer. These questions are fully regulated by the standards.14 At
the opposite extreme lie questions to which no answer is excluded by the
standards. These questions are fully unregulated by the standards. In
between these extremes lie partially regulated questions, to which the
standards preclude some answers but allow others. Of course, the fact that
a hypothetical standard applies to a certain legal question tells us nothing
about the legally correct answer to the question in the real world. Only
legally valid standards dictate legally correct answers in the real world.
For any legal question that is at least partially regulated by a standard,
it is possible for a judge to give an answer that departs from that standard.
This is what I call deviating from a standard as opposed to adhering to it.15
These are the central concepts of this book. The verb “to adhere” is my
term of art, referring to the judicial act of correctly applying a legal stan-
dard. I use it to distinguish adherence from the generic action of obeying
or following a standard. Anyone can obey a standard, but only those in law-
applying roles, such as judges, can adhere to one. Likewise, deviating from
a standard is distinct from the generic action of disobeying a standard,
which anyone can do.16

14. Some writers refer to cases in which all legal questions are fully regulated as “easy
cases,” in contrast to “hard cases.” See, e.g., David Lyons, “Derivability, Defensibility, and the
Justification of Judicial Decisions,” in Moral Aspects of Legal Theory (Cambridge: Cambridge
University Press, 1993), p. 119; Frederick Schauer, “Easy Cases,” Southern California Law
Review 58 (1985): 399–440; Ronald Dworkin, “Hard Cases,” in Taking Rights Seriously (Cam-
bridge, Mass.: Harvard University Press, 1977). However, the process of identifying the
legally correct result may be difficult, in practice, even when there is a definite correct
answer. Compare complicated arithmetic problems.
15. Others use different terminology. Alexander and Sherwin call deviation “defiance.”
Larry Alexander and Emily Sherwin, The Rule of Rules (Durham, N.C.: Duke University
Press, 2001), p. 47.
16. As I have defined the terms, deviating from a legal standard is a redundant phrase, as
is adhering to a legal standard, but I will sometimes use these phrases, and others, such as
deviating from the law, for stylistic reasons. In chapter 10, I shall consider the claim that
judges have a legal duty to adhere to the law. If one law requires judges to adhere to another
law, then adhering to the second law is, in fact, obeying the first, and deviating from the
second is disobeying the first.
40 Limits of Legality: Part I

The cases that interest me involve conflicts between legally valid

standards and other standards, legally valid or not. Consider cases in
which two opposing standards or sets of standards apply to the case,
the first standard or set is legally valid in the jurisdiction, and the first
standard or set requires less individual judgment to apply than the
second. There are three categories of cases with these characteristics.
One category comprises cases in which the first standard is a valid legal
rule, and the second is another type of legal standard such as a princi-
ple.17 The rule “Do not exceed thirty-five miles per hour” requires less
judgment to apply than does the standard “Do not exceed a safe speed.”
In the second category, the first standard is the most locally applicable
legal standard whereas the second standard is a more remotely applica-
ble legal standard.18 In the third category, the first is an entire set of
applicable, valid legal standards (rules and/or principles) whereas the
second is a moral principle or set of principles that is not found in the
law at all.
In each of these three categories I am interested in the conditions
under which a judge is morally permitted to deviate from the first stan-
dard, or set of standards, in favor of the second. I think my arguments in
this book have implications for all three categories of cases, including
cases of legally authorized deviation from legal rules. However, I con-
centrate on cases in the third category: conflicts between legal standards
and extralegal principles. To defer to extralegal principles in such cases
is to deviate from the law simpliciter. This is my paradigm case of devia-
tion. For any legal question that is at least partially regulated by law, it
is possible to give a legally incorrect answer—to deviate from the law
simpliciter.
In the next three sections I shall briefly examine some other types of
decisions that might easily be confused with deviation from the law.

3.5 DEPARTURES FROM VALID LEGAL RULES

A decision deviates from a valid legal rule if and only if it conflicts with a
rule that is legally valid when the decision is made. Therefore, we must
distinguish between invalidating legal rules and deviating from valid legal
rules. Judges often have legal authority to invalidate legal rules, as when
appellate courts invalidate previously valid legal rules announced by lower
courts. A judge also invalidates a valid legal rule when she decides that

17. See Ronald Dworkin, “The Model of Rules I,” in Taking Rights Seriously (Cambridge,
Mass.: Harvard University Press, 1977).
18. For the concept of locality in this context see Ronald Dworkin, Law’s Empire (Cam-
bridge, Mass.: Harvard University Press, 1986), pp. 251–54; Frederick Schauer, Playing by the
Rules: A Philosophical Examination of Rule-Based Decision-Making in Law and in Life (Oxford:
Oxford University Press, 1991), pp. 188–89.
Deviating from Legal Standards 41

federal law preempts a state statute, in whole or in part.19 These powers

of invalidation are universally recognized as lawful and legitimate.
Courts also invalidate rules announced in their own prior decisions—a
legally accepted practice that, nevertheless, is sometimes criticized.20
Courts invalidate statutes as unconstitutional.21 This practice has long
been accepted, although its political legitimacy and wisdom are still some-
times questioned.22 Historically, courts also have legal authority under the
doctrine of desuetude, which allows that “under some circumstances stat-
utes may be abrogated or repealed by a long-continued failure to enforce
them.”23 Some scholars would even extend to courts the legal authority to
revise or invalidate statutes that have become obsolete, without finding
them unconstitutional.24 Such proposals remain highly controversial.25
The U.S. Supreme Court has long recognized that courts have legal
authority to depart from the plain meaning of statutory text when adher-
ing would produce “absurd” results.26 But even this practice, when legally
authorized, could be understood as modification of the rule, rather than
deviation from it. If a court has legal authority to invalidate a valid rule,
then it does not deviate from the rule in so doing. There is no conflict
between the rule and the result at the time of application. That conflict
defines what I call deviation.

19. Bethlehem Steel Co. v. New York State Labor Relations Bd., 330 U.S. 767 (1947)
(holding that the National Labor Relations Act preempts the New York Labor Relations
Board from allowing foremen to unionize).
20. See, e.g., Payne v. Tennessee, 501 U.S. 808, 844–45 (1991) (Marshall, J., dissenting).
21. Marbury v. Madison, 5 U.S. 137 (1803); McCulloch v. Maryland, 17 U.S. 316
(1819).
22. See, e.g., Jeremy Waldron, “The Core of the Case against Judicial Review,” Yale Law
Journal 115 (2006): 1346–406; Larry D. Kramer, The People Themselves: Popular Constitution-
alism and Judicial Review (Oxford: Oxford University Press, 2004); Jeremy Waldron, Law
and Disagreement (Oxford: Oxford University Press, 1999); Mark Tushnet, Taking the Con-
stitution Away from the Courts (Princeton, N.J.: Princeton University Press, 1999).
23. Arthur E. Bonfield, “The Abrogation of Penal Statutes by Nonenforcement,” Iowa
Law Review 49 (1964): 389–440, p. 394.
24. See, e.g., Douglas E. Edlin, Judges and Unjust Laws: Common Law Constitutionalism
and the Foundations of Judicial Review (Ann Arbor: University of Michigan Press, 2008);
William N. Eskridge, Jr., Dynamic Statutory Interpretation (Cambridge, Mass.: Harvard University
Press, 1994); Guido Calabresi, A Common Law for the Age of Statutes (Cambridge, Mass.:
Harvard University Press, 1982).
25. See, e.g., Samuel Estreicher, “Judicial Nullification: Guido Calabresi’s Uncommon
Common Law for a Statutory Age,” New York University Law Review 57 (1982): 1126–73.
Even the routine practice of distinguishing prior cases constitutes a kind of invalidation.
Suppose prior cases have been cited for a certain doctrine. The doctrine requires a certain
result in the instant case, but the court distinguishes the instant case from the prior cases,
creating a doctrinal exception where none previously existed. It thereby revises or elaborates
the doctrine. The revised doctrine is consistent with the prior cases, but the new exception
allows a result in the instant case that the original, exceptionless doctrine would have for-
bidden. See further discussion in §3.10.
26. See John F. Manning, “The Absurdity Doctrine,” Harvard Law Review 116 (2003):
2387–486.
42 Limits of Legality: Part I

Of course, courts do not have the authority to invalidate most rules. If

a court invalidates a rule that it lacks the authority to invalidate, then it
deviates from the law simpliciter, in addition to deviating from the rule. It
is also possible for a court to deviate from a rule that it has the legal
authority to change, if it fails to change the rule before rendering its
decision.27

3.6 DEVIATIONAL DISCRETION

As I explained in the previous section, many departures from legal rules

are instances of legally authorized invalidation: federal preemption of
state law, reversal of prior decisions, invalidation of unconstitutional stat-
utes, desuetude, absurdity, distinguishing of precedents, et cetera. Such
departures do not satisfy my definition of deviation. Coauthors Mortimer
and Sanford Kadish describe decisions that come closer to satisfying my
definition. They suggest that judges sometimes have legal authority to
deviate from legal rules.28 They introduce the idea of “recourse roles”
within the legal system: “roles that enable their agents to take actions in
situations where the role’s prescribed ends conflict with its prescribed
means, including grants of discretion, broad or narrow.”29
The Kadishes defend at length the idea that jurors in criminal trials
play recourse roles.30 A juror, they claim, “considers whether literal adher-
ence to the judge’s instructions will advance or impede the goals of crim-
inal justice as well as the institutional and background ends of the society
more generally.”31 Characterizing the juror’s role as a recourse role allows
one to conclude that jury nullification32 can be legally justified. According
to the Kadishes this is so even if the juror’s decision to nullify is morally
reprehensible. They use the example of “a Southern jury that acquits a
white segregationist of killing a civil rights worker, on the grounds that in
the public interest carpetbag troublemakers must be discouraged from
venturing into their community, and that in any event the defendant’s act
was a political act that should not be punished as a common crime.”33

27. If a court has the authority to change a certain rule (e.g., a rule announced in hori-
zontal precedent) and it makes a decision that conflicts with that rule, then observers are
likely to conclude that the court has, by that very act, changed the rule. This inference is
reasonable unless the court explicitly announces its intention not to change the rule from
which it deviates.
28. Mortimer R. Kadish and Sanford H. Kadish, Discretion to Disobey: A Study of Lawful
Departures from Legal Rules (Stanford, Calif.: Stanford University Press, 1973), pp. 85–91.
29. Ibid., p. 35.
30. Ibid., pp. 55–68.
31. Ibid., p. 61.
32. Jury nullification is “a jury’s ability to acquit a criminal defendant despite finding facts
that leave no reasonable doubt about violation of a criminal statute.” Darryl K. Brown, “Jury
Nullification within the Rule of Law,” Minnesota Law Review 81 (1997): 1149–200, p. 1149.
33. Kadish and Kadish, Discretion to Disobey, p. 68.
Deviating from Legal Standards 43

They characterize this acquittal as a “legitimated rule departure” and deny

that this jury has acted “lawlessly,” even if it is “egregiously wrong in its
interpretation of the ends of its role.”34
The Kadishes subsequently assert that the judicial role, too, is a recourse
role, although they admit that they have not defended this claim.35 The
supporting data seem to be that judges sometimes depart from legal rules
when they believe that applying them would contradict the “prescribed
ends” of the judicial role. Characterizing the judicial role as a recourse role
allows the conclusion that judicial nullification of legal rules can be legally
justified. If the judge deviates from legal rules and his “judgment is consci-
entiously made on his view of [the] ends [of the judicial role],”36 then his
deviation is legally legitimate, according to the Kadishes.37
The exercise of deviational discretion, as the Kadishes describe it, comes
close to satisfying my definition of deviation, but one difference remains.
The Kadishes insist that judges do not act lawlessly by exercising deviational
discretion—they have legal authority to deviate. The Kadishes are not con-
cerned to defend deviation from the law simpliciter, but rather to argue that
deviation from legal rules can be legally legitimate. They would not contest
Robert Summers’ claim that the rule of law requires that rule departures be
legitimated, in the sense “that any exceptional power of courts or other tribu-
nals to modify or depart from anterior law at point of application be a power
that, so far as feasible, is itself explicitly specified and duly circumscribed in
rules, so that this is a power the exercise of which is itself law-governed.”38

3.7 LEGAL PRINCIPLES

When legal scholars talk about “following the law” they often have in
mind the following of rules such as we find in statutes, regulations, and
case law. In this book I will accept arguendo that rules comprise at least
part of the law. However, not everyone agrees that the law consists
entirely of posited rules—rules that “come into existence at particular
times and places” by way of enactment or announcement.39 Some natural
lawyers believe that the law incorporates moral standards that have not
been enacted or announced.40 Dworkinians and inclusive positivists

34. Ibid.
35. Ibid., pp. 85–91.
36. Ibid., p. 69.
37. A recent defense of the thesis that Anglo-American judges have the legal authority
to overturn unjust laws is Edlin, Judges and Unjust Laws.
38. Robert S. Summers, “The Principles of the Rule of Law,” Notre Dame Law Review 74
(1999): 1691–712, p. 1694. Cf. Edlin, Judges and Unjust Laws (judges have a legal duty to
“develop the law,” which requires not enforcing highly unjust laws).
39. Alexander and Sherwin, The Rule of Rules, p. 27.
40. See, e.g., Michael Moore, “Law as a Functional Kind,” in Natural Law Theory: Con-
temporary Essays, ed. Robert P. George (Oxford: Oxford University Press, 1992).
44 Limits of Legality: Part I

believe that the law contains, in addition to rules, other kinds of legal
standards, such as principles and policies, which can legally authorize
departures from legal rules.41 In some cases the rules require a result of
which the presiding judge disapproves, for some reason, but she can some-
times identify legal principles that authorize a preferable result, rules not-
withstanding. This is the reconciliation strategy that Dworkin thinks the
courts used to reach putatively just results, unsupported by rules, in Riggs
v. Palmer and Henningsen v. Bloomfield Motors.42
Disputes about whether the law includes standards other than rules
arise within the theories of law and legal content. I shall not attempt to
resolve these disputes. This book aims for compatibility with a wide range
of theories of content so I shall not assume either that legal standards
other than rules exist or that they do not.43 When I talk about “following
the law” or “adhering to the law” I shall understand the law as broadly as
possible. If principles are part of the law, then so be it.
I am interested in the conditions under which judges are morally per-
mitted to deviate from the law simpliciter, however we define law. It is
simplest, however, to discuss cases in which the applicable law consists
exclusively of codified rules such as statutory provisions. My paradigm
case is one in which the only controlling legal authority is a legal rule
that dictates an unwelcome result. I concentrate on such cases for three
reasons.
First, everyone agrees that modern legal systems contain rules, whereas
it remains controversial whether principles should even be classified as a
distinct type of legal standard.44
Second, compared to principles and other legal standards, rules require
the least judgment to apply. Ceteris paribus, it is easier to ascertain
whether a result conforms to a certain rule than it is to ascertain whether
a legal principle supports that result.
Third, the judicial obligation to adhere to the law must apply to rules
if it applies to anything. This is not to insist that judges must always
adhere to rules. It is just to observe that if judges have an obligation to

41. See, e.g., Mark Greenberg, “How Facts Make Law,” Legal Theory 10 (2004): 157–98;
Jules Coleman, The Practice of Principle (Oxford: Oxford University Press, 2001); Matthew
Kramer, “How Moral Principles Can Enter into the Law,” Legal Theory 6 (2000): 83–108;
Kenneth Einar Himma, “Judicial Discretion and the Concept of Law,” Oxford Journal of
Legal Studies 19 (1999): 71–82; Nicos Stavropoulos, Objectivity in Law (Oxford: Clarendon
Press, 1996); W. J. Waluchow, Inclusive Legal Positivism (Oxford: Clarendon Press, 1994);
Stephen R. Perry, “Second-Order Reasons, Uncertainty, and Legal Theory,” Southern Califor-
nia Law Review 62 (1989): 913–94.
42. Dworkin, “The Model of Rules I”; Riggs v. Palmer, 115 N.Y. 506 (1889); Henningsen
v. Bloomfield Motors, 32 N.J. 358 (1960).
43. Here I follow Schauer, Playing by the Rules, p. 11.
44. See, e.g., Larry Alexander and Ken Kress, “Against Legal Principles,” in Law and
Interpretation: Essays in Legal Philosophy, ed. Andrei Marmor (Oxford: Clarendon
Press, 1995).
Deviating from Legal Standards 45

adhere to the law, then there are some rules to which judges must
adhere. If, conversely, we conclude that judges are morally permitted to
deviate from a result dictated by clear, codified rules, when these are
the only controlling legal authority, then it follows easily that they are
also morally permitted to deviate from a result dictated by a wider
range of legal standards, such as principles and policies. If we conclude
that judges are never morally obligated to adhere to legal rules, then it
follows a fortiori that they are never obligated to adhere to other legal
standards.45

3.8 LEGALLY UNAUTHORIZED DEVIATION FROM THE LAW

Judicial departures from legal rules are commonplace. If such a depar-

ture is legally authorized in a certain case, then by definition it does not
constitute deviation from the law. Deviation from legal rules can be
authorized by more authoritative legal standards, such as higher level
rules, constitutional provisions, or (arguably) sufficiently important
legal principles. I shall write that a decision deviates from the law sim-
pliciter if and only if it conflicts with all applicable legal standards,
whether or not these take the form of rules.46 Of course, if it turns out
that rules are the only legal standards (contra Dworkin), then the afore-
mentioned decision is impossible because a decision that conflicts with
legal rules is ipso facto inconsistent with the law. We cannot know which
decisions count as deviant until we have determined the correct theory
of legal content, whatever it may be, but we can explore the concept of
deviation without committing ourselves to a definite theory of legal
content.
A judge engages in deviation from the law simpliciter when she reach-
es a result that cannot be supported by any reasonable reading of the
controlling legal authority. Deviation from the law simpliciter occurs
when a judge deviates from a legally valid rule without any higher legal
authority to do so. This is the limit case of deviation—as deviant as devi-
ation can be.
So my topic differs from that of Dworkin and the Kadishes in two
related respects. First, they address departures from legal rules,
whereas I address deviation from the law simpliciter. Second, they
address legally authorized departures, whereas I am interested in cases
of legally unauthorized deviation. The judicial duty that occupies me
in the remainder of this book is the duty to correctly apply the law
simpliciter.

45. This paragraph assumes arguendo that principles sometimes require particular
results, just as rules do. See Dworkin, “The Model of Rules I,” pp. 35–36.
46. Hereafter, unless otherwise noted, I use “deviates from the law” interchangeably
with “deviates from the law, simpliciter.”
46 Limits of Legality: Part I

3.9 DISCRETION VERSUS DEVIATION

Various ideas have been discussed in the jurisprudential literature under

the label of judicial discretion,47 so I must explain the relationship between
discretion as discussed in the literature and deviation as I understand it.
I shall write that a judge enjoys objective discretion with respect to two
incompatible choices that are open to her if and only if the complete set
of valid legal standards permits her to make either choice.
Jurisprudential discussions of judicial discretion have focused on the
question of whether judges ever enjoy objective discretion. Does the law
contain gaps such that cases can arise in which there is no unique, legally
correct answer to the question posed?48 Some think so.49 Others disagree.50
Some believe that the law consists entirely of rules and that cases arise to
which the rules do not apply. Others believe that certain “closure rules”
close all such gaps.51 Still others believe that gaps exist between rules, but
that binding principles always exist to fill those gaps.52 And still others
believe that, even if judges invoke principles, cases arise in which the law
(understood to include both rules and principles) remains indetermi-
nate.53
If the law consists only of rules, then a judge enjoys objective discretion
when valid rules do not dictate a unique result, as when a statute man-
dates a fine between $1,000 and $5,000 and no other rule mandates any-
thing more specific. Again, assuming that law consists only of rules, a judge
also enjoys objective discretion when a case falls within the open texture

47. See, e.g., Marisa Iglesias Vila, Facing Judicial Discretion: Legal Knowledge and Right
Answers Revisited (Boston: Kluwer, 2001); Himma, “Judicial Discretion and the Concept of
Law”; H. L. A. Hart, The Concept of Law, 2nd ed. (Oxford: Oxford University Press, 1994),
pp. 141–47; Keith Hawkins, ed., The Uses of Discretion (Oxford: Oxford University Press,
1992); Aharon Barak, Judicial Discretion, trans. Yadin Kaufmann (New Haven, Conn.: Yale
University Press, 1989); Dworkin, “The Model of Rules I,” pp. 31–39; Ronald Dworkin, “The
Model of Rules II,” in Taking Rights Seriously (Cambridge, Mass.: Harvard University Press,
1978), pp. 68–71.
48. “There is a gap in the law when a legal question has no complete answer.” Joseph
Raz, “Legal Reasons, Sources, and Gaps,” in The Authority of Law: Essays in Law and Morality
(Oxford: Clarendon Press, 1979), p. 70. Do not confuse this kind of gap with the “gap” intro-
duced in Larry Alexander, “The Gap,” Harvard Journal of Law and Public Policy 14 (1991):
695–701, on which see §5.4.
49. See, e.g., Schauer, Playing by the Rules, pp. 222–26; Raz, “Legal Reasons, Sources, and
Gaps.”
50. See, e.g., John Gardner, “Concerning Permissive Sources and Gaps,” Oxford Journal
of Legal Studies 8 (1988): 457–61.
51. Hans Kelsen, “On the Theory of Interpretation,” Legal Studies 10 (1990): 127–35,
p. 132.
52. Neil Duxbury, “Faith in Reason: The Process Tradition in American Jurisprudence,”
Cardozo Law Review 15 (1993): 601–705, p. 614 (attributing to Roscoe Pound the view that
“principles fill in the gaps where the positive law is found wanting”).
53. See, e.g., Stephen J. Burton, Judging in Good Faith (Cambridge: Cambridge Univer-
sity Press, 1992), especially pp. 185–91; Andrew Altman, “Legal Realism, Critical Legal
Studies, and Dworkin,” Philosophy and Public Affairs 15 (1986): 205–35.
Deviating from Legal Standards 47

of applicable rules, as when an ordinance forbids vehicles in the park but

does not specify whether a moped constitutes a vehicle. The law might
provide the judge with guidelines in these kinds of cases, but if these are
merely advisory then she still has discretion—what the Kadishes call “del-
egated discretion.”54 This is not the type of discretion that interests me.
Suppose, however, that the law includes principles as well as rules. In
that case, if legal principles uniquely resolve the question, then the judge
does not enjoy objective discretion. If legal principles do not uniquely
resolve the case, then she still enjoys objective discretion. Again, this is not
the type of discretion that interests me.
Whereas objective discretion exists prior to a decision, the type of dis-
cretion that occupies me exists only when a judge makes a decision. It is
exercised when the full set of valid legal standards clearly apply and
clearly dictate a unique result, but the judge reaches a different result for
some reason. Consider a judge who faces the statute mandating a fine
between $1,000 and $5,000, but who imposes a fine of $500 or $6,000.
His decision is actually forbidden by the statute. He deviates from this
rule.55 If the law consists only of rules, then he deviates from the law sim-
pliciter. Such cases arise even if we assume, with Dworkin, that the law
includes principles. If no other legal standard, such as a principle, permits
the judge to deviate from the statute, then he deviates from the law sim-
pliciter.
The debates about judicial discretion are important, and my project
has implications for them. However, I challenge some of the presupposi-
tions of these debates. The most basic presupposition is that if the law
dictates a unique result, then judges are morally obligated to reach that
result. The greater the number of people who accept this presupposition,
the more tempted some people will be, under certain conditions, to mis-
represent the law. Imagine someone (a lawyer, judge, or commentator) for
whom the following is true:
1. She believes (correctly or not) that her intended audience believes
that, if the law dictates a unique result, then judges are morally
obligated to reach that result.
2. She believes (correctly or not) that her audience believes that the
law requires a particular result in a certain case.
3. She believes with great confidence that this result is unjust or bad
policy.
This individual will want to persuade her audience that the law does
not, in fact, require this result. She might try to show her audience the
errors in their legal reasoning. But in so doing she may succumb to wishful

54. Kadish and Kadish, Discretion to Disobey, pp. 42–44.

55. United States v. Mosley, 965 F.2d 906, 916 (10th Cir. 1992) (holding that a sentenc-
ing judge has no authority to sentence below a statutory mandatory minimum); Alexander
and Sherwin, The Rule of Rules, p. 47 (on “renegade officials”).
48 Limits of Legality: Part I

perception in which the applicable law, which precludes her favored

result, appears to her less determinate than it actually is. I agree with
Frederick Schauer that we should resist this kind of wishful perception:
[W]hatever we might say about well-behaved or anaesthetized dogs, they are
still dogs, and thus still literally within the semantic scope of a “No dogs allowed”
prohibition. It may be that in such circumstances a decision-maker, such as a
judge, should then refuse to apply the rule, or should revise the rule, but the
(arguable) desirability of such an approach, and the (arguable) desirability of
saving the legal system from occasional absurd results, ought not to be dis-
guised in an implausible theory of meaning.56

We should admit, at least to ourselves, that the law is not so indetermi-

nate as to always permit our favored results. That goes for judges, too. The
practical question is, what should judges do in such cases?

3.10 TYPES OF LEGALLY UNAUTHORIZED DEVIATION

Legally unauthorized deviation involves either misapplying or disregarding

applicable sources of law. A court misapplies a legal standard when it
incorrectly presents the standard as a reason to reach a result or incorrectly
treats the standard as a reason to reach the result. A court disregards a legal
standard when it incorrectly fails to present the standard as a reason to
reach a result or incorrectly fails to treat the standard as a reason to
reach a result. Such cases include misapplying or disregarding statutes57
and misapplying or disregarding constitutional provisions.58
With respect to precedent, matters are more complicated. Consider
the following sequence of events. While visiting Noah’s residence in order
to buy something that Noah has advertised, Jacob trips on a loose brick
and suffers injury. He sues Noah for damages. Imagine (implausibly) that
this is a case of first impression in the state. It reaches the state supreme
court, which announces that a homeowner is responsible when someone
on his property suffers an injury due to his negligence. The court upholds
an award of tort damages to Noah.
In the same jurisdiction, Isaac, a dinner guest of Abraham, trips on a
loose brick and suffers injury. Isaac sues Abraham for damages. The rule
relied upon by the supreme court in Jacob’s case would seem to entail an
award for Isaac. But the trial court distinguishes the cases: Jacob was an
invitee (one present for commercial reasons), Isaac a licensee (a social

56. Schauer, Playing by the Rules, p. 59.

57. See, e.g., Joseph P. Bauer, “Addressing the Incoherency of the Preemption Provision
of the Copyright Act of 1976,” Vanderbilt Journal of Entertainment & Technology Law 10
(2007): 1–119, p. 58 (Fourth Circuit “misreads” Act).
58. See, e.g., Frank J. Macchiarola, “Why the Decision in Zelman Makes So Much Sense,”
N.Y.U. Annual Survey of American Law 59 (2003): 459–67, p. 461 (U.S. Supreme Court
“misreads” Establishment Clause in Everson v. Board of Education, 330 U.S. 1 (1947)).
Deviating from Legal Standards 49

guest).59 The trial court announces that property owners owe a higher
duty of care to invitees than licensees. This rule is consistent with the
supreme court’s opinion in Jacob’s case. The trial court rules in favor of
Noah. It thereby deviates from the rule of Jacob’s case, but it announces
and adheres to a new rule that also supports the result in Jacob’s case. The
trial court is legally authorized to deviate from the rule of Jacob’s case.
A judge who properly distinguishes all prior cases does not deviate
from the law simpliciter. Although the prior rule on its own does not sup-
port his result in the instant case, the judge extends the existing rule in
a way that is consistent with all prior cases and with his result in the
instant case. If the extended rule had been applied in all prior cases, then
it would have upheld the results reached in those cases. To deviate is to fail
to apply the law as it exists at the time of decision. Because the process of
properly distinguishing specifies or elaborates the rule, effectively creating
a new rule before deciding the case, proper distinguishing does not consti-
tute true deviation.60
However, it is also possible to distinguish cases improperly. The court
might, for example, fail to distinguish the instant case from some relevant
prior cases. If the trial court had ruled in Noah’s favor without drawing
the invitee/licensee distinction, then it would have deviated from the rule
announced by the supreme court in Isaac’s case. The trial court lacks legal
authority to do so. It would have deviated from the law simpliciter. A court
engages in legally unauthorized deviation when it reaches a result that is
inconsistent with precedent without adequately distinguishing the case.61
An especially brazen lower court could even purport to “reverse” a verti-
cal precedent.62
Deviating and distinguishing precedent are also distinct from reversing
precedent. Courts in the United States often reverse subordinate and hor-
izontal precedents. They are understood to have the legal authority to do
so in many cases. Of course, the overruled precedent itself might have
been legally correct. A federal appellate court could, for example, reverse

59. See William Lloyd Prosser et al., Prosser and Keeton on the Law of Torts, 5th ed.
(St. Paul, Minn.: West., 1984) §§ 57–60. I am assuming, again, that this is a case of first
impression, so let us assume that this authority has not yet been written.
60. It resembles in this respect norm specification. See Henry S. Richardson, “Specifying
Norms as a Way to Resolve Concrete Ethical Problems,” Philosophy and Public Affairs 19
(1990): 279–310.
61. See, e.g., Thomas E. Roberts, “Facial Takings Claims Under Agins-Nectow: A Proce-
dural Loose End,” Hawaii Law Review 24 (2002): 623–55, p. 654 (“Ninth Circuit misreads
Supreme Court precedent”); Ann C. Hodges, “Protecting Unionized Employees against Dis-
crimination: The Fourth Circuit’s Misinterpretation of Supreme Court Precedent,” Employee
Rights and Employment Policy Journal 2 (1998): 123–74.
62. I have found no actual case of a lower court explicitly purporting to reverse a higher
court, as opposed to merely disregarding vertical precedent. See State Oil Co. v. Khan, 118
S. Ct. 275, 284 (1997) (noting in dictum that Court of Appeals was correct to follow prec-
edent with which it disagreed because “it is this Court’s prerogative alone to overrule one of
its precedents”).
50 Limits of Legality: Part I

its own sound interpretation of the Constitution in favor of an incorrect

interpretation. In such cases the reversal constitutes deviation, not because
it is a reversal, but because it is incorrect—it fails properly to apply the
law. Whereas if a lower court disregards vertical precedent, but in so doing
correctly interprets the underlying law, then the court still deviates from
the law. A federal district court that defies the law of its circuit deviates,
even if everyone else agrees that the appellate court is mistaken and the
Supreme Court ultimately vindicates the district court.63

3.11 DEVIATION AND “JUDICIAL ACTIVISM”

I mentioned earlier that I avoid the phrase judicial activism.64 Now I shall
explain how my concept of deviation relates to the ideas denoted by that
phrase. “Judicial activism” and its cognates appear thousands of times
annually in scholarship and news media.65 I think the phrase is meaningful
if carefully defined, but too often it is not.66 Indeed, scholars spend time
just differentiating definitions. Black’s Law Dictionary defines it as “judi-
cial decision-making whereby judges allow their personal views about
public policy, among other factors, to guide their decisions.”67 My deviat-
ing judges do, indeed, “allow their personal views about public policy,
among other factors, to guide their decisions.” But the dictionary defini-
tion also seems to encompass decisions in which a judge allows his per-
sonal views about policy to guide his exercise of discretion, as when a
judge imposes the minimum sentence allowed by law because he objects
to the law on policy grounds. The dictionary calls this a case of judicial
activism but it is not a case of deviation in my sense.
A useful recent study of usage by Keenan Kmiec identifies five core
meanings of judicial activism: “(1) invalidation of the arguably constitu-
tional actions of other branches, (2) failure to adhere to precedent, (3)
judicial ‘legislation,’ (4) departures from accepted interpretive methodol-
ogy, and (5) result-oriented judging.”68 For each of these five practices,
there are at least some instances of the practice that constitute deviation
as I understand it. I shall consider each practice in turn.

63. See Evan H. Caminker, “Why Must Inferior Courts Obey Superior Court Prece-
dents?” Stanford Law Review 46 (1994): 817–73; Paul L. Colby, “Two Views of the Legiti-
macy of Nonacquiescence in Judicial Opinions,” Tulane Law Review 61 (1987): 1041–69.
64. See §1.8.
65. Keenan D. Kmiec, “The Origin and Current Meanings of ‘Judicial Activism,’” Cali-
fornia Law Review 92 (2004): 1441–77, p. 1442 nn.5–6 (Westlaw searches).
66. “‘[J]udicial activism’ is defined in a number of disparate, even contradictory, ways;
scholars and judges recognize this problem, yet persist in speaking about the concept with-
out defining it. Thus, the problem continues unabated: people talk past one another, using
the same language to convey very different concepts.” Ibid., p. 1443.
67. Bryan A. Garner, ed., Black’s Law Dictionary, 8th ed. (St. Paul, Minn.: Thomson/
West, 2004), p. 862.
68. Kmiec, “The Origin and Current Meanings of ‘Judicial Activism,’” p. 1444.
Deviating from Legal Standards 51

Invalidating the actions of other branches of government constitutes

deviation if those actions are, in fact, constitutional. However, many
actions are both arguably constitutional and arguably unconstitutional.
Kmiec’s language suggests that a court is activist if it invalidates such
actions. To avoid activism, in Kmiec’s terms, courts should follow a pre-
sumption of constitutionality, invalidating an action only if it is not even
arguably constitutional.69 In contrast, a court that invalidates an action
that is arguably constitutional does not deviate if the action is, in fact,
unconstitutional. Kmiec calls a court activist if it invalidates an action that
is, in fact, unconstitutional, if it is arguably constitutional. The charge of
activism is apt when there is some uncertainty about constitutionality.
The charge of deviation is apt when there is no uncertainty.
Kmiec’s second meaning of judicial activism involves “failure to adhere
to precedent.” Deviation often takes this form, but a departure from prec-
edent constitutes deviation only if it is legally unauthorized. As discussed
earlier, all departures from vertical precedent fall in this category.70 So do
departures from horizontal precedent unless the precedent is explicitly
reversed. But Kmiec’s definition implies that any departure from horizon-
tal precedent constitutes activism, even if (perhaps especially if) the court
reverses the precedent. So the second meaning of activism encompasses
more than deviation.
Judicial “legislation,” Kmiec’s third meaning, occurs when judges change
legal standards without legal authorization. Such decisions always count
as deviant in my terms. For example, courts are sometimes accused of
creating new constitutional rights and ignoring existing ones.71 Of course,
commentators disagree about whether a certain right is “new” or not and
I shall not enter these debates. The important point is that if one believes
that a court has created a new right or ignored an existing one, without
legal authorization, then one believes that the court has deviated.
Do courts deviate when they “depart from accepted interpretive meth-
odology” (Kmiec’s fourth definition)? It depends. If, by accepted, Kmiec
means “mandatory,” then the answer is affirmative. If, however, a method-
ology can be accepted but not mandatory, then we must conclude that
courts deviate only if they depart from methodologies that they are
required to use. For example, the standard canons of statutory construc-
tion are seen as advisory, not mandatory.72
Finally, Kmiec’s fifth definition of activism as “result-oriented judging”
is perhaps the closest to my definition of deviation. He states that a judge

69. Compare James B. Thayer, “The Origin and Scope of the American Doctrine of
Constitutional Law,” Harvard Law Review 7 (1893): 129–56.
70. See §3.10.
71. See, e.g., Antonin Scalia, A Matter of Interpretation: Federal Courts and the Law
(Princeton, N.J.: Princeton University Press, 1997), pp. 37–47.
72. Chickasaw Nation v. United States, 534 U.S. 84, 94 (2001) (canons of construction
are not mandatory rules); Connecticut National Bank v. Germain, 503 U.S. 249, 253 (1992).
52 Limits of Legality: Part I

engages in result-oriented judging if and only if her decision “departs from

some ‘baseline’ of correctness,” and she makes the decision for an “ulterior
motive.”73 If correctness means “legal” correctness, then it is true by defini-
tion that deviant decisions are those that depart from a baseline of cor-
rectness, so all cases of result-oriented judges are cases of deviation. If
ulterior motives include the desire to avoid normatively undesirable
results, then the cases of deviation that interest me are all cases of result-
oriented judging. Therefore, one might describe my central topic as “the
ethics of judicial activism, understood as result-oriented judging.”

3.12 JUDICIAL MENTAL STATES

A judge who deviates from the law does so either purposely, knowingly,
recklessly, negligently, or innocently.74 Much deviation is merely negli-
gent: the judge tries to adhere, but he misreads sources that a reasonable
judge would have read correctly. In 1960, a magistrate was charged with
a number of errors, including releasing a defendant on parole who was
ineligible because of a prior felony conviction. The magistrate was not
disciplined for these errors because it was determined that he had hon-
estly misunderstood the scope of his authority (which is not to say that
his misunderstanding was reasonable).75
Deviation is occasionally reckless. One judge dismissed the first crimi-
nal charge he heard on the bench simply because he had once “promised
himself” that he would do so.76 Others have decided cases on the basis of
coin flips.77 One judge polled the courtroom audience on whether to grant
a request to file a criminal complaint.78 Such judges decide recklessly. If
they reach legally incorrect results, then they deviate recklessly.79

73. Kmiec, “The Origin and Current Meanings of ‘Judicial Activism,’” p. 1476.
74. A judge deviates innocently if and only if a reasonable judge could not be expected
to understand the law in question. For example, if laws are too complicated, then judges will
sometimes deviate innocently. See Lon L. Fuller, The Morality of Law, 2nd rev. ed. (New
Haven, Conn.: Yale University Press, 1969), p. 36.
75. Gerald Stern, “Is Judicial Discipline in New York State a Threat to Judicial Indepen-
dence?” Pace Law Review 7 (1987): 291–388, p. 315. The California Supreme Court has held
that mere negligence does not suffice for discipline. Broadman v. Commission on Judicial
Performance, 959 P.2d 715, 720–21 (Cal. 1998).
76. In re DeRose, 1980 Annual Report 181, 183 (New York State Commission on Judi-
cial Conduct, Nov. 13, 1979).
77. McCartney v. Commission on Judicial Quality, 526 P.2d 268 (Cal. 1974); In re
Daniels, 340 So.2d 301 (La. 1976).
78. In re Friess, 1984 Annual Report 84 (New York State Commission on Judicial Con-
duct, Mar. 30, 1983).
79. Of course, using these arbitrary methods is unethical even when it happens to yield
a correct decision. They are unethical even when used to make a decision within the range
of the judge’s legal discretion, as when a judge decided between a twenty-day and thirty-day
sentence on the basis of a coin flip. Ibid.
Deviating from Legal Standards 53

The cases that interest me, however, involve judges who knowingly
deviate in order to avoid results of which they disapprove.80 Within that
category we can distinguish between express and tacit deviation. A judge
deviates expressly if and only if she admits it publicly. Otherwise she
deviates tacitly.
Most judicial deviation is tacit, but express deviation occasionally occurs,
making for considerable drama. A well-publicized episode arose in the Cal-
ifornia case of Morrow v. Hood Communications, Inc.81 Morrow involved a
stipulated reversal, in which opposing parties settle a case after a trial court
has made a final ruling on the merits. The loser pays the winner to consent
to have the ruling reversed by an appeals court. This practice had been
endorsed by the California Supreme Court in the case of Neary v. Regents of
the University of California.82 In Morrow, Presiding Justice J. Anthony Kline
of the California Court of Appeal opened his dissenting opinion as follows:
There are rare instances in which a judge of an inferior court can properly
refuse to acquiesce in the precedent established by a court of superior juris-
diction. This is, for me, such an instance.
I acknowledge that the opinion of the California Supreme Court in
Neary v. Regents of University of California requires that the motion before
us be granted. I would deny the motion, however, because I cannot as a mat-
ter of conscience apply the rule announced in Neary.
I do not refuse to acquiesce in Neary because I believe the opinion is
analytically flawed and empirically unjustified, though, as I have elsewhere
explained at length, that is my view. My refusal is instead based on my
deeply felt opinion that the doctrine of stipulated reversal announced in
Neary—a doctrine employed in no other jurisdiction in this nation and
unanimously repudiated by the Supreme Court of the United States—is
destructive of judicial institutions.83

Another example of express deviation, perhaps even more shocking, is

that of Justice Tom Parker of the Alabama Supreme Court, whose platform

80. A word about purposeful deviation is in order. Few judges who deviate do so pur-
posely under that description. Rarely does a judge act with the “conscious object” of deviat-
ing from the law. One exception is the traitorous judge who intends to undermine the
reputation of the judiciary and who therefore makes a decision because it is deviant. Even
corrupt judges deviate for their own benefit. But most judges who knowingly deviate do not
do so for such reasons. They deviate in order to avoid objectionable results. In the cases that
interest me, fulfilling this purpose requires decisions that happen to be deviant.
81. See, e.g., Morrow v. Hood Communications, Inc., 69 Cal. Rptr. 2d 489, 494–95 (Cal.
Ct. App. 1997) (Kline, P.J., dissenting). For discussions of the case see, e.g., Sambhav N.
Sankar, “Disciplining the Professional Judge,” California Law Review 88 (2000): 1233–80;
Pamela S. Karlan, “Two Concepts of Judicial Independence,” Southern California Law Review
72 (1999): 535–58; Stephen C. Yeazell, “Good Judging and Good Judgment,” Court Review
35 (1998): 8–10; Howard Mintz, “Disciplinary Case against Judge Raises Legal Uproar,” San
Jose Mercury News, July 11, 1998, p. 1A.
82. Neary v. Regents of the University of California, 834 P.2d 119 (Cal. 1992).
83. Morrow v. Hood Communications, Inc. at 494–95. Note that Kline’s deviation
occurred only in a dissenting opinion and thus had no legal force.
54 Limits of Legality: Part I

during his 2006 campaign for chief justice centered on the fact that he,
unlike his colleagues, was willing to ignore the U.S. Supreme Court when
he believed it to be wrong.84

3.13 PRIVATE REASONS TO DEVIATE

In the next chapters, I shall examine various reasons that a judge might
knowingly deviate from the law simpliciter. Before closing this chapter,
I shall consider private reasons to deviate, mainly incentives.
An incentive obviously exists when someone offers the judge a bribe to
deviate: he will receive the bribe if he rules in favor of a specified party,
despite the fact that a dispositive legal case has been or will be made
against the latter.85 Even in the absence of a bribe, a judge might have
private reasons to deviate in favor of a relative or friend who is a party to
a case and is disfavored by the law. Of course, codes of judicial conduct
require judges to recuse themselves in cases of conflict of interest and
they face sanctions for failing to do so.86 But a judge who can avoid such
sanctions could have a private reason to hear the case and to deviate in
favor of his relative.
More common than cases involving relatives or friends are those in
which the judge develops a personal like or dislike for a party and a con-
comitant desire for her to win or lose, respectively. If the law requires the
opposite result, then the judge’s feelings give him a private reason to
deviate.
Even when the case does not involve bribes, conflicts of interest, per-
sonal affection, or antipathy, judges can still have incentives to deviate.
Identifying and applying the correct law can be difficult and time-
consuming. Adhering to the law requires knowing the law and applying it
to the facts of the case. A judge could simplify his job in various ways: by
ignoring the law and deciding on some simpler basis, by writing a superfi-
cial opinion, or by deciding difficult cases summarily, without writing an
opinion.87 Even if the judge happens to reach the legally correct result,

84. David White, “Parker Says He’s Willing to Defy High Court,” Birmingham (Ala.)
News, May 26, 2006, p. 1C.
85. Bracy v. Gramley, 520 U.S. 899, 909 (1997) (holding that petitioner, who was con-
victed before a judge who was himself later convicted of taking bribes, showed “good cause”
for discovery on claim of actual bias).
86. Model Code of Judicial Conduct, Canon 3E (2007); Code of Conduct of United States
Judges, Canon 3C.
87. He could, for example, grant a sua sponte motion for summary judgment. See Arthur
R. Miller, “The Pretrial Rush to Judgment: Are the ‘Litigation Explosion,’ ‘Liability Crisis,’
and Efficiency Cliches Eroding Our Day in Court and Jury Trial Commitments?” New York
University Law Review 78 (2003): 982–1134, pp. 1006, 1052, 1055–56 (widespread federal
practice of judges employing Federal Rule of Civil Procedure 16, providing case manage-
ment authority, to effectively implement summary judgment merits review sua sponte).
Deviating from Legal Standards 55

these shortcuts involve deviant legal reasoning—reasoning that does not

aim at reaching the legally correct result.
This book contains little more about private reasons to deviate. I shall
be concerned mainly with impartial, moral reasons to deviate.88 I shall
argue that judges have pro tanto moral reasons to deviate whenever the
legally required result is unjust or otherwise objectionable as a matter of
public policy. These reasons are often outweighed by stronger reasons to
adhere, but that story must await part II.

88. Introduced in chapter 6.

4

The Legal Duties of Judges

4.1 ADHERENCE RULES

Do judges have a legal duty to apply the law correctly? To ask this ques-
tion is not to query the tautology: does the law require judges to do what
the law requires judges to do? That the law requires certain results does
not logically entail that any particular individual has a legal duty to
decide cases accordingly. It might seem that the concept of judge entails
a legal duty to apply the law, but this is incorrect. Asking whether a
judge has a legal duty to apply the law is like asking whether a police
officer has a legal duty to knock before entering your apartment or
whether a physician has a legal duty to perform an emergency tracheot-
omy on a fellow airline passenger. These are questions about the legal
duties of particular professionals. The concept of judge does not entail a
legal duty to apply the law any more than the concept of doctor entails a
legal duty to perform tracheotomies. In each of these cases we can learn
the answer only by studying what the law demands of these profes-
sionals.
Judges have various professional duties while they remain on the bench.
These include duties to oversee court proceedings, to maintain decorum,
to study the law, to rule on motions, to examine evidence presented, to
answer questions from jurors, to decide cases, et cetera.1 These duties can
be expressed in adjudication rules—secondary rules addressed to adjudica-
tors.2 Some adjudication rules are codified in constitutions, statutes, codes
of judicial conduct, and case law.3 They impose various requirements:
judges must appear in court when assigned;4 they must recuse themselves

1. Although they are not enumerated, these are presumably some of the duties to which
the Code of Conduct for federal judges refers when it states that “a judge should perform
the duties of the office . . . diligently.” Code of Conduct of United States Judges, Canon 3.
2. They contrast with primary rules addressed to ordinary legal subjects. H. L. A. Hart,
The Concept of Law, 2nd ed. (Oxford: Oxford University Press, 1994), pp. 93–94.
3. Although some writers would not classify rules found only in case law as codified, the
classification serves my purposes.
4. Model Code of Judicial Conduct, Canon 3B(1) (1990) (“A judge shall hear and decide
matters assigned to the judge except those in which disqualification is required”).

56
The Legal Duties of Judges 57

in cases that present conflicts of interest;5 they must not practice law
while sitting on the bench,6 et cetera.
Are judges legally obligated to obey codified adjudication rules? Judges
have many of the same legal duties as the rest of us, certainly: to pay
income taxes, to testify under subpoena, to care for dependent children,
to fulfill contractual obligations, and so forth. Some scholars believe that
judicial duties, likewise, are legal duties.7 My specific question is whether,
as some believe, judges have a legal duty to adhere to the law simpliciter.8
Justice Cardozo appears to think so: “Judges have, of course, the power,
though not the right, to ignore the mandate of a statute, and render judg-
ment in despite of it. They have the power, though not to right, to travel
beyond the walls of the interstices, the bounds set to judicial innovation
by precedent and custom. Nonetheless, by that abuse of power, they vio-
late the law.” 9
A judicial duty to adhere is entailed by the legal validity of certain
adherence rules: adjudication rules that require judges to decide cases in
accordance with certain decision rules. Adherence rules are mandatory
rules: rules that “when accepted, furnish reasons for action simply in
virtue of their existence qua rules.”10 Adherence rules, if authoritative,
provide judges with reasons to adhere to the law.11

5. Ibid., Canon 3E(1).

6. Ibid., Canon 4G.
7. Darrell L. Keith, “The Court’s Charge in Texas Medical Malpractice Cases,” Baylor
Law Review 48 (1996): 675–814, p. 778 (trial judge has a “legal duty to preside over the
trial”); Joseph Raz, “Law and Value in Adjudication,” in The Authority of Law (Oxford: Clar-
endon Press, 1979), p. 197 (discussing judges’ official duties as “legal duties”).
8. Philip Hamburger, “Law and Judicial Duty,” George Washington Law Review 72
(2003): 1–41, p. 24 (“to the extent their office was defined by law and their oaths were
imposed by law, the judges could be considered legally obliged to decide in accord with
law”); Stephen J. Burton, Judging in Good Faith (Cambridge: Cambridge University Press,
1992), p. 35 (“judges do not fulfill their legal duty if they act only on parts of the law with
which they agree”). To be more precise, one should say that a judge has a conditional legal
duty: if he chooses to decide a case, then he has a legal duty to decide it according to law,
when the law provides an answer. This formulation reflects the fact that judges may be
legally permitted to avoid deciding any given case. Judges are legally permitted to recuse
themselves in certain cases. Moreover, they have no legal duty to remain on the bench. Judi-
cial service is not indentured servitude. Judges are always legally permitted to resign.
9. Benjamin N. Cardozo, The Nature of the Judicial Process (New Haven, Conn.: Yale
University Press, 1921), p. 129. Neil MacCormick states that “the judge has a duty” to reach
the legally required result. Neil MacCormick, Legal Reasoning and Legal Theory (Oxford:
Clarendon Press, 1978), p. 33. See also Sambhav N. Sankar, “Disciplining the Professional
Judge,” California Law Review 88 (2000): 1233–80, p. 1240 (“[T]he integrity of the legal
system relies upon each individual judge’s adherence to precedent, compliance with orders
from superior courts, and respect for the validity of prior judgments”).
10. Frederick Schauer, Playing by the Rules: A Philosophical Examination of Rule-Based
Decision-Making in Law and in Life (Oxford: Oxford University Press, 1991), p. 5.
11. Adherence rules are entirely separate from conduct rules addressed to legal subjects
who are not judges. A judge who convicts someone of income tax evasion in a bench trial
does not thereby “obey” the tax code as he does when he submits his own tax return.
58 Limits of Legality: Part I

My arguments in this book are consistent with the view that every
judge has a pro tanto legal duty to decide according to law every case that
he chooses to decide. I shall now consider arguments that support this
proposition because some important objections to my ultimate position
depend upon it.12 Three issues seem relevant to the question whether
judicial duties, such as the duty to adhere, are legal duties. First, do recog-
nized sources of law include, or incorporate by reference, any adherence
rules? Second, does the law provide sanctions for judges who break the
rules? Third, are sanctions imposed, in practice?

4.2 CODIFIED ADHERENCE RULES

Adherence rules are codified in the two main codes of judicial conduct
in the United States: the Model Code of Judicial Conduct of the American
Bar Association (ABA) and the Code of Conduct for United States Judges
adopted by the Judicial Conference of the United States. The 1990 revi-
sion of the ABA Model Code states, “A judge shall respect and comply with
the law and shall act at all times in a manner that promotes public confi-
dence in the integrity and impartiality of the judiciary,”13 in which law is
defined as “court rules as well as statutes, constitutional provisions and
decisional law.”14 The Model Code also states, “A judge shall be faithful to
the law and maintain professional competence in it. A judge shall not be
swayed by partisan interests, public clamor or fear of criticism.”15 The
2007 revision of the Model Code states, “A judge shall uphold and apply
the law, and shall perform all duties of judicial office fairly and impartially.”16
The Commentary to this section reads, “Although each judge comes to
the bench with a unique background and personal philosophy, a judge
must interpret and apply the law without regard to whether the judge
approves or disapproves of the law in question.”17
Several provisions of the Code of Conduct for United States Judges also
address the adherence duty. Canon 2A states, “A judge should respect and
comply with the law and should act at all times in a manner that pro-
motes public confidence in the integrity and impartiality of the judiciary.”18
Canon 3A(1) states, “A judge should be faithful to and maintain pro-
fessional competence in the law, and should not be swayed by partisan

12. See chapter 10.

13. Model Code of Judicial Conduct, Canon 2A (1990). The 2007 revision of the Model
Code states, “A judge shall comply with the law, including the Code of Judicial Conduct.”
Model Code of Judicial Conduct, Rule 1.1 (2007).
14. Model Code of Judicial Conduct, Definitions (1990).
15. Ibid., Canon 3B(2).
16. Ibid., Rule 2.2.
17. Ibid., Comment on Rule 2.2.
18. Code of Conduct of United States Judges, Canon 2A.
The Legal Duties of Judges 59

interests, public clamor, or fear of criticism.”19 Also noteworthy is the

commentary to Canon 1:
Deference to the judgments and rulings of courts depends upon public confi-
dence in the integrity and independence of judges. The integrity and independ-
ence of judges depend in turn upon their acting without fear or favor. Although
judges should be independent, they should comply with the law, as well as the
provisions of this Code. Public confidence in the impartiality of the judiciary is
maintained by the adherence of each judge to this responsibility. Conversely,
violation of this Code diminishes public confidence in the judiciary and there-
by does injury to the system of government under law.20 (emphasis added)

The Model Code has been adopted, in whole or in part, in forty-nine of

the fifty states,21 usually including provisions stating that “judges shall be
faithful to the law.” 22 These codes have not, however, been adopted dire-
ctly by state legislatures. Some have been drafted by agencies or judicial
councils which themselves enjoy statutory or constitutional authoriza-
tion.23 Although state codes of judicial conduct look very much like state
statutes, the former have not actually been incorporated into the latter.
Rather, the high court of each state has issued an order containing a code
of judicial conduct. Such an order might be sufficient to give a code legal
force, absent statutory incorporation, but it might not be.
Federal judges are not bound by these state codes. Moreover, at the
federal level, the Judicial Conference of the United States holds no spe-
cific statutory grant of authority to enact binding ethical rules.24 The Code
of Conduct states, “The Code may . . . provide standards of conduct for
application in proceedings under the Judicial Councils Reform and Judi-
cial Conduct and Disability Act of 1980. . . .” But the Code of Conduct
cannot thereby bestow legislative authority upon itself in this area. Fed-
eral legislation provides conditions under which judges are required to
disqualify themselves but it does not incorporate any other provisions of
the Code of Conduct.25 Therefore, one could deny that federal judges have

19. Ibid., Canon 3A(1).

20. Ibid., Commentary to Canon 1.
21. Leslie W. Abramson, “Appearance of Impropriety: Deciding When a Judge’s Impar-
tiality ‘Might Reasonably Be Questioned,’” Georgetown Journal of Legal Ethics 14 (2000):
55–102, p. 55. Montana is the holdout.
22. See, e.g., California Rules of Court, Appendix Division II, Code of Judicial Ethics,
Canon 3 (2006); New York CLS Judicial Appendix, Code of Judicial Conduct, Canon 100.3
(2006).
23. See, e.g., N.Y. Const., Art. VI, § 22 (2008) (establishing commission on judicial
conduct).
24. See 28 U.S.C. § 331 (establishing Judicial Conference of the United States); In re
Cargill, 66 F.3d 1256, 1267 (1st Cir. 1995) (Campbell, J., dissenting) (noting that although
“the Judicial Conference of the United States, which adopted the [Code, does not] hold a
specific statutory grant of authority to enact binding ethical rules,” it should still be accorded
“great persuasive weight”); Debra Lyn Bassett, “Judicial Disqualification in the Federal
Courts,” Iowa Law Review 87 (2002): 1213–56, p. 1230.
25. 28 U.S.C. §455 (disqualification of justice, judge, or magistrate judge).
60 Limits of Legality: Part I

a legal duty to obey the adherence rules contained in the Code of Conduct.
Circuit Judicial Councils are now authorized to investigate allegations of
judicial misconduct, but they have narrow disciplinary powers.26
In addition to codes of judicial conduct, there are some other putative
sources of legal authority for a judicial adherence duty. Judicial opinions
recognize that courts have a duty to apply the law or are “bound by the
law.” 27 One may also infer that a judge believes in a duty to adhere to the law
if he claims to have adhered in a certain case despite expressing disagreement
or regret over what it “requires” him to do.28 In such cases one could argue
that a duty to adhere forms part of the reasoning necessary to the decision if
the judge claims to adhere to the law, despite the fact that she disagrees with
the legally required result and “wishes” that she could reach a different one.
Finally, some judicial oaths could be read as implying a duty to obey
adherence rules. Mississippi judges, for example, swear that they will

26. Steven Lubet, “Judicial Discipline and Judicial Independence,” Law and Contempo-
rary Problems 61 (1998): 59–74, p. 59. See also 28 U.S.C. §§ 331, 332, 372, 604.
27. See, e.g., Maryland St. Dept. of Educ., Div. of Rehabilitation Servs. v. United States
Dept. of Veterans Affairs, 98 F.3d 165, 168 (4th Cir. 1997) (“Courts are charged with the duty
to apply the law that Congress enacted”); Bittaker v. Enomoto, 587 F.2d 400, 402 n. 1 (9th
Cir. 1978) (“It is . . . a district court’s duty to apply the law of the appropriate circuit to all
persons presenting claims within its jurisdiction”); Prima Paint Corp. v. Flood & Conklin Mfg.
Co., 388 U.S. 395, 406 (1967) (“Federal courts are bound to apply rules enacted by Congress
with respect to matters . . . over which it has legislative power”); Jackson v. Lykes Bros. S.S.
Co., No. 575, 386 U.S. 731, 737 (1967) (Stewart, J., dissenting) (“It is our duty to apply the
law, not to repeal it”); United States v. Mirsky, 17 F.2d 275, 276 (S.D.N.Y. 1926) (court has
“duty to apply the law as it is now written”); States v. Schooner Peggy, 1 Cranch 103, 109
(1801) (“[T]he court must decide according to existing laws”); United States v. Callender, 25
F. Cas. 239, 257 (C.C.D. Va. 1800) (“No position can be more clear than that all the federal
judges are bound by the solemn obligation of religion, to regulate their decisions agreeably
to the constitution of the United States, and that it is the standard of their determination in
all cases that come before them”).
28. See, e.g., Jack B. Weinstein, “Every Day Is a Good Day for a Judge to Lay Down His
Professional Life for Justice,” Fordham Urban Law Journal 32 (2004): 131–70 (praising judg-
es who criticize the law); Charles Fried, “Scholars and Judges: Reason and Power,” Harvard
Journal of Law and Public Policy 23 (2000): 807–32, p. 811 (“The judge is also free to criticize
the law as she finds it—and there are some notable examples of judges doing just that while
accepting and carrying on in a course of decision they regret or even deplore”); Gerald
Gunther, Learned Hand: The Man and the Judge (Cambridge, Mass.: Harvard University
Press, 1998), p. 149 (“[B]owing to precedent did not prevent [Hand] from expressing sharp
and thoughtful criticism of the prevailing law, or from suggesting a better approach”); United
States v. Shonubi, 895 F. Supp. 460 (E.D.N.Y. 1995) (criticizing Second Circuit rule of
automatic obstruction of justice enhancement under section 3C1.1 of the Sentencing Guide-
lines, but imposing enhancement); United States v. Tropiano, 898 F. Supp. 90 (E.D.N.Y. 1995)
(following appellate mandate despite serious reservations); United States v. Ekwunoh, 888 F.
Supp. 369 (E.D.N.Y. 1994) (noting that it is important to consider mens rea in drug sentenc-
ing, despite court of appeals’ instruction to disregard defendant’s reasonable belief); United
States v. Isgro, 974 F.2d 1091, 1093–94 (9th Cir. 1992) (criticizing rule of U.S. Supreme
Court that prosecutors have no duty to present substantial exculpatory evidence to the
grand jury, as stated in United States v. Williams, 504 U.S. 36 (1992)); United States v. Orjuela,
809 F. Supp. 193, 197 (E.D.N.Y. 1992) (“Regrettably, [Williams] compels the conclusion that
this court lacks power to remedy the government’s apparent abuse of its power”).
The Legal Duties of Judges 61

perform their duties “agreeably to the Constitution of the United States

and the Constitution and laws of the State of Mississippi.”29 One could
read this language as a promise to apply the law correctly, in which case
one could argue that judges who take the oath acquire a moral obligation
to apply the law regardless of whether they have a legal obligation to do
so. But one could instead read the language as a promise to obey the law
on the job, which is a different matter.

4.3 SANCTIONS

Even if Congress and state legislatures decided to codify codes of judicial

conduct into statutes, someone could still insist that no judge actually has
a legal duty to adhere to the law unless he is credibly threatened with
formal sanctions for deviating. This suggestion might reflect a discredited
“command model” of law according to which each legal duty must have
an associated sanction.30 At least since H. L. A. Hart, legal theorists
have recognized that the law can impose genuine legal duties without
sanctions, just as it can grant legal powers without any sanctions being
specified.31 Nevertheless, it is useful to determine what formal sanctions
the law provides for judges who deviate. Legal duties do not logically
entail sanctions, but whether sanctions are provided may still be relevant
to the existence of legal duties.
Some commentators assert that encouraging adherence to the law is a
legitimate reason to discipline judges,32 but American law has a long tra-
dition of reluctance to sanction judges for decisions made on the merits of
the case. In 1872, the U.S. Supreme Court declared that “[a] judge shall
be free to act upon his own convictions without apprehension of personal
consequence to himself.” 33 Jeffrey Shaman, a leading scholar of judicial
ethics, observes that “[i]mposing discipline upon a judge for an incorrect
legal ruling is an extremely sensitive issue because it comes closer than
any other ground of discipline to threatening judicial independence.” 34
Another eminent ethicist, Stephen Lubet, argues that “judicial indepen-
dence is most gravely threatened when judges face sanctions for ‘decisional

29. Miss. Const. Ann., Art. 6, §155 (2008).

30. See John Austin, The Province of Jurisprudence Determined, ed. Wilfrid E. Rumble
(New York: Cambridge University Press, 1995), p. 29 (first published 1832). One might also
infer that the importance of the duty is proportional to the severity of the sanction, on the
command model.
31. Hart, The Concept of Law, ch. 3.
32. “A . . . reason for imposing discipline is to enforce adherence to legal norms. . . . ,”
Sankar, “Disciplining the Professional Judge,” p. 1239.
33. Bradley v. Fisher, 80 U.S. 335, 347 (1872). See also Jeffrey M. Shaman, “Judicial
Ethics,” Georgetown Journal of Legal Ethics 2 (1988): 1–20, p. 11 (“In this nation there has
been a long-standing belief in judicial independence so that judges can be free to decide
cases without fear of retribution or the need to curry favor”).
34. Shaman, “Judicial Ethics,” p. 8.
62 Limits of Legality: Part I

conduct,’ which may be defined as discipline based on the merits of a

ruling,” 35 although he concedes that even some purely decisional conduct
should be punished.36
The mere fact that a judge knowingly deviates from the law does not
subject him to civil or criminal liability. Of course, an act of deviation can
constitute a criminal act or give rise to a civil cause of action for other
reasons: a judge who deviates from the law in furtherance of some other
crime could end up in prison. But knowingly violating codes of judicial
conduct does not, per se, constitute a crime.37
It is also well-settled law that judges enjoy absolute immunity from
civil lawsuits, “so long as the contested action was judicial in nature and
was not taken in the complete absence of jurisdiction.” 38 Judges are some-
times sued for their decisions, but no civil judgment can be entered against
them on the basis of their decisional conduct, even when egregiously
deviant.39
Judges are, however, subject to various formal sanctions for violating
codes of judicial conduct, including reprimand, private or public censure,
suspension, impeachment, and removal.40 But even mild disciplinary
sanctions are imposed only in cases involving criminal activity, intoxica-
tion on the job, judicial patronage, inappropriate courtroom demeanor,
racist or sexist remarks, sexual harassment, or similar misconduct.41
Another example is failing to disqualify oneself when required, as when
the judge has had ex parte communications with one of the parties.42

4.4 LEGAL ERROR AS JUDICIAL MISCONDUCT

When, if ever, does deviation from the law constitute judicial miscon-
duct? It is commonly said that “mere legal error” does not rise to that
level.43 The commentary to Rule 2.2 of the Model Code states, “When

35. Lubet, “Judicial Discipline and Judicial Independence,” p. 59.

36. Ibid., p. 72.
37. “[T]he Code is not designed or intended as a basis for civil liability or criminal pros-
ecution.” Commentary, Code of Conduct of United States Judges.
38. Stump v. Sparkman, 435 U.S. 349, 356–57 (1978). See also Lewis v. Green, 629 F.
Supp. 546 (DC Dist. Col. 1986) (28 U.S.C. § 453, which sets forth oath of allegiance to
Constitution taken by federal judges, does not create substantive cause of action against
federal judges for violating oath by acting contrary to Constitution); James J. Alfini et al.,
Judicial Conduct and Ethics, 4th ed. (Newark, N.J.: LexisNexis, 2007).
39. But see Abimbola A. Olowofoyeku, Suing Judges: A Study of Judicial Immunity
(Oxford: Clarendon Press, 1993).
40. See generally, Alfini et al., Judicial Conduct and Ethics.
41. Ibid.
42. See, e.g., In re Cooks, 694 So.2d 892 (La. 1997) (judge engages in punishable mis-
conduct by failing to recuse herself after extensive ex parte communications with a party).
43. “While the courts have often said that ‘mere’ legal error does not amount to judicial
misconduct, that does not mean that legal error can never constitute misconduct.” Shaman,
“Judicial Ethics,” p. 8.
The Legal Duties of Judges 63

applying and interpreting the law, a judge sometimes may make good-
faith errors of fact or law. Errors of this kind do not violate this Rule.”44
The California Supreme Court has held that “[m]ere legal error, without
more, . . . is insufficient to support a finding that a judge has violated the
Code of Judicial Ethics and thus should be disciplined.”45 Gerald Stern
asserts that “investigations of judges are not warranted merely because the
judges abused their discretion or otherwise committed judicial error,”46
and he notes that “[f ]or most of the past ten decades, arbitrary conduct in
court that deprived litigants or other persons of their guaranteed rights,
with few exceptions, has not been a basis for discipline.”47 Shaman writes
that “[u]nder the modern judicial disciplinary system, judges may not be
censured or penalized for making erroneous or unpopular decisions.”48
The normal remedy for legal error is the appeal process.49 Most com-
plaints filed with state judicial conduct commissions allege errors of law
or fact and are dismissed as more properly handled via appeal.50 However,
the Model Code and Code of Judicial Conduct contain some provisions that
support the idea that deviation can constitute judicial misconduct.51 So
why do leading scholars insist that mere legal error does not constitute
judicial misconduct?52
At a minimum, these commentators are asserting that innocent, non-
negligent deviation does not constitute misconduct. Perhaps they are also
asserting that deviation does not constitute misconduct if it is merely neg-
ligent. It is not obvious to me that this assertion is literally correct, but I
shall not pursue the issue.53 It is still less obvious that the exemption extends
to reckless or knowing deviation. Several commentators assert that inten-
tional deviation manifests unfitness for judicial office. Stern, for instance,
suggests that judicial independence “was not intended to afford protection

44. Model Code of Judicial Conduct, Commentary to Rule 2.2 (1990).

45. Oberholzer v. Commission on Judicial Performance, 975 P.2d 663, 680 (Cal. 1999).
46. Gerald Stern, “Is Judicial Discipline in New York State a Threat to Judicial Indepen-
dence?” Pace Law Review 7 (1987): 291–388, p. 314. Abuse of discretion could involve, for
example, a trial court interpreting in an unreasonable way the facts presented.
47. Ibid., p. 322.
48. Shaman, “Judicial Ethics,” p. 12.
49. Ibid., pp. 8–9; Lubet, “Judicial Discipline and Judicial Independence,” p. 74 (“It is
widely understood that disciplinary complaints should not be misused as an alternative to
appeal by disgruntled litigants”). See also Harrod v. Illinois Courts Commission, 372 N.E.2d
53 (Ill. 1977); In re Mattera, 168 A.2d 38 (N.J. 1961); Murtagh v. Maglio, 9 A.D.2d 515 (N.Y.
1960).
50. Cynthia Gray, “The Line between Legal Error and Judicial Misconduct: Balancing
Judicial Independence and Accountability,” Hofstra Law Review 32 (2004): 1245–80, p.
1245.
51. See §4.2.
52. See quotations, above.
53. “In some cases it is also said that legal error can amount to misconduct if a judge
should have known the law. . . .” Shaman, “Judicial Ethics,” p. 9.
64 Limits of Legality: Part I

to judges who ignore the law. . . .”54 According to Shaman, “[i]ntentional

refusals to follow the law are . . . [a] manifestation of unfitness for judicial
office.”55 I shall address intentional deviation at greater length in §4.4.2.
Some judges, and other officials assigned to discipline judges, appear to
agree with Stern and Shaman to some extent. Although mere legal error
does not amount to misconduct, judges have been disciplined for “obvious”
legal error, “egregious legal error, legal error motivated by bad faith, or a
continuing pattern of legal error. . . .”56 Legal error, especially a pattern of
it, has been held to demonstrate unfitness for judicial office.57 Discipline
has been imposed pursuant to Canons 2 and 3 of the Model Code.58 For
example, judges in New York have been disciplined for ignoring the weight
of the evidence, for intentionally disregarding the law, and for showing bias
toward certain lawyers or parties.59 They have been censured for showing
bias in favor of tenants.60 Under law, New York judges who abuse their
power or disregard fundamental rights can be removed from office.61
A Texas judge was censured for stating that he gave less than a life sentence
to a murderer because his victims had been gay men “cruising” for teenage
boys.62 A municipal judge in South Carolina was removed from office for
issuing arrest warrants based solely on his own daughter’s affidavit.63
Similarly, the California Supreme Court has held that a judge is subject
to investigation for “legal error which, in addition, clearly and convinc-
ingly reflects bad faith, bias, abuse of authority, disregard for fundamental
rights, intentional disregard of the law, or any purpose other than the
faithful discharge of judicial duty. . . .”64 The same high court has found
that a judge engaged in willful misconduct by directing a jury to find a
defendant guilty when the judge at least should have known that the law
did not give him this authority.65

54. Stern, “Is Judicial Discipline in New York State a Threat to Judicial Independence?”
p. 304.
55. Shaman, “Judicial Ethics,” p. 9.
56. Ibid., p. 8.
57. See, e.g., Harrod v. Illinois Courts Commission at 65 (“[W]here the law is clear on its
face, a judge who repeatedly imposes punishment not provided for by law is subject to dis-
cipline”).
58. See cases cited in Shaman, “Judicial Ethics,” pp. 8–9 nn. 61, 62.
59. Stern, “Is Judicial Discipline in New York State a Threat to Judicial Independence?”
p. 313.
60. Ibid.
61. Ibid.
62. Inquiry Concerning a Judge No. 52 (Hampton) (Texas Commission on Judicial Con-
duct 1989).
63. In re McKinney, 478 S.E.2d 51 (S.C. 1996).
64. Oberholzer v. Commission on Judicial Performance at 680. One commentator inter-
prets this language to imply that Justice Kline, discussed in §3.12, would have been subject
to discipline had he written the majority opinion, rather than a dissent. Sankar, “Disciplining
the Professional Judge,” p. 1273 n. 234.
65. McCullough v. Commission on Judicial Performance, 776 P.2d 259, 262 (Cal. 1989).
The Legal Duties of Judges 65

Consider, by contrast, the case of Howard R. Broadman, a California

trial judge. Judge Broadman ordered a female defendant to receive an
implanted birth control device as a condition of her probation.66 In
another case, he made “not getting pregnant” a condition of a woman’s
five-year probation.67 The California Commission on Judicial Performance
brought “improper sentencing” charges against Judge Broadman, both of
which were ultimately dismissed after lengthy hearings.68 In a similarly
lenient decision, the Supreme Court of Alaska held that a judge should
not even be privately reprimanded for ordering the overnight imprison-
ment of a prosecuting witness in order to ensure that she would appear in
court sober the next day.69
Finally, recall the case of Justice Kline, whose dissenting opinion
announced his unwillingness to grant a stipulated reversal.70 The California
Commission on Judicial Performance charged Justice Kline with “willful
misconduct in office, conduct prejudicial to the administration of justice
that brings the judicial office into disrepute, improper action, and derelic-
tion of duty.” Lubet observes that Kline’s case “may well be the very first
time that a state judicial conduct organization has pursued disciplinary
charges solely on the basis of the substance of a judge’s written ruling.”71
The commission’s decision was widely criticized,72 and the charges against
Justice Kline were ultimately dismissed.73

4.4.1 Patterns of Error

So far I have mostly cited isolated instances of deviation, rather than pat-
terns of deviation by a single judge over time. Lubet states that
a pattern of repeated and uncorrected legal error is obviously more serious
than an isolated instance. Judges who fail to learn and apply the law fall into a
distinctly different category from those who simply hold minority—or innova-
tive—opinions. The commission of multiple errors, or unacceptable judging

66. Broadman v. Commission on Judicial Performance, 959 P.2d 715, 725 (Cal. 1998).
67. People v. Zaring, 10 Cal. Rptr. 2d 263, 265 (5th Dist. 1992).
68. Broadman v. Commission on Judicial Performance.
69. In re Curda, 49 P.3d 255 (Alaska 2002) (reversing recommendation of Commission
on Judicial Conduct).
70. See §3.12.
71. Lubet, “Judicial Discipline and Judicial Independence,” p. 66. See also Nancy
McCarthy, “Judge Faces Discipline: Commission Charge Unleashes Protest in Legal Com-
munity,” California State Bar Journal, August 1998.
72. McCarthy, “Judge Faces Discipline.”
73. Two facts about Kline’s case are especially interesting. First, Justice Kline was
charged with misconduct for language in a dissenting opinion that, as such, had no legal
force. Second, Kline advocated deviation in order to protect the judiciary, rather than a party
to the case. In fact, his position would block a transaction favored by both parties to the case,
whereas deviation ordinarily benefits the legally disfavored party at the expense of the
other.
66 Limits of Legality: Part I

that continues over a period of years, may indicate that the judge has not main-
tained professional competence in the law.74

Indeed, most judges who are actually disciplined for legal errors have
patterns of the same type of error:
Judges have been sanctioned for patterns of failing to advise defendants of their
rights . . . during criminal proceedings; imposing sentences in excess of statu-
tory authority; accepting guilty pleas using a form that did not comply with
statutory requirements; holding trials in absentia; violating procedural require-
ments when conducting arraignments; disregard of and indifference to fact or
law in criminal and juvenile cases; illegally incarcerating individuals in non-
criminal matters to satisfy a civil fine; accepting guilty pleas without obtaining
proper written plea statements; a practice of stating, for the record, that defend-
ants had waived their rights to have speedy preliminary examination or timely
trial without obtaining the defendants’ personal waivers of these rights; requir-
ing pro se defendants who requested jury trials to answer an in-court “jury trial
roll call” once a week and to discuss plea bargains with the prosecutor; and fail-
ing to advise litigants in family court cases of their statutory rights to counsel,
a hearing, and the assistance of counsel.75

4.4.2 Bad Faith

Legal errors can also bring sanctions when committed in “bad faith.” Bad
faith can include a “corrupt purpose” or an improper motive, such as
revenge or anger, as when a Massachusetts judge set an unusually high bail
for four African-American defendants and proceeded to tell a court clerk,
“That’s what blacks get for voting against my brother.”76 Since at least the
late nineteenth century, judges in New York State have been subject to
discipline for conduct that “reflects bias, malice, or intentional disregard of
the law.”77
Lubet suggests that “[a] willful refusal to follow the law, as distinct
from an honest and acknowledged difference of opinion or interpretation,
may manifest unfitness for judicial office.”78 Some courts have counted
intentional failure to follow the law as bad faith, even when the judge acts
from a benign motive. The Court of Appeals of New York (the high court
of the state) sanctioned a judge who failed to set bail, as required by law,
in twenty-four cases, although he did so because he believed that the

74. Lubet, “Judicial Discipline and Judicial Independence,” p. 72. See, e.g., Kloepfer v.
Commission on Judicial Conduct, 782 P.2d 239 (Cal. 1989).
75. Gray, “The Line between Legal Error and Judicial Misconduct,” pp. 1263–65. See
also Shaman, “Judicial Ethics,” p. 9.
76. In re King, 568 N.E.2d 588, 594 (Mass. 1991). The judge’s brother had fared badly
in minority voting districts in his recent run for governor. Lubet, “Judicial Discipline and
Judicial Independence,” p. 73.
77. Stern, “Is Judicial Discipline in New York State a Threat to Judicial Independence?”
p. 303.
78. Lubet, “Judicial Discipline and Judicial Independence,” p. 73.
The Legal Duties of Judges 67

defendants (many of them homeless) preferred to remain in jail. In a dif-

ferent case, the same court removed a judge for knowingly disregarding
the law by dismissing cases “in the interests of justice, using the guise of
factual insufficiency” when he “thought it was right to do it.”79

4.4.3 Obvious or Egregious Errors

Judges can also be sanctioned, even for an isolated legal error, when it is
especially serious or “egregious.”80 Several state high courts have held that
only “obviously” mistaken rulings are subject to discipline. In New York
and Indiana, judges are not to be disciplined for legal error if the correct-
ness of the decision is “sufficiently debatable.”81 The Supreme Court of
Maine has stated that discipline is proper only if “a reasonably prudent
and competent judge would consider [the judge’s decision] obviously and
seriously wrong in all the circumstances.”82 Lubet suggests that
legal error becomes serious enough to warrant discipline when judges deny
individuals their basic or fundamental procedural rights, as when a judge pro-
ceeds to adjudication without advising a defendant of the right to counsel,
declines to hold a full hearing, or coerces a guilty plea. The same may occur
when judges act beyond their lawful jurisdiction, as by sentencing defendants
to jail when only a fine is authorized by law or sentencing defendants to incar-
ceration for a period longer than the maximum allowed by statute.83

The following are some examples of legal errors in criminal cases that
have been considered sufficiently serious that one or two such errors war-
ranted discipline:
[F]inding a defendant guilty without a guilty plea or trial, revoking a defend-
ant’s probation without the defendant’s attorney being present, accepting a
defendant’s guilty plea without an attorney present and adjudicating a criminal
matter for which there was no formal case opened, sentencing a defendant
under the wrong statute, failing to follow proper procedures when a defendant
failed to pay a fine, refusing to allow a self-represented defendant to cross-
examine a police officer in a trial on a speeding ticket, knowingly convicting a
defendant of an offense that had not been charged and was not a lesser includ-
ed offense, refusing to set appeal bonds for misdemeanor defendants when
clearly obligated by law to do so, issuing bench warrants for the arrests of mis-
demeanor defendants when their attorneys had been late even though the

79. In re Duckman, 699 N.E.2d 872, 875 (N.Y. 1998). The judge had not given the
prosecution notice or allowed it an opportunity to be heard or to redraft charges. Nor had
he required written motions or consent of the prosecutor, as required by law.
80. Lubet, “Judicial Discipline and Judicial Independence,” p. 73; In re Quirk, 705 So.2d
172, 178 (La. 1997) (holding that even a single instance of serious legal error may constitute
judicial misconduct, especially if it involves denial of fundamental rights).
81. In re LaBelle, 591 N.E.2d 1156, 1161 (N.Y. 1992); In re Spencer, 798 N.E.2d 175,
183 (Ind. 2003).
82. In re Benoit, 487 A.2d 1158, 1163 (Me. 1985).
83. Lubet, “Judicial Discipline and Judicial Independence,” p. 73.
68 Limits of Legality: Part I

defendants themselves had been in court, forcing a defendant to enter a plea of

guilty in the absence of his counsel, using the criminal process to collect a civil
debt, . . . detaining a juvenile for nearly six weeks before he had the assistance
of counsel and without taking any evidence, and twice convicting a defendant
in the defendant’s absence and without a guilty plea.84

4.5 IMPEACHMENT, CONVICTION, AND REMOVAL

Could a judge be impeached and removed from office in the United States
solely for deviating from the law? This is a subtle question. There is no
consensus on the exact scope of the congressional impeachment power,85
so Congress could, perhaps, impeach a federal judge for deviating from the
law. But Congress impeached only thirteen federal judges between 1789
and 1991,86 just seven of whom were eventually removed from the bench.87
None was removed because legislators disagreed with his decisions.88
Most state judges in the United States are elected to finite terms of
office. They can, of course, lose reelection for any reason that motivates
the electorate, including deviation. But that possibility does not entail
that elected judges have any legal obligation to adhere, as opposed to a
political incentive to do so. Impeachment and removal are more relevant
to my question because these remedies require a legal basis. It is signifi-
cant, therefore, that since 1785 no state judge has ever been impeached,
convicted, and removed from office based on political or policy disagree-
ments with the judge’s decision.89 However, since 2001, there have been
several cases of state legislatures threatening impeachment of judges
based solely on a decision. In a widely publicized case from 2006, the
governor and legislators in Vermont called upon District Judge Edward
Cashman to resign for the sixty-day sentence he handed down in a serial
child molestation case.90 Justice James Heiple of the Illinois Supreme

84. Gray, “The Line between Legal Error and Judicial Misconduct,” pp. 1270–72.
85. “Although no consensus has ever developed on the exact scope of the impeachment
power, its reach is illuminated by two hundred years of experience. Congress has removed
judges for various forms of official and personal misconduct, but it has not done so because
it disagreed with the outcome of cases.” Report of the National Commission on Judicial
Discipline and Removal, 152 F.R.D. 265, 282 (Aug. 1993).
86. Warren S. Grimes, “Hundred-Ton-Gun Control: Preserving Impeachment as the
Exclusive Removal Mechanism for Federal Judges,” UCLA Law Review 38 (1991): 1209–55,
p. 1214 n. 32. See also Mary L. Volcansek, Judicial Impeachment: None Called for Justice
(Urbana: University of Illinois Press, 1993), p. 89; Eleanore Bushnell, Crimes, Follies and
Misfortunes: The Federal Impeachment Trials (Urbana: University of Illinois Press, 1992).
87. Grimes, “Hundred-Ton-Gun Control.”
88. Ibid. See also Charles Gardner Geyh, “Informal Methods of Judicial Discipline,”
University of Pennsylvania Law Review 142 (1993): 243–331.
89. John O. Haley, “The Civil, Criminal and Disciplinary Liability of Judges,” American
Journal of Comparative Law 54 (2006): 281–91.
90. Wilson Ring, “Vt. Judge Criticized for Molester Sentence,” Associated Press Online,
Jan. 10, 2006. Judge Cashman subsequently accepted the prosecution’s motion for reconsid-
eration and increased the sentence.
The Legal Duties of Judges 69

Court was widely lambasted in the press and by politicians for ordering
that “Baby Richard” be returned to his biological father, who had not con-
sented to the adoption, despite the ties the infant had formed to his adop-
tive family. The Illinois House of Representatives initiated an impeachment
inquiry against Judge Heiple, ostensibly for several traffic incidents and
administrative matters.91
Recent years have seen other, less notorious, cases as well. In 2006, the
legislature of New Hampshire considered removing Superior Court Jus-
tice Kenneth R. McHugh for a decision, made years earlier, in which he
ruled that a petitioner’s pleadings in a divorce case were frivolous. The
same year the Ohio House of Representatives considered removing Judge
John Connor for his sentencing of a sex offender. The Speaker of the
House issued a press release saying the House was “reviewing the pro-
cesses by which Judge Connor may be removed from the bench.” 92 In
Colorado, a bill of impeachment was introduced against Judge John W.
Coughlin in 2004 for his order in a child custody case.93
Notice, however, that these judges were not primarily accused of devi-
ating from the law. Judges who make unpopular decisions receive criti-
cism from the news media, the general public, and politicians, but the
critic’s usual objection is that he believes the decision to be substantively
unfair, unjust, or inexpedient. In some cases, of course, a critic explicitly
accuses the judge of deviating from (“ignoring,” “disregarding”) the law.
But the most despised decisions are often exercises of lawful discretion, as
in the aforementioned state cases.
Therefore, we must not conflate criticism of judges with accusations of
deviation. Critics who consider a result unjust or unfair often argue that
the judge has deviated from the law, and present this as a reason to take
adverse action against him, when they would have remained silent had
the judge deviated from the law in order to reach a result of which they
approve. The reasons presented by a critic for disciplining a judge may not
reflect the critic’s actual motivations. Nevertheless, critics do occasionally
present deviation, per se, as the target of their disapprobation, whatever
their actual motivations. In 2005, for example, a bill was introduced in the
Tennessee Senate that would make any decision that “deviates from a rule
of law” or precedent a presumptive act of judicial misconduct.94 Similar
bills have been introduced recently in other states. In 2006, the Judiciary

91. Lubet, “Judicial Discipline and Judicial Independence,” p. 69.

92. See William E. Raftery, “The Legislatures, the Ballot Boxes, and the Courts,” Court
Review 43 (2006): 102–7.
93. Ibid.
94. The bill required the judge to “present clear and convincing evidence that, before
ruling, the adjudicator competently and thoroughly researched the law on the question
controlling [and] cite uncontradicted and controlling precedent . . . confirming that the
question was one of first impression.” The bill died in committee. S.B. 3522, 104th General
Assembly, Second Session (Tenn. 2006).
70 Limits of Legality: Part I

Committee of the New Jersey State Assembly considered resolutions

seeking the impeachment of the entire New Jersey Supreme Court for
their allegedly deviant ruling on same-sex marriage.95
I conclude that judges could, in theory, be impeached and removed
from office for deviating from the law. But this possibility has not yet
materialized in the United States. As Judge Richard Posner notes, “detec-
tion and sanctioning [of judges who] bend or break the rules for the sake
of other values [are] difficult.” 96 For my purposes it is also significant that
even the most severe sanction—impeachment and removal—is qualita-
tively different from typical criminal and civil penalties, which involve
deprivations of property or personal liberty. The act of knowingly deviat-
ing from the law, in itself, can bring nothing worse than public censure,
impeachment, and removal from the bench. The disrobed judge retains
his property and personal liberties. He loses his job, but he remains
employable. He may not even lose his license to practice law.
Therefore, one might wonder whether judges really have a legal duty to
apply the law. If they have such a duty, then the legal system does not
appear to take it very seriously. At least the legal system does not provide
the sanctions or remedies that it provides for violations of other important
legal duties.97 I have, of course, noted that legal duties do not entail sanc-
tions.98 Nevertheless, I cannot conclude with complete confidence that
judges have a legal obligation to apply the law, much less a stringent one,
because the legal sanctions for deviating from the law are relatively weak
and so rarely applied. Nor can I rule out the possibility of a legal obliga-
tion to adhere. Therefore, in the remainder of this book I shall simply
assume arguendo that judges have such an obligation. If this assumption
proves mistaken, then so much the better for my central argument.99

4.6 OTHER JUDICIAL INCENTIVES TO ADHERE

Despite the weakness and rare application of formal sanctions for deviat-
ing, judges have various incentives to adhere. Most judicial decisions in
Anglo-American systems are subject to reversal on appeal. Most judges

95. A.R. 217 (Justice Albin), 212th Legislature, Second Session (N.J. 2007); A.R. 218
(Justice LaVecchia), 212th Legislature, Second Session (N.J. 2007); A.R. 219 (Justice Rivera-
Soto), 212th Legislature, Second Session (N.J. 2007); A.R. 220 (Justice Wallace), 212th
Legislature, Second Session (N.J. 2007); A.R. 221 (Chief Justice Poritz), 212th Legislature,
Second Session (N.J. 2007); A.R. 222 (Justice Zazzali), 212th Legislature, Second Session
(N.J. 2007); A.R. 223 (Justice Long), 212th Legislature, Second Session (N.J. 2007).
96. Richard A. Posner, “What Do Judges and Justices Maximize? (The Same Thing
Everybody Else Does),” Supreme Court Economic Review 3 (1993): 1–41, p. 30.
97. Of course, this may be because the system has not, thus far, found it necessary to
impose severe sanctions for deviation.
98. See §4.3.
99. When I write that I am “accepting a proposition arguendo” I mean that I am assum-
ing it to be true even though my overall argument in the book is easier to make if it is false.
The Legal Duties of Judges 71

dislike having their decisions reversed.100 Insofar as deviation increases

one’s likelihood of reversal, the desire to avoid reversal, for those who
have it, constitutes an incentive to adhere. Higher courts are “less revers-
ible” than lower courts. At the limit are decisions on questions of state law
by state supreme courts and decisions by the U.S. Supreme Court. But
statutory decisions of high courts are “reversible” by legislatures, and even
the constitutional decisions of high courts in the United States are revers-
ible by constitutional amendment (state or federal). In principle, every
American judge can be “reversed.”
Most judges also want to maintain good professional reputations.101 In
a community in which fidelity to law is valued, the perception that a
judge has deviated may harm her professional reputation. Deviating
increases the likelihood that others will conclude that one has deviated.
Therefore, under these conditions, a judge who cares about his reputation
has an incentive to adhere. There have been a few studies of the methods
of “informal discipline” that judges use to police one another.102 Neil Mac-
Cormick observes:
It would be strange if a judge’s opinion as to the normative quality of the alter-
native decisions confronting him in litigation were not for him a motivating
factor in making up his mind what decision to give, that is, what order to make.
He does after all have to state publicly in open court the reasons for which he
is deciding the case as he is. Given the institutional pressures within the legal
system—the opinion of the profession, the possibility of an appeal, etc.—and
given the external pressures of adverse press publicity and Parliamentary com-
ment and the like, it would be so strange as to be barely imaginable that a judge
having established the justifiability of one decision by logical argument from
sound legal premises and findings of fact should then issue some diametrically
different order. So institutionally and psychologically it is highly unlikely that a
judge will so conduct himself, but it is not impossible.103

Many judges also want to receive interesting and/or prestigious oppor-

tunities, both remunerative and nonremunerative, in the future. Many

100. Posner, “What Do Judges and Justices Maximize? (The Same Thing Everybody
Else Does),” p. 14 (“Judges don’t like to be reversed. . . . I speak from experience. . . .”). Posner
asserts, nevertheless, that “aversion to reversal does not figure largely in the judicial utility
function.” See also Donald R. Songer, Martha Humphries Ginn, and Tammy A. Sarver, “Do
Judges Follow the Law When There Is No Fear of Reversal?” Justice System Journal 24 (2003):
137–61; Richard S. Higgins and Paul H. Rubin, “Judicial Discretion,” Journal of Legal Studies
9 (1980): 129–38, p. 130 (“For reasons not completely understood, judges seem to desire to
avoid being reversed”).
101. See, e.g., Posner, “What Do Judges and Justices Maximize? (The Same Thing
Everybody Else Does),” p. 13; Robert D. Cooter, “The Objectives of Private and Public Judges,”
Public Choice 41 (1983): 107–32, p. 129.
102. Geyh, “Informal Methods of Judicial Discipline”; Sankar, “Disciplining the Profes-
sional Judge,” pp. 1254–56.
103. MacCormick, Legal Reasoning and Legal Theory, pp. 33–34. See also Scott J.
Shapiro, “Judicial Can’t,” Noûs 35, Supp. 1 (2001): 530–57, p. 546.
72 Limits of Legality: Part I

wish to be promoted to higher office, judicial or otherwise. Insofar as a

record of deviation reduces one’s attractiveness for interesting, lucrative,
and prestigious opportunities, judges have an incentive to adhere.104 How-
ever, one early study found that reversal rates do not affect the chance of
promotion for federal district judges.105
Some scholars have also suggested that the less faithful to the law other
judges (and scholars and lawyers) perceive a certain judge to be, the less
often will they cite or follow him.106 If this empirical hypothesis is true,
then judges have another prudential reason to adhere.
Judge Posner places greatest emphasis upon yet another incentive for
judges to adhere: it gives them pleasure. Posner analogizes judges to citi-
zens voting for public officials, audience members watching stage plays,
and players of games. He observes that “[a] chess player would reduce
rather than enhance the pleasure he received from playing a game if he
violated the rules, and so would a theatergoer who refused to enter into
the lives of the characters on the stage, on the ground that they were not
real people; and likewise the judge who violates the rules of the judicial
game.” 107
And also:
[M]any people do not cheat at games even when they are sure they can get
away with cheating. The pleasure of judging is bound up with compliance with
certain self-limiting rules that define the “game” of judging. It is a source of
satisfaction for a judge to vote for the litigant who irritates him, the lawyer who
fails to exhibit proper deference to him, the side that represents a different
social class from his own; for it is by doing such things that you know that you
are playing the judge role, not some other role, and judges for the most part are
people who want to be—judges.108

Finally, in many cases simply adhering to the law is the easiest and
least time-consuming option. Deciding whether one has good reasons to

104. “[I]f judges who were most faithful to the rules were the ones most likely to be
elevated to higher positions, then the hope of this reward might for many judges provide the
incentive for taking rules to be reasons for action.” Schauer, Playing by the Rules, pp. 123–24.
For research findings see, e.g., Thomas J. Miceli and Metin M. Cogel, “Reputation and Judi-
cial Decision-Making,” Journal of Economic Behavior and Organization 23 (1994): 31–51;
Erin O’Hara, “Social Constraint or Implicit Collusion? Toward a Game Theoretic Analysis of
Stare Decisis,” Seton Hall Law Review 24 (1993): 736–78; Mark A. Cohen, “The Motives of
Judges: Empirical Evidence from Antitrust Sentencing,” International Review of Law and
Economics 12 (1992): 13–30.
105. Higgins and Rubin, “Judicial Discretion.”
106. See, e.g., Lewis A. Kornhauser, “Modeling Collegial Courts I: Path Dependence,”
International Review of Law and Economics 12 (1992): 169–85; Lewis A. Kornhauser, “Mod-
eling Collegial Courts II: Legal Doctrine,” Journal of Law, Economics, and Organization 8
(1992): 441–70; Richard A. Posner, Economic Analysis of Law, 4th ed. (Boston: Little, Brown,
1992), pp. 534–36, 541–42.
107. Posner, “What Do Judges and Justices Maximize? (The Same Thing Everybody
Else Does),” p. 26.
108. Ibid., p. 28.
The Legal Duties of Judges 73

deviate often requires one to consider a wider range of factors than would
blind adherence to the law. Insofar as judges value their leisure109 and are
“cognitively lazy,” 110 they have another incentive to adhere in such cases.
Of course, these incentives should not be confused with a legal or moral
duty to adhere.

109. Ibid., p. 11.

110. Karl E. Weick, Sensemaking in Organizations (Thousand Oaks, Calif.: Sage, 1995),
pp. 61–62.
5

The Normative Classification

of Legal Results

5.1 OPTIMAL VERSUS SUBOPTIMAL RESULTS

In this chapter, I shall discuss the evaluation of court decisions in normative

terms that go beyond the law. In addition to forming an opinion about the
legal correctness of a court’s ruling, one can form an opinion that charac-
terizes it as more or less just, more or less fair, more or less conducive to
the general welfare, et cetera. Judicial decisions always have effects on
someone. Most obviously, a criminal conviction has negative effects on the
convict. It also has negative effects on his family members, friends, employer,
employees, and so forth. An acquittal may have negative effects on the
general public, on future victims, and on victims of the crime itself.
In civil cases, a decision has negative effects on the losing litigant. A
losing defendant is typically made worse off than he was before the judg-
ment (although he may be left no worse off than he was before he com-
mitted the delict). A losing plaintiff is typically not made worse off than
he was before the judgment, although he may be left worse off than he
was before the delict against him.
In theory, if all values were taken into account, each result in a court
case could be classified into one of two moral categories: optimal or subop-
timal. These could be further broken down into four subcategories. Some
results are not just morally optimal, but morally mandatory, meaning that
it is morally wrong, all things considered, for a court to reach any other
result in that case. Other results are optimal but not mandatory, meaning
that a court is morally permitted but not morally required to reach that
result.
The worst subcategory of suboptimal results comprises those that are
impermissible, meaning that it is morally wrong, all things considered, for
a court to reach such a result. All other suboptimal results are morally
permissible for a court to reach, but still suboptimal because a morally
superior result exists.
At this point some readers will be impatient for concrete illustrations,
which I shall provide shortly. However, in this book I stay as neutral as
possible regarding which results are optimal. For me the distinctions
between optimal, indifferent, and suboptimal results are conceptual

74
The Normative Classification of Legal Results 75

distinctions, like the philosopher’s distinction between justice and injus-

tice. To evaluate my arguments you need only distinguish for yourself
between optimal and suboptimal results. You and I need not agree
regarding the proper classification of any particular result. Our opinions
will be influenced by our beliefs about empirical matters and by our
probably divergent theories of corrective justice, distributive justice,
criminalization, and punishment.
In the rest of this book some readers will find themselves repeatedly
plagued by questions such as the following: how can a judge know that
she is correct when she determines that a certain result is morally subopti-
mal? What if she is wrong? Are there even “correct” answers to difficult
moral questions? Some readers may be unable to get past these questions.
I may appear to be making some naïve assumptions: that determining
whether a result is suboptimal is always an easy, mechanical process; that
judges possess perfect moral insight; that judges generally agree on nor-
mative issues. These are unrealistic assumptions. My arguments are
designed to succeed without them, although you will have to decide if
they do. I am simply trying to respect what John Rawls calls “the fact of
reasonable pluralism.” 1 However, I cannot permanently ignore the fact
that judges will disagree with one another on many moral questions. I
shall address the fact of moral disagreement in chapter 14.2 I beg your
patience until then.
Again, my objective is not to persuade you to adopt my general norma-
tive views or my opinions on particular policy issues. For my purposes the
following definitions suffice:
• In any court case to be decided, a possible result, r1, is normatively
superior to another possible result, r2, if and only if it is true that
the judge would have a stronger reason to reach r1 than r2 if the law
permitted him to reach either.
• A result is normatively optimal if and only if no other result is
superior to it.
• A result is normatively suboptimal if and only if it is not optimal.3
Notice that my definition of superior does not specify whether the law
in the actual world requires the judge to reach either r1 or r2. It defines
superior in terms of the reasons the judge would have in a specific possible

1. See John Rawls, “The Idea of Public Reason Revisited,” University of Chicago Law
Review 64 (1997): 765–807, pp. 765–66 (“[A] basic feature of democracy is the fact of rea-
sonable pluralism—the fact that a plurality of conflicting reasonable comprehensive doc-
trines, religious, philosophical, and moral, is the normal result of its culture of free
institutions”). See generally John Rawls, Political Liberalism (New York: Columbia University
Press, 1993).
2. See §14.8.
3. Frederick Schauer, Playing by the Rules: A Philosophical Examination of Rule-Based
Decision-Making in Law and in Life (Oxford: Oxford University Press, 1991), p. 204.
76 Limits of Legality: Part I

world: one that is exactly similar to the actual world except that the law
permits the judge to reach either result (which the law in the actual world
might or might not permit). Consider the straightforward example of a
judge presiding over the bench trial of a defendant charged with criminal
battery for maiming someone without justification. Which result is opti-
mal: conviction or acquittal? On my definition of optimal the answer
depends on what the judge would have an all-things-considered reason to
do if the law gave him unguided discretion to convict or acquit on this
evidence. In this trivial case, reasonable people will agree that the judge
would have an all-things-considered reason to convict if the law permit-
ted him to do so: conviction is an optimal result. Other cases, of course,
will invite moral controversy.
Maiming is an offense malum in se.4 Having used this example I must
forestall a potential misunderstanding that could arise in connection with
any malum prohibitum offense: one that is “not inherently immoral, but
becomes so because its commission is expressly forbidden by positive
law.”5 Examples include failing to file tax returns, counterfeiting, running
stop signs when it is safe to do so, et cetera. Because these actions are not
inherently immoral, one might jump to the conclusion that a judge would
have an all-things-considered reason to impose no penalty whatsoever for
them if the law permitted her to do so, which would entail that a convic-
tion for a malum prohibitum offense is always a suboptimal result on my
definition. I shall explain why this is not so.
Imagine a world, otherwise like ours, with one difference in the law:
although counterfeiting is a crime, the law gives a certain judge unguided
discretion in bench trials either to convict or to acquit defendants accused
of counterfeiting.6 This judge has an all-things-considered reason to con-
vict counterfeiters, despite the fact that she is not legally required to do
so. Whatever reasons one accepts for convicting counterfeiters in the real
world apply with the same force in this hypothetical. Circulating counter-
feit bills still debases the national currency, so the hypothetical society
should still discourage counterfeiting. One could also argue that counter-
feiting currency is immoral in the hypothetical society. Knowingly circu-
lating counterfeit currency is fraudulent, so it is still morally wrong.
Insofar as counterfeiting currency facilitates its circulation, one could
argue that counterfeiting is wrong for that reason. Counterfeiting also
violates the fair play principle if most people obey the law.7 Therefore, the
hypothetical judge has most of the same reasons as a real-world judge to

4. “[I]nherently and essentially evil, that is immoral in its nature and injurious in its
consequences, without any regard to the fact of its being noticed or punished by the law of
the state.” Bryan A. Garner, ed., Black’s Law Dictionary, 8th ed. (St. Paul, Minn.: Thomson/
West, 2004).
5. Ibid.
6. Obviously, this is bad policy.
7. Richard Dagger, “Playing Fair with Punishment,” Ethics 103 (1993): 473–88.
The Normative Classification of Legal Results 77

convict a counterfeiter. Conviction is the optimal result. Similar analyses

apply to other mala prohibita offenses.

5.2 THE NORMATIVE CLASSIFICATION OF REGULATED CASES

IN TERMS OF LEGALLY DICTATED RESULTS

Legally regulated cases divide into four categories corresponding to the

four normative categories of results: required; optimal-but-not-required;
suboptimal-but-permissible; and impermissible. I shall refer to required-
result and optimal-but-not-required-result cases, collectively, as optimal-
result cases. Suboptimal-but-permissible-result and impermissible-result
cases, collectively, are suboptimal-result cases. An optimal-result case is one
in which the law authorizes the judge to reach a result that would other-
wise be morally optimal, all things considered, absent the law. All other
cases are suboptimal-result cases: those in which the law requires the
judge to reach a result that would otherwise be morally suboptimal, all
things considered, absent the law. In such a case the law requires her to
rule in favor of a certain party although she would otherwise have an all-
things-considered moral reason to rule against that party, ceteris paribus.
In other words, a case is a suboptimal-result case if and only if the law
either (1) requires the judge to reach a result that she would have all-
things-considered reason to avoid, if the law allowed her to avoid it, or (2)
forbids her to reach a result that she would have all-things-considered
reason to reach, if the law allowed her to reach it. By the definition of
suboptimal-result case, it is possible for a judge who makes no legal or fac-
tual errors to reach a suboptimal result in such a case.
It is important to remember that the adjective suboptimal in the phrase
suboptimal-result case modifies the result required by the law, not the result
actually reached by a court. A suboptimal-result case is not one in which
some court reached a suboptimal result. It is one in which the law, prop-
erly understood, requires (or required) the court to reach a suboptimal
result, whatever the court actually does (or did). The same holds for the
other three categories, mutatis mutandis.
Here are some examples of cases in the four categories mentioned
above:
1. Criminal defendant has a moral right to be acquitted. He is morally
or factually innocent. The law also requires the judge to acquit.
We have natural duties not to punish morally innocent people, so
the judge has a moral duty to acquit. The law requires the judge
to commit or omit actions that he has a natural duty to commit or
omit, respectively. This is a required-result case.
2. Neither party to a certain civil matter has a moral right to a
favorable judgment, but a judgment for plaintiff would benefit
plaintiff very greatly and harm defendant, at most, only minimally
78 Limits of Legality: Part I

(and no other individuals will be affected more than negligibly by

either result). Therefore, the judge has a natural moral reason to
rule for plaintiff, but no natural moral duty to do so. If the law
requires the judge to rule for plaintiff, then the judge has a
legal duty to rule for plaintiff. The judge has a judicial duty
to commit or omit actions that he has natural reasons, but
no natural duty, to commit or omit, respectively. This is an
optimal-but-not-required-result case.
3. Neither party has a moral right to a favorable judgment, but a
judgment for plaintiff would benefit plaintiff very greatly and
harm defendant, at most, only minimally (and no other individuals
will be much disadvantaged by either result). The law requires the
judge to rule for defendant. The judge has no moral duty to rule
for plaintiff, but he has a natural moral reason to do so. The judge
has a judicial duty to commit or omit actions that he has natural
reasons, but no natural duty, to omit or commit, respectively. This
is a suboptimal-but-permissible-result case.
4. Criminal defendant has a moral right to be acquitted. He is morally
or factually innocent, but the law requires the judge to convict
and sentence him. The judge has a judicial duty to commit or omit
actions that he has a natural duty to omit or commit, respectively.
This is an impermissible-result case.

5.3 SUBOPTIMAL-RESULT CASES

I shall further illustrate the concept of a suboptimal-result case using

a hypothetical scenario—a standard case of economic hardship.8 Yasmin
is an impoverished widow who is renting an apartment from Rafael, a
wealthy landowner who inherited the property and has never worked a
day in his life. Yasmin has missed several rent payments, and Rafael has
filed an eviction petition before Judge Lucas. I stipulate that Yasmin is
morally innocent. She is too old and frail to work, and her pension does
not cover her rent. Neither retributive justice nor corrective justice
requires a judgment against Yasmin. She has done nothing to deserve
eviction as a moral matter (although legally, she is delinquent). Nor does
Judge Lucas violate anyone’s moral rights if he rules for Yasmin. That
decision violates Rafael’s legal rights, but Rafael was not morally entitled
to Yasmin’s rent in the first place, absent the law. The property rights
regime in place, let us suppose, is distributively unjust, because it pro-
vides an inadequate social safety net for people in Yasmin’s situation. Nor

8. This case is adapted from one discussed in Richard A. Wasserstrom, The Judicial
Decision: Toward a Theory of Legal Justification (Stanford, Calif.: Stanford University Press,
1961), p. 141; and in Alan H. Goldman, Practical Rules: When We Need Them and When We
Don’t (Cambridge: Cambridge University Press, 2002), p. 43.
The Normative Classification of Legal Results 79

will Yasmin’s eviction maximize the combined utility of Yasmin and

Rafael, because Yasmin needs her apartment more than Rafael needs her
money. If the law did not require eviction, then Judge Lucas would have
an all-things-considered moral reason to deny Rafael’s petition. However,
the regulating statute requires him to grant the petition, thereby evicting
Yasmin.
This is my example of a suboptimal-result case, but I offer it merely
for the sake of being concrete. My point is not to persuade you that evict-
ing Yasmin is a suboptimal result. That conclusion depends on certain
theories of distributive justice that I shall not ask you to accept. If you
are glad to see Yasmin evicted, then you can imagine some other case,
one in which the law requires a result that you consider to be suboptimal.
Below is a list of results, inspired by actual cases and chosen to stimulate
your imagination. The fact of reasonable pluralism entails that for most
of these results, you should find it easy to imagine an intelligent, informed
person who considers the result to be suboptimal, whether or not you
agree:
1. A court orders the deportation of an undocumented immigrant
who has worked hard and contributed a great deal to his commu-
nity.
2. A court forbids a student from attending a public school outside
his own district, despite the fact that his own school system offers
an inadequate and inferior learning environment.9
3. Under mandatory minimum sentencing laws, a court imposes a
life sentence on someone who was convicted of one assault with a
deadly weapon and two subsequent, nonviolent felonies.10
4. A court upholds a jury’s decision to execute a defendant who
received a fair trial.
5. A court orders the forfeiture of assets from a convict’s innocent
family members.11
6. A court excludes illegally obtained evidence that is necessary for
conviction of a defendant accused of serial child molestation.
7. A court sentences to prison an animal rights activist who has been
convicted of stealing animals from a research laboratory in order to
liberate them.12
8. A court sentences a seventeen-year-old male to a prison term of
ten years for receiving consensual fellatio from his fifteen-year-old
girlfriend.13

9. San Antonio v. Rodriguez, 411 U.S. 1 (1973) (upholding local funding scheme for
public schools despite substantial inter-district disparities in school expenditures).
10. Lockyer v. Andrade, 538 U.S. 63 (2003).
11. Bennis v. Michigan, 116 S. Ct. 994 (1996).
12. State v. Troen, 786 P.2d 751 (Or. Ct. App. 1990).
13. Wilson v. State, 279 Ga. App. 459 (Ga. Ct. App. 2006). The girl in this case was not,
in fact, the defendant’s girlfriend, but this makes no legal difference.
80 Limits of Legality: Part I

9. A court dismisses the First Amendment free speech and free

exercise claims of a student who wishes to lead his fellow
students in a Christian prayer at an event sponsored by a
public school.14
10. A court dismisses criminal charges brought against a doctor for
performing a previability abortion.15
11. A court upholds the criminal conviction of a doctor who
performed an abortion via intact dilation and evacuation
(a “partial-birth abortion”) when the doctor believed this
procedure to be necessary in order to protect the health of the
pregnant woman.16
12. A court upholds the federal possession conviction of a defendant
who grows marijuana for medical purposes.17
13. A court dismisses a case against a defendant who has trafficked
in “virtual” child pornography—computer-generated material
produced without using actual children as subjects.18
14. A court imposes a long sentence on a battered woman who has
killed her partner in his sleep.
15. A federal court orders a public official to remove a monument of
the Ten Commandments from a public building.19
16. A federal court overturns a ban on handgun ownership in the Dis-
trict of Columbia.20
Many commentators, regardless of their political views, believe that
the United States Supreme Court under Chief Justice Earl Warren
repeatedly deviated from the law—that some of the Court’s opinions
cannot be reconciled with any legally correct reading of the Constitution.
Political conservatives who take this position condemn the decisions as
unjustified instances of deviation.21 Many liberals deny that the Warren
Court deviated from the law at all, although they often admit that the
justices deviated from certain rules of law. These liberals deny that the law
actually required suboptimal results in the controversial cases. Fashioning
a theory of law that is consistent with that position is one of the chal-
lenges faced by these liberal lawyers.22 Conservative lawyers are, needless

14. Santa Fe Independent School District v. Doe, 530 U.S. 290 (2000).
15. Planned Parenthood v. Casey, 505 U.S. 833 (1992).
16. Gonzales v. Carhart, 550 U.S. 124 (2007) (upholding Partial-Birth Abortion Ban
Act of 2003).
17. Gonzales v. Raich, 545 U.S. 1 (2005).
18. Ashcroft v. Free Speech Coalition, 535 U.S. 234 (2002).
19. Glassroth v. Moore, 335 F.3d 1282 (11th Cir. 2003).
20. District of Columbia v. Heller, 128 S. Ct. 2783 (2008).
21. See, e.g., Robert H. Bork, The Tempting of America (New York: Simon & Schuster,
1990), ch. 3; Raoul Berger, Government by Judiciary: The Transformation of the Fourteenth
Amendment (Cambridge, Mass.: Harvard University Press, 1977), p. 265.
22. See, e.g., David A. Strauss, “The Common Law Genius of the Warren Court,”
William & Mary Law Review 49 (2007): 845–79.
The Normative Classification of Legal Results 81

to say, unpersuaded by their efforts. Nor is there much consensus among

liberal lawyers regarding how to justify the Warren-era decisions in legal
terms.23
My point, however, is that many liberal lawyers reluctantly agree that
the Warren Court sometimes deviated from the law of its day. They
see the cases in question as suboptimal-result cases in which the Court
deviated—perhaps fortunately, perhaps even heroically.24 As Brian Tama-
naha explains, even Brown v. Board of Education25 invites this reaction:
The Brown decision arguably was not supported by legal principles or reasoned
elaboration or neutral principle. Hence the decision could not be reconciled
comfortably with pre-existing Constitutional law and understandings. Yet it
was undoubtedly correct from a moral standpoint, at least in the view of many
in the mainstream of legal academia. Other progressive Warren Court deci-
sions raised the same dilemma: morally correct in content, as far as mainstream
legal academics were concerned, but nigh impossible to justify in purely legal
terms. These decisions, evidently a product of the political views of the justices,
were the antithesis of the rule of law, according to infuriated critics, one more
indication of its breakdown.26

The liberal law professors to whom Tamanaha refers were, and are,
divided among themselves. Some believe that the Court’s putative devia-
tion in Brown was ethically justified, albeit unlawful.27 Other liberals con-
clude, often with regret, that the justices acted unethically, despite reaching
just results. Just results could have been reached by the political branches
and, on this view, should not have been reached by the judiciary.28

23. Bruce Ackerman, We the People: Foundations (Cambridge, Mass.: Belknap, 1991), ch.
6; J. M. Balkin and Bruce A. Ackerman, eds., What Brown v. Board of Education Should Have
Said (New York: New York University Press, 2001); J. M. Balkin, ed., What Roe v. Wade
Should Have Said (New York: New York University Press, 2005). Warren Burger was Chief
Justice when Roe v. Wade was decided.
24. Learned Hand and Herbert Wechsler were antisegregation liberals who famously
declared Brown v. Board of Education deviant. Learned Hand, The Bill of Rights (Cambridge,
Mass.: Harvard University Press, 1958), p. 55; Herbert Wechsler, “Toward Neutral Principles
of Constitutional Law,” Harvard Law Review 73 (1959): 1–35. The liberal John Hart Ely
reached the same judgment about Roe v. Wade in John Hart Ely, “The Wages of Crying Wolf:
A Comment on Roe v. Wade,” Yale Law Journal 82 (1973): 920–49.
25. Brown v. Board of Education, 347 U.S. 483 (1954) (racial segregation in public
schools violates Equal Protection Clause).
26. Brian Z. Tamanaha, On the Rule of Law: History, Politics, Theory (Cambridge:
Cambridge University Press, 2004), p. 80.
27. “Brown . . . had serious, thoughtful critics among people who were opposed to seg-
regation. Some of those critics . . . said that Brown was a lawless act. . . .” David A. Strauss,
“Little Rock and the Legacy of Brown,” Saint Louis University Law Journal 52 (2008): 1065–
86, p. 1066. See also William H. Simon, “Should Lawyers Obey the Law?” William & Mary
Law Review 38 (1996): 217–53, p. 222 (“The balance of burden and benefit in the legal
order of the day was not fairly struck for African-Americans; the . . . arguably lawless [Brown]
decision inarguably pushed the balance toward greater fairness”).
28. See, e.g., Wechsler, “Toward Neutral Principles of Constitutional Law”; Hand, The
Bill of Rights, p. 55.
82 Limits of Legality: Part I

Judges themselves lament conflicts between their moral convictions

and the law as they understand it. Justice Blackmun firmly opposed the
death penalty on moral grounds, but dissented in Furman v. Georgia,
which forbade states to give sentencing juries “unguided discretion” in
capital cases. This effectively ended capital punishment in the United
States until 1976. Blackmun wrote, “Although personally I may rejoice at
the Court’s result, I find it difficult to accept or to justify as a matter of
history, of law, or of constitutional pronouncement. I fear the Court has
overstepped. It has sought and has achieved an end.” 29
Conservatives also face hard choices between swallowing politically
unpalatable outcomes and supporting judicial deviation from the law.
Scott Shapiro claims that

[e]ven the most vociferous conservative critics of the regulatory state do

not advocate that judges should ignore duly enacted law. Their claim is that
the regulatory state must be dismantled through the proper legislative
channels, by repealing certain law[s], passing others and shifting jurisdiction
from larger entities, such as the federal government, to smaller more
accountable ones, such as states and municipalities. In the meantime, how-
ever, “the law’s the law.” 30

Of course, the fact that conservatives often condemn deviation does

not entail that conservative judges never deviate.

5.4 CLASSIFYING SUBOPTIMAL-RESULT CASES

In this section, I use the previously introduced classifications to divide the

universe of cases into four categories: optimal rule/optimal result; subop-
timal rule/suboptimal result; suboptimal rule/optimal result; and optimal
rule/suboptimal result. I call these categories positive closure, negative
closure, positive gap,31 and negative gap, respectively.
A negative-closure case is one in which a suboptimal rule generates a
suboptimal result. The laws of many states in the antebellum United
States permitted the ownership of slaves. Because slavery is unjust, virtu-
ally every case in which the law supported the property rights of slave
owners in their slaves was a suboptimal-result case.
A positive-gap case is one in which a suboptimal rule generates an
optimal result. Although a law permitting slavery usually yields bad
results, on rare occasions it will, by coincidence, permit the harsh treat-
ment of an individual who deserves it, morally, such as a rapist who has

29. Furman v. Georgia, 408 U.S. 238, 414 (1972) (Blackmun, J., dissenting).
30. Scott J. Shapiro, “Judicial Can’t,” Noûs 35, Supp. 1 (2001): 530–57, p. 532.
31. This use of gap originates in Larry Alexander, “The Gap,” Harvard Journal of Law
and Public Policy 14 (1991): 695–701.
The Normative Classification of Legal Results 83

avoided arrest. These unusual cases are positive-gap cases.32 I will say little
more about them.
A positive-closure case is one in which an optimal rule generates an
optimal result. An example would be a case in which someone drives her
automobile on the left side of a two-way street when it is very dangerous
to do so and she is convicted under a law that forbids driving on the left.
Finally, here is an example of a negative-gap case. On a particular occa-
sion, Michael drives his automobile on the left, thereby knowingly breaking
the law. No one else is present, and it is safe to do so. Michael would never
drive unsafely. But for the law, he would not deserve a sanction, nor would
a sanction promote the good in this case, because it is not necessary to deter
Michael or others from driving unsafely. An official who was legally autho-
rized to sanction Michael would have no all-things-considered moral reason
to do so if no legal rule required her to do so. Therefore, sanctioning Michael
is a suboptimal result. However, the rule that Michael has violated may be
an optimal rule. It might be counterproductive to try to add a textual
exception for cases such as Michael’s. If so, then his is a negative-gap case.33

5.5 ASSUMPTIONS UNDERLYING THE EXISTENCE OF

SUBOPTIMAL-RESULT CASES

My discussion is addressed to people who believe that suboptimal-result

cases actually exist. Anyone who believes this is committed to two
assumptions. First, at least some legal questions have legally correct
answers. Second, the law sometimes requires judges to reach results that
they would have all-things-considered reasons to avoid, if the law permit-
ted. I discuss these assumptions in this section.

5.5.1 Partial Determinacy

Suboptimal-result cases cannot exist unless at least some legal questions
have legally correct answers.34 The antithesis—that no legal questions

32. One might also refer to them as “Gettier cases,” by analogy with Edmund L. Gettier,
“Is Justified True Belief Knowledge?” Analysis 23 (1963): 121–23.
33. It is “a platitude that justifiable rules can sometimes have morally regrettable applica-
tions.” David Lyons, “Derivability, Defensibility, and the Justification of Judicial Decisions,” in
Moral Aspects of Legal Theory (Cambridge: Cambridge University Press, 1993), p. 128. See also
Larry Alexander and Emily Sherwin, The Rule of Rules (Durham, N.C.: Duke University Press,
2001), pp. 53–61; Heidi M. Hurd, Moral Combat (Cambridge: Cambridge University Press,
1999), pp. 185–202; Alexander, “The Gap”; Schauer, Playing by the Rules, pp. 128–34; Gerald
J. Postema, Bentham and the Common Law Tradition (Oxford: Oxford University Press, 1986),
p. 407 (“[E]ven the best and most just of general rules may yield injustice in some particular
cases”); Rolf Sartorius, Individual Conduct and Social Norms (Encino, Calif.: Dickenson, 1975),
pp. 56–57. The idea dates back at least to Aristotle, Nicomachean Ethics V.x.5.xi.
34. See M. B. E. Smith, “Do Appellate Courts Regularly Cheat?” Criminal Justice Ethics
16 (1997): 11–20, p. 11.
84 Limits of Legality: Part I

have legally correct answers—is radical indeterminacy.35 Valid, mandatory

legal rules purport to give their addressees at least pro tanto reasons for
action. However, if a rule is radically indeterminate, then it does not pro-
vide such reasons. This entails that judges have no reason to adhere to the
rule because the rule does not really require anything.36 If legal rules are
all radically indeterminate, then the principle that ought implies can
entails that there is no judicial duty to adhere to the law and my remaining
arguments are otiose.
A commitment to radical indeterminacy is often attributed to the crit-
ical legal studies movement,37 although I hesitate to impute the thesis to
any particular author without defining it more precisely.38 The radical
indeterminacy thesis has been subjected to severe and extensive critique,
along with several other interrelated claims commonly ascribed to critical
legal studies.39 This debate is complicated. Some proponents of critical
legal studies, and other defenders of indeterminacy, claim that their views
were misunderstood or misrepresented.40 I shall take no position on that
question. Many have observed that critical legal studies is no longer a

35. Also known as strong indeterminacy. See, e.g., Lawrence B. Solum, “Indeterminacy,”
in A Companion to Philosophy of Law and Legal Theory, ed. Dennis Patterson (Oxford: Black-
well, 1996), p. 491.
36. “If one thought that rules are maximally indeterminate—that is, it is impos-
sible to say whether a rule applies in any individual case—there would be no point in asking
. . . whether they constrain those judgments once made.” Shapiro, “Judicial Can’t,”
p. 557 n. 26.
37. See, e.g., Solum, “Indeterminacy”; Andrew Altman, Critical Legal Studies: A Liberal
Critique (Princeton, N.J.: Princeton University Press, 1990), p. 19; Ken Kress, “Legal Indeter-
minacy,” California Law Review 77 (1989): 283–337, p. 283.
38. The view has been attributed to Joseph Singer, Duncan Kennedy, Clare Dalton,
Mark Tushnet, Gary Peller, and many others. See, e.g., Kress, “Legal Indeterminacy,” pp. 286–
87 (Singer, Kennedy); Altman, Critical Legal Studies: A Liberal Critique, pp. 19, 58 (Dalton,
Tushnet, Peller).
39. See, e.g., W.J. Waluchow, “Indeterminacy,” Canadian Journal of Law and Jurispru-
dence 9 (1996): 397–409; Christian Zapf and Eben Moglen, “Linguistic Indeterminacy and
the Rule of Law: On the Perils of Misunderstanding Wittgenstein,” Georgetown Law Journal
84 (1996): 485–520; Jules L. Coleman and Brian Leiter, “Determinacy, Objectivity, and
Authority,” University of Pennsylvania Law Review 142 (1993): 549–637; Brian Bix, Law,
Language, and Legal Determinacy (Oxford: Clarendon Press, 1993); Stephen J. Burton, Judg-
ing in Good Faith (Cambridge: Cambridge University Press, 1992); Kent Greenawalt, Law
and Objectivity (Oxford: Oxford University Press, 1992), pp. 3–89; Peter Drahos and
Stephen Parker, “Rule Following, Rule Scepticism and Indeterminacy in Law: A Conven-
tional Account,” Ratio Juris 5 (1992): 109–19; Altman, Critical Legal Studies: A Liberal
Critique; Scott Landers, “Wittgenstein, Realism, and CLS: Undermining Rule Scepticism,”
Law and Philosophy 9 (1990): 177–203; Kress, “Legal Indeterminacy”; Lawrence B. Solum,
“On the Indeterminacy Crisis: Critiquing Critical Dogma,” University of Chicago Law Review
54 (1987): 462–503.
40. See, e.g., Duncan Kennedy, A Critique of Adjudication (fin de siècle) (Cambridge,
Mass.: Harvard University Press, 1997), p. 24; Anthony D’Amato, “Aspects of Deconstruc-
tion: Refuting Indeterminacy with One Bold Thought,” Northwestern University Law Review
85 (1990): 113–27.
The Normative Classification of Legal Results 85

viable scholarly movement.41 Its demise may be partially attributable to

the critique of the indeterminacy thesis, but I shall not try to make that
case. Whatever the reason for its demise, the current consensus is that
most mandatory rules of law are not radically indeterminate and that even
partially indeterminate rules can guide judges.42 In Tamanaha’s words,
“the theoretical debate over legal indeterminacy expired with a consensus
that a degree of indeterminacy coexists with a substantial amount of pre-
dictability.” 43 Leading critical scholar Roberto Unger agrees that “[t]he
radicalization of indeterminacy is . . . a mistake.” 44 So I shall assume that
at least some legal questions have a finite number of legally correct
answers—one or more—plus some legally incorrect answers.

5.5.2 Legal Content

Although I shall assume that legal standards dictate determinate results in
some cases, however they do it, I shall not assume any particular theory of
legal content. My topic is what Dworkin calls the force of law—“the rela-
tive power of any true proposition of law to justify coercion in different
sorts of exceptional circumstance”—rather than the grounds of law—
“circumstances in which particular propositions of law should be taken to
be sound or true.” 45 I invite you to assume whatever theory of legal con-
tent you favor: inclusive positivist, exclusive positivist, Dworkinian, et
cetera. I shall not assume, for example, that legal content can always be
identified without evaluative argument, as exclusive positivists insist.46
My arguments do not presuppose a positivist theory of legal content,
much less an exclusive positivist theory, although they are compatible
with these theories. They are even compatible with what Hart calls “for-

41. Brian Z. Tamanaha, Law as a Means to an End: Threat to the Rule of Law (New York:
Cambridge University Press, 2006), p. 132 (critical legal studies “no longer exists”); Cass
R. Sunstein, “On Academic Fads and Fashions,” Michigan Law Review 99 (2001): 1251,
p. 1251 (critical legal studies has “disappear[ed]”); Jack M. Balkin, “Bush v. Gore and the
Boundary between Law and Politics,” Yale Law Journal 110 (2001): 1407–58, p. 1441 (crit-
ical legal studies was “dead as a doornail” in 2000).
42. See, e.g., Burton, Judging in Good Faith; Coleman and Leiter, “Determinacy, Objec-
tivity, and Authority.”
43. Tamanaha, On the Rule of Law, p. 124. See also, Alexander and Sherwin, The Rule of
Rules, p. 97 (“[W]e reject skepticism about the possibility of communication through
language”).
44. Roberto Mangabeira Unger, What Should Legal Analysis Become? (London: Verso,
1996), p. 121. See also Duncan Kennedy, “Freedom and Constraint in Adjudication: A Crit-
ical Phenomenology,” Journal of Legal Education 36 (1986): 518–62, p. 527 (“[P]erhaps there
are some results that you simply can’t reach through correct legal reasoning”).
45. Ronald Dworkin, Law’s Empire (Cambridge, Mass.: Harvard University Press, 1986),
p. 110.
46. See, e.g., Joseph Raz, The Authority of Law (Oxford: Clarendon Press, 1979); Scott
J. Shapiro, “On Hart’s Way Out,” in Hart’s Postscript, ed. Jules Coleman (Oxford: Oxford
University Press, 2000).
86 Limits of Legality: Part I

malist theories,” according to which rules alone always dictate legally cor-
rect answers.47 I shall not assume that the law consists exclusively of rules
as opposed to legal principles and other types of legal standards. My argu-
ments are compatible with, although they do not entail, Dworkin’s claim
that theories of legal content “need not leave entirely open the question
how judges should decide actual cases.” 48

5.5.3 Imperfection
One could agree that legal rules have determinate content while denying
that suboptimal-result cases exist. Indeed, many people who reject inde-
terminacy still appear to believe that the law, properly applied, never
actually forbids their preferred results. I affectionately call these people
Panglossians. Panglossians may complain when judges reach what they
regard as bad results, but they do not distinguish between morally bad and
legally incorrect. They condemn certain decisions as both morally bad and
legally incorrect.49 They praise other decisions as both good and faithful
to the law. Justice Antonin Scalia laments “[t]he inevitable tendency of
judges to think that the law is what they would like it to be.” 50
A Panglossian position could take either a conceptual or a contingent
form. Imagine a theory of legal content entailing that, by definition, a
positive rule is not legally valid if it dictates a suboptimal result. Accord-
ing to this theory, if a court decision is suboptimal, then ipso facto it is
legally incorrect. This entails that suboptimal-result cases do not exist. I
am unaware of any argument for such a theory of legal content, although
one might exist. Certain natural lawyers almost accept such a theory, but
even they do not go this far. Michael Moore notes that Cicero, Blackstone,
and other natural lawyers occasionally state that the justness of a norm is
sufficient for legal validity.51 No one now holds this view.52 Even this view
does not entail that the law always permits reaching optimal results,
because even just norms dictate suboptimal results in negative-gap cases.

47. H. L. A. Hart, The Concept of Law, 2nd ed. (Oxford: Oxford University Press, 1994),
ch. 7.
48. Dworkin, Law’s Empire, p. 112.
49. See, e.g., Michael Stokes Paulsen, “Accusing Justice: Some Variations on the Themes
of Robert M. Cover’s Justice Accused,” Journal of Law and Religion 7 (1989): 33–97, p. 35
(Roe v. Wade is “a lawless and immoral decision”). I am not suggesting that Paulsen is a
Panglossian.
50. Antonin Scalia, “Originalism: The Lesser Evil,” University of Cincinnati Law Review
57 (1989): 849–65, p. 864.
51. Michael Moore, “Law as a Functional Kind,” in Natural Law Theory: Contemporary
Essays, ed. Robert P. George (Oxford: Oxford University Press, 1992), p. 197.
52. Mark C. Murphy, “Natural Law Jurisprudence,” Legal Theory 9 (2003): 241–67, p.
242 (natural law theorists reject the “stupid position . . . that a norm’s status as a correct
moral norm is sufficient for its being a legal norm”).
The Normative Classification of Legal Results 87

Moore’s own position comes still closer to the view that suboptimality
entails legal incorrectness. In addition to endorsing Augustine’s slogan, lex
iniusta non est lex,53 Moore asserts that “an unjust court decision is not the
law of the case so decided.” 54 This entails that judges are never legally or
morally obligated to reach unjust results. If Moore would also classify as
unjust every result that counts as suboptimal on my definition, then his
position would entail that the law never requires suboptimal results.
However, I suspect that Moore would not classify as unjust every result
that counts as suboptimal on my definition. Recall the case of the widow,
Yasmin. I stipulated that Judge Lucas would have an all-things-considered
reason to rule in Yasmin’s favor if, counterfactually, the law gave him dis-
cretion to reach either result. That makes evicting Yasmin a suboptimal
result by definition. But does that fact entail that evicting Yasmin is unjust,
given that the law actually requires Judge Lucas to evict her? I do not
know what Moore’s answer would be, so I cannot attribute to him the
view that there are no suboptimal-result cases.
If a natural lawyer wanted to maintain that the law never requires sub-
optimal results, then he would have to go farther than Moore does. He
would have to insist that “the law,” properly understood, always reflects
morality rather than positive law whenever the two conflict. However, if
this claim were to gain wide currency with judges, then positive law could
no longer perform the important role of guiding action in morally contro-
versial areas. For the pro-choice judge, the law would permit abortion. For
the pro-life judge, it would forbid abortion. The public could have no
coherent law on topics about which judges morally disagree. So the
Panglossian natural lawyer would have to assume either that judges share
all moral opinions that bear on legal questions, or that having settled law
never matters more than doing justice, as one sees it, to the parties. These
are implausible assumptions.
Rather than defining suboptimal-result cases out of existence via one’s
theory of legal content, one could achieve the same goal by denying as a
contingent matter that the laws of one’s legal system ever require subop-
timal results. This is the path of a positivist Panglossian. She believes her
lawmakers to be exceptionally enlightened. Their laws require judges to
reach results that happen to be precisely the ones that each judge would
have had an all-things-considered reason to reach, had the law not required
anything of him. This would be a happy state of affairs. There would
be no suboptimal-result cases. Modern people, however, do not see their
lawmakers as infallible. I cannot name any positivist Panglossians.
One scholar who could be mistaken for a Panglossian is Douglas
Edlin. As do the Kadishes,55 Edlin contends that judges are legally

53. “An unjust law is no law at all.”

54. Moore, “Law as a Functional Kind,” p. 198.
55. See §3.6.
88 Limits of Legality: Part I

permitted to deviate from applicable legal standards under certain con-

ditions. But Edlin goes farther, arguing that judges have a legal obligation
to deviate from extremely unjust laws, one that outweighs their legal
obligation to apply the law.56 Consider the Fugitive Slave Act, which
required judges to return escaped slaves to their masters. Edlin argues
that judges did not have an all-things-considered legal obligation to
adhere to the act.57 Therefore, Edlin’s position implies that a case brought
under the act was not actually a suboptimal-result case, even though the
act required suboptimal results. However, his position does not entail
the nonexistence of suboptimal-result cases. Edlin delineates a tiny cat-
egory of cases in which judges have a legal obligation to deviate. This
category constitutes a small subset of the set of cases in which positive
law dictates suboptimal results. The rest of the cases in the latter set still
constitute suboptimal-result cases even if Edlin is correct. Edlin places
several restrictions upon deviation. I shall mention only two. First, he
insists upon a stringent “epistemic threshold” for deviation.58 He states
that the judge’s legal obligation to deviate overrides her legal obligation
to adhere only if she is extremely confident that the result otherwise
required by law is extremely unjust: “no judge can properly invoke
common law review unless she is as certain as she can be that a mistake
was made by a prior court or a legislature and that this mistake concerns
a matter of grave social importance that violates the judge’s deepest
convictions.” 59
Second, Edlin insists that the moral convictions of judges cannot justify
deviation unless those convictions have been incorporated into “authori-
tative legal sources.” 60 For example, a judge may not refuse to enforce the
death penalty simply because of her religious convictions. She must be
able to “express the wrongness of capital punishment in terms of a con-
crete legal violation.” 61 A judge whose conviction is not reflected in any
authoritative legal source will be unable to make a legal argument for
deviation based upon that conviction.
Because Edlin limits the scope of permissible deviation in these
ways and others, his position does not preclude the existence of subop-
timal-result cases. A judge who embraces Edlin’s philosophy will still
confront cases in which she has a legal obligation to reach results that
she believes to be suboptimal. Edlin appears to recognize and accept
this conclusion. He is no Panglossian, although he comes as close as
anyone I have found.

56. Douglas E. Edlin, Judges and Unjust Laws: Common Law Constitutionalism and the
Foundations of Judicial Review (Ann Arbor: University of Michigan Press, 2008), chs. 5–7.
57. Ibid., p. 133.
58. Ibid., pp. 139–43.
59. Ibid., p. 167.
60. Ibid., p. 15.
61. Ibid.
The Normative Classification of Legal Results 89

5.6 LEGAL PRINCIPLES

Dworkin observes that judges sometimes invoke legal principles in order

to decide cases in which the rules, taken alone, require suboptimal results.62
Inclusive positivists agree that legal principles can be authoritative.63 One
might wonder whether, on that premise, suboptimal-result cases even
exist. But even Dworkin understands that they do.64 A legal system with
suboptimal principles still generates suboptimal-result cases.65 Dworkin
appears to believe that the legal systems of the United States contain no
suboptimal principles, but that would be a happy contingency, and some-
one could reasonably disagree. One can imagine someone believing the
following propositions:

1. The principles to which Dworkin appeals in his arguments

about abortion, euthanasia, and affirmative action are, indeed,
authoritative principles of American law.66
2. Dworkin applies these principles correctly.
3. The results that Dworkin defends in these cases are morally
repugnant.

One who believes these three propositions could conclude that

Dworkin’s principles are morally defective, despite being legally authori-
tative. The cases mentioned are, from that individual’s perspective, subop-
timal-result cases. So suboptimal-result cases can arise even if the law
includes principles as well as rules. Reaching an optimal result sometimes

62. See §3.7.

63. See, e.g., Jules Coleman, The Practice of Principle (Oxford: Oxford University Press,
2001); Matthew Kramer, “How Moral Principles Can Enter into the Law,” Legal Theory 6
(2000): 83–108; W. J. Waluchow, Inclusive Legal Positivism (Oxford: Clarendon Press,
1994).
64. Dworkin writes the following:

Perhaps the law of the United States, properly interpreted in deference to integrity,
did include the Fugitive Slave Act enacted by Congress before the Civil War. If a
judge’s own sense of justice condemned that act as deeply immoral because it
required citizens to help send escaped slaves back to their masters, he would have to
consider whether he should actually enforce it on the demand of a slave owner, or
whether he should lie and say that this was not the law after all, or whether he should
resign.

Dworkin, Law’s Empire, p. 219.

65. See, e.g., Ken Kress, “Why No Judge Should be a Dworkinian Coherentist,” Texas
Law Review 77 (1999): 1375–427; Alexander and Sherwin, The Rule of Rules, ch. 8.
66. See, e.g., Ronald Dworkin, Freedom’s Law: The Moral Reading of the American Con-
stitution (Cambridge, Mass.: Harvard University Press, 1996); Ronald Dworkin, Life’s Domin-
ion: An Argument about Abortion, Euthanasia, and Individual Freedom (New York: Knopf,
1993); Ronald Dworkin, A Matter of Principle (Cambridge, Mass.: Harvard University Press,
1985).
90 Limits of Legality: Part I

requires judges to deviate not just from the rules but from all applicable
legal standards. This is deviation from the law simpliciter.
It is confusing that some authors call decisions that deviate from legal
rules exercises of “judicial discretion.” 67 This usage works for cases in
which the judge has some kind of legal authorization to deviate from the
rules. However, we should distinguish decisions in which the judge has
such legal authorization from those in which he lacks it. As noted, I use
discretion exclusively to refer to cases of legally authorized choice.68 A
judge can have discretion, in this sense, to depart from legal rules. By con-
trast, I use deviation to refer to legally unauthorized decisions—those that
contradict applicable law simpliciter. One cannot, by definition, have legal
discretion to deviate from the law simpliciter.
I have now reviewed and found plausible several assumptions underly-
ing the claim that suboptimal-result cases exist. In the legal systems that
interest me, the law sometimes requires the judge to reach a result that
she would have an all-things-considered reason to avoid if the law permit-
ted her to do so. In the next chapter, I shall ask whether judges have
pro tanto moral reasons to deviate in such cases. Before turning to this
question, I wish to draw another distinction between my topic and a
traditional concern of jurisprudence.

5.7 HARD CASES

For over a century, lawyers have used the adjective hard to refer to cases
that are especially problematic in various respects.69 One might assume
that suboptimal-result case is just a new term for hard cases. It is not. Hard
case has various meanings. The relation between suboptimal-result cases
and hard cases depends upon definitions. David Dyzenhaus, in Hard Cases
in Wicked Legal Systems, uses hard case to refer to cases in which “informed
lawyers disagree about the proper result.” 70 This usage is common, but
lawyers can disagree about different things: the legally correct result or
the result that the judge should reach, all things considered. Dyzenhaus
appears to intend the latter when he defines a hard case as one in which
“lawyers disagree about what the judge should decide.” 71 But on the same

67. The Kadishes call it deviational discretion. Mortimer R. Kadish and Sanford H.
Kadish, Discretion to Disobey: A Study of Lawful Departures from Legal Rules (Stanford, Calif.:
Stanford University Press, 1973), p. 42.
68. See §3.9.
69. See, e.g., Henchey v. City of Chicago, 41 Ill. 136, 141 (1866) (“Hard cases make bad
law”).
70. David Dyzenhaus, Hard Cases in Wicked Legal Systems: South African Law in the
Perspective of Legal Philosophy (Oxford: Clarendon Press, 1991), p. 1, citing Kent Greenawalt,
“Discretion and Judicial Decision: The Elusive Quest for the Fetters that Bind Judges,”
Columbia Law Review 75 (1975): 359–99, p. 386.
71. Dyzenhaus, Hard Cases in Wicked Legal Systems, p. 1.
The Normative Classification of Legal Results 91

page, he describes a case as hard if its “decision turns on contested points

of law,” 72 which suggests disagreement about legal correctness.
Although lawyers often disagree about how judges should decide sub-
optimal-result cases, they do not always disagree. All lawyers might agree
that a constitutional provision requiring people to eat their children is
legally valid, but they might also agree that judges should deviate from it.
All lawyers might agree that a parking ordinance dictates a slightly subop-
timal result in a certain case, but they might also agree that judges should
adhere to it. A case constitutes a suboptimal-result case not because
lawyers disagree about the morally optimal result, or about the legally
correct result, but because the law requires a suboptimal result. Some
suboptimal-result cases, but not all, are also hard in the first and/or second
senses.
Dyzenhaus also rejects the definition of hard cases as those in which
“there is no answer at law.” He thinks this definition begs the question
against Dworkin and in favor of positivism.73 Dworkin describes hard
cases as those that “cannot be brought under a clear rule of law, laid down
by some institution in advance,” 74 although he believes that there is a
legally correct answer in such cases.75 However, there are many cases in
which the only applicable legal standards—whether rules, principles, or
policies—dictate a unique result that is suboptimal. These are suboptimal-
result cases.
There is, however, a more generic sense of hard case that encompasses
most suboptimal-result cases. Suboptimal-result cases are usually hard
insofar as they present a difficult practical dilemma for anyone who
believes both that judges should adhere to the law and that judges should
avoid suboptimal results. Many lawyers hold both beliefs. Some give more
weight to the first, others to the second. Furthermore, many lawyers dis-
agree about what the law requires and about which results are subopti-
mal. As a result, lawyers will disagree sharply about “what judges should
decide” in many suboptimal-result cases. But that fact is not constitutive
of a suboptimal-result case.

72. Ibid.
73. Ibid., p. 1 n. 2.
74. Ronald Dworkin, “Hard Cases,” in Taking Rights Seriously (Cambridge, Mass.:
Harvard University Press, 1977), p. 81.
75. Ibid.
6

Reasons to Deviate

Assuming that suboptimal-result cases exist, how should judges handle

them? Dworkin acknowledges the issue:

[T]here are of course cases in which the institutional right is clearly settled by
established legal materials, like a statute, and clearly conflicts with background
moral rights. In these cases the judge seeking to do what is morally right is faced
with a familiar sort of conflict: the institutional right provides a genuine reason,
the importance of which will vary with the general justice or wickedness of the
system as a whole, for a decision one way, but certain considerations of moral-
ity present an important reason against it.1

A judge who faces a legally regulated question has four options:

1. Adhere to the law.

2. Deviate from the law.
3. Disqualify (recuse) herself from the case.
4. Resign from the bench.2

A wide range of reasons can bear on the judge’s choice, depending on

the question before her. In what follows I shall consider the following
questions. Do judges have a legal duty to adhere to the law? What incen-
tives do they have to adhere? What incentives, and other private reasons,
do they have to deviate, self-recuse, or resign? What moral reasons do they
have to adhere, deviate, self-recuse, or resign in suboptimal-result cases?
Does the validity of the regulating law change the judge’s reasons in these
cases? Can valid law give the judge reasons to reach a result that she
would otherwise have an all-things-considered reason to avoid, if the law
permitted her to avoid it? How does the law perform this trick, if it
does?
If a judge chooses to decide a suboptimal-result case, rather than self-
recusing or resigning, then she has two options: adhere or deviate. Two
related questions arise. First, is it always objectively wrong for her to

1. Ronald Dworkin, “A Reply to Critics,” in Taking Rights Seriously (Cambridge, Mass.:

Harvard University Press, 1978), pp. 326–27.
2. Macabre as it sounds, suicide counts as “resignation” for my purposes.

92
Reasons to Deviate 93

deviate, even when the law requires objectively bad results? Second, is it
always subjectively wrong for her knowingly to deviate, even when she
sincerely believes that the law requires bad results? In other words, is a
judge who knowingly deviates always blameworthy for doing so?3
There are many possible answers to these questions. Affirmative
answers are often suggested by the rhetoric of politicians and pundits who
criticize judges for deviation.4 According to the most restrictive principle
of judicial fidelity, deviation is never morally permissible, even when the
law mandates extreme violations of basic human rights. A judge who dis-
approves of the law may express his disapproval as forcefully as he wishes,
on this view, but he must adhere.5
I am not sure how many critics actually believe that judges should
unconditionally adhere. Critics of the bench do not always distinguish
clearly between disapproval of a result on the basis of their own opinions
about justice or policy, and belief that the judge has deviated. One won-
ders if critics would complain about judicial deviation if they were other-
wise pleased with the results. Historical laws have authorized genocide,
slavery, penal torture, and other horrors. Critics who insist that deviation
is always wrong, all things considered, do not seem to have these laws in
mind. But perhaps they believe that judges should simply resign from the
bench rather than apply such laws.
In opposition to the objective absolutist position, one could claim that
the ethical permissibility of deviation somehow depends on whether the
result in a certain case is desirable overall, as a matter of justice and public
policy. This would be to assert that a judge can have decisive moral rea-
sons to disregard the law—reasons that trump whatever legal or moral
reasons she may have to adhere to it.
In opposition to the subjective absolutist position, one might claim
that the subjective ethical permissibility of knowingly deviating from the
law depends on whether the judge believes (or reasonably believes) that
the legally required result in a certain case is desirable overall, as a matter
of justice and public policy. At the far extreme lies the view that judges
are always permitted to reach optimal results, even when the law forbids
it: they are always permitted to deviate in suboptimal-result cases. This
is a form of legal particularism, analogous to moral particularism. Moral
particularists claim that particular moral judgments can be justified
without appeal to any moral rules or principles.6 Legal particularists

3. See §2.3 on objective/subjective.

4. See cases mentioned in §1.1.
5. See, e.g., Khan v. State Oil Co., 93 F.3d 1358, 1362–64 (7th Cir. 1996) in which Judge
Richard A. Posner criticizes at length, but follows, the rule announced by the Supreme
Court in Albrecht v. Herald Co., 390 U.S. 145 (1968).
6. See, e.g., Jonathan Dancy, Ethics without Principles (Oxford: Oxford University Press,
2004). Dancy’s position is criticized in Sean McKeever and Michael Ridge, Principled Ethics:
Generalism as a Regulative Ideal (Oxford: Oxford University Press, 2006).
94 Limits of Legality: Part I

claim that particular legal judgments can be justified without appeal to

any legal rules or principles.7 Extreme legal particularism entails that
judges are always permitted to reach optimal results, notwithstanding
the law.

6.1 COGNITIVE DISSONANCE

Any judicial opinion raises four questions:

1. Does a sound legal argument for the holding exist?
2. Does the opinion advance a sound legal argument?
3. Is the holding optimal?
4. Is filing the opinion morally permissible, all things considered?
The easiest positions to understand give the same answer to all four
questions.8 Critics of a verdict often attack whatever argument the court
offers for it, however sound it may be, whereas supporters of the verdict
are often tempted to rally in defense of the argument, however weak.
Nevertheless, even the proponents of a holding can criticize the reasoning
and opponents can praise it. If someone’s answer to the third question
diverges from his answer to the first or second, then his answer to the
fourth becomes interesting. One possible view is that a negative answer to
the first question entails a negative answer to the fourth: it is morally im-
permissible for a court to reach a verdict for which no sound legal argu-
ment exists (even if the verdict itself is optimal). A stronger view is that a
negative answer to the second question entails a negative answer to the
fourth: it is morally impermissible for a court to advance a fallacious legal
argument, even if a sound legal argument exists for the verdict (itself
optimal).
Accepting either principle can cause cognitive dissonance. I shall illus-
trate using the famous 1965 case of Griswold v. Connecticut,9 in which the
U.S. Supreme Court held that a state statute criminalizing the use of con-
traceptives violated the constitutional rights of married couples. I choose
Griswold as a case in which a reasonable person might endorse the hold-
ing but reject the argument. It combines a result that is widely thought
optimal with reasoning that many legal scholars—including many who

7. See, e.g., Larry Alexander and Emily Sherwin, The Rule of Rules (Durham, N.C.: Duke
University Press, 2001), p. 28 (particularistic legal decision making is reasoning directly from
moral principles to particular decisions); Frederick Schauer, Playing by the Rules: A Philo-
sophical Examination of Rule-Based Decision-Making in Law and in Life (Oxford: Oxford
University Press, 1991), pp. 77–78.
8. A negative answer to the first question, of course, entails a negative answer to the
second. If one answers the second and third questions affirmatively, then I cannot imagine
why one would answer the fourth question negatively.
9. Griswold v. Connecticut, 381 U.S. 479 (1965).
Reasons to Deviate 95

support the result—find unsatisfactory.10 I shall not ask you to agree with
them that the Griswold argument is fallacious. Think of some other such
decision if you prefer.
Assume arguendo that Connecticut’s law was unjust. The ideal solu-
tion, then, would be to change the law: Connecticut could repeal its stat-
ute or the people of the United States could amend their Constitution
explicitly to ban such legislation. But neither event had occurred as of
1965. So we turn to a second-best solution: sound constitutional argu-
ments for invalidation. If such arguments existed in 1965, then the best
possible Supreme Court opinion would have used one. However, many
people believe that the majority opinion in Griswold contains no sound
legal argument for invalidation.11 Perhaps no such argument was even
available to the Court in 1965. An individual can rationally hold one of
these beliefs and also believe that Connecticut’s law was unjust and should
have been repealed. My question is this: assuming that such an individual
is correct on all counts, was it objectively wrong, all things considered, for
the Court to use a fallacious argument for invalidation? If one answers
affirmatively, then one may experience cognitive dissonance: a conflict
between one’s opposition to banning contraceptives and one’s judgment
that the Griswold Court acted wrongly. One could relieve this dissonance
by acquiring the belief that the Court was morally permitted to deviate in
Griswold.
Of course, the fact that acquiring this belief would relieve someone’s
cognitive dissonance does not bear on its truth. Defending its truth would
require, among other things, a substantive moral argument against banning
contraceptives. That belongs in another book. My upcoming arguments in
this book will support a conditional conclusion: if the Griswold result was
optimal, then reaching it was not necessarily impermissible, even if no sound
legal argument existed for it. Defending this conclusion does not require
any premises about the justice or injustice of banning birth control.

6.2 REASONS TO DEVIATE

I have explained how judges are in the business of force.12 They instruct
other officials to use force and to create, withdraw, and block threats of
force. The judge’s decision will influence how force is used no matter how

10. Robert H. Bork, The Tempting of America (New York: Simon & Schuster, 1990), pp.
95–100; Raoul Berger, Government by Judiciary: The Transformation of the Fourteenth Amend-
ment (Cambridge, Mass.: Harvard University Press, 1977), p. 18; Louis Henkin, “Privacy and
Autonomy,” Columbia Law Review 74 (1974): 1410–33, pp. 1421–22; Hugo L. Black, A Con-
stitutional Faith (New York: Knopf, 1968), p. 9; Paul G. Kauper, “Penumbras, Peripheries, Ema-
nations, Things Fundamental and Things Forgotten: The Griswold Case,” Michigan Law Review
64 (1965): 235–58, pp. 252–53; Griswold v. Connecticut at 527 (Stewart, J., dissenting).
11. See authors cited above.
12. See §2.8.
96 Limits of Legality: Part I

she decides the case. If she rules for the defendant, then she refrains from
threatening him with force and protects him from threats of force by
other officials. If she rules against him, then she instructs subordinate of-
ficials to threaten him with force, and to use force if necessary, and she
withdraws threats to use force against them if they do so.
In optimal-result cases, judicial reasons are straightforward. If the law
permits the judge to favor the defendant, when this is a superior outcome,
then the judge has pro tanto moral reasons to refrain from threatening him
with force and to protect him from threats of force by other officials. If
the law permits the judge to favor the plaintiff or prosecution, when this
is a superior outcome, then the judge has pro tanto moral reasons to
instruct subordinate officials to threaten the defendant with force. She
also has pro tanto moral reasons to withdraw threats to use force against
them if they so threaten the defendant.
In suboptimal-result cases, by definition, the judge would have an all-
things-considered reason not to perform the legally required actions were
they not required by law. If the law favors the plaintiff or prosecution,
then the judge would have all-things-considered reason to refrain from
threatening the defendant with force and to protect her from threats of
force by other officials, if the law permitted him to do so. If the law favors
the defendant, then the judge would have all-things-considered reason to
instruct subordinate officials to threaten the defendant with force, if the
law permitted him to do so. He also would have all-things-considered
reason to withdraw threats of force against subordinate officials if they
were so to threaten the defendant.
If the law permits a judge to avoid suboptimal results, then she has all-
things-considered reasons to avoid them. What happens to these reasons
in suboptimal-result cases? Compare two hypothetical cases of judges
sentencing convicts to prison. Both judges are legally authorized to impose
these sentences, which neither convict deserves as a moral matter. Sup-
pose that in the first case, but not the second, the law mandates the sen-
tence. According to the undermining principle,13 the fact that the law in
the first case requires the judge to impose the sentence undermines what-
ever reasons, including moral reasons, she might otherwise have to avoid
this result, at least in reasonably just legal systems. The principle entails
that judges in reasonably just legal systems have all-things-considered
moral reasons to adhere to the law in all cases. Not everyone believes the
undermining principle, but it is a popular view, perhaps the view of most
lawyers, judges, law professors, and legal philosophers.
In the rest of this chapter I shall present and criticize every argument
for the undermining principle of which I am aware. I conclude that none
is sound. I think all-things-considered moral reasons to avoid suboptimal
results are not undermined when the law requires such results, but rather

13. See §2.9.

Reasons to Deviate 97

survive as pro tanto reasons to deviate. It is, of course, a further question

whether these reasons are overridden by stronger reasons to adhere. I shall
examine that question in part II.
Although I conclude that we should reject the undermining princi-
ple, I do not know how to disprove it, nor shall I try to do so. Instead,
having challenged several arguments in favor of the undermining princi-
ple, I develop in part II a theory of adjudication that is assumes it to be
false.

6.3 NATURAL ROLES

Individuals can acquire new moral obligations—role obligations—when

they enter institutional roles. My critic might suggest that the very con-
cept of the judicial role entails that someone who takes into account rea-
sons for deviation is not, by definition, a judge anymore, just as a man who
marries is no longer a bachelor. This is to say that the judicial role is what
Arthur Applbaum calls a “natural role,” and that this role requires judges
to disregard reasons to deviate. Applbaum defines a natural role “by anal-
ogy to natural law,” as a role the obligations of which “follow from some
truths about the kind of creatures we are,” such that occupants of a role
who pursue ends incompatible with the natural ends of the role make a
“conceptual mistake.”14 However, Applbaum argues against natural roles
in favor of “practice positivism,” the view that “the concept of a practice
does not impose any general content requirements or restrictions on the
rules of all practices.”15 His conclusion seems correct to me, but I refer
skeptical readers to his argument.16
Practice positivism entails that roles can change over time. Even if
everyone today believes that the judicial role entails the undermining
principle, this could change. There is nothing conceptually incoherent or
irrational about a judge adopting for herself an alternative conception of
her role, one that allows her to consider moral reasons to deviate. If enough
judges begin to see their role in this way, then a new judicial role emerges.
I shall argue, moreover, that we have no reason to conceive of the judicial
role in accordance with the undermining principle.
Even if Applbaum is mistaken about natural roles and the judicial role
is a natural role that requires ignoring reasons to deviate, that would not
settle the practical question. If a judge, by definition, is someone who
ignores reasons to deviate, then we must ask whether anyone should
become a judge, in that sense. Perhaps there is something to be said for a
different and less restrictive role.

14. Arthur Isak Applbaum, Ethics for Adversaries: The Morality of Roles in Public and
Professional Life (Princeton, N.J.: Princeton University Press, 1999), p. 48.
15. Ibid., p. 51.
16. Ibid., pp. 51–58.
98 Limits of Legality: Part I

6.4 LEGAL POSITIVISM

This book does not plunge directly into the byzantine controversies about
the nature of law that have occupied legal philosophers for decades.
However, I often write as though some favorite positivist theses were
true, including the separability thesis,17 the social fact thesis,18 the sources
thesis,19 and the thesis that law consists at least partly of rules.20 I remain
agnostic about these four theses and shall not evaluate them. For me,
treating them as true is a convenient expository fiction: they make my
arguments easier to understand and more interesting.21 My ideal adver-
sary is a positivist who also believes that judges are morally obligated to
adhere in all cases, at least in reasonably just societies.22 I have no interest
in persuading him to abandon positivism, but I want him to abandon the
latter principle.
My tolerance of positivism could cause confusion. Someone might
suggest that positivism entails the undermining principle. This is true,
but only on a particular and somewhat outdated definition of positivism
as a theory of adjudication: the view that judges must decide cases only
on the basis of posited legal norms. Dworkin calls this view “strict
conventionalism.”23 It resembles what Dyzenhaus attacks as “posi-
tivism.”24 Strict conventionalism does, indeed, entail the undermining
principle. My arguments will entail, inter alia, that strict conventional-
ism is false.
However, strict conventionalism is not what most self-described posi-
tivists today mean by positivism. Schauer’s statement is typical:
Nothing about positivism compels the idea that only legally pedigreed rules
should guide judicial decisions. . . . Positivism is about legal validity and not
about ultimate action, and nothing in positivism commits any decision-maker,

17. See Jules Coleman, The Practice of Principle (Oxford: Oxford University Press,
2001), p. 151; John Austin, The Province of Jurisprudence Determined, ed. Wilfrid E. Rumble
(New York: Cambridge University Press, 1995), p. 157 (“The existence of law is one thing;
its merit or demerit is another”).
18. See, e.g., Coleman, The Practice of Principle, pp. 75–76. Cf. Benjamin C. Zipursky,
“The Model of Social Facts,” in Hart’s Postscript: Essays on the Postscript to the Concept of Law,
ed. Jules Coleman (New York: Oxford University Press, 2001).
19. Joseph Raz, “Legal Positivism and the Sources of Law,” in The Authority of Law
(Oxford: Clarendon Press, 1979), p. 47.
20. See, e.g., H. L. A. Hart, The Concept of Law, 2nd ed. (Oxford: Oxford University
Press, 1994), ch. 7.
21. Also, positivism is so popular that even a book aimed exclusively at positivists would
have a large readership.
22. This is what I call restrictive rule in §7.3.
23. Ronald Dworkin, Law’s Empire (Cambridge, Mass.: Harvard University Press, 1986),
p. 124.
24. David Dyzenhaus, “Positivism’s Stagnant Research Programme,” Oxford Journal
of Legal Studies 20 (2000): 703–22; David Dyzenhaus, Hard Cases in Wicked Legal
Systems: South African Law in the Perspective of Legal Philosophy (Oxford: Clarendon Press,
1991).
Reasons to Deviate 99

including a judge in a court of law, to treating positivistic norms as the exclusive

input into decision-making.25

Modern positivists understand themselves as advancing theories of

legal validity and legal content, not theories of adjudication.26 Their
position is compatible with many theories of adjudication, including
mine.

6.5 ORDINARY DISCOURSE

A supporter of the undermining principle might object that she rarely

hears an observer describe a judge as acting “immorally” or violating his
“moral duty” by rendering a decision, provided that the observer believes
that the judge has adhered to the law. If, as I claim, judges have moral
reasons to deviate in suboptimal-result cases, then why do observers rarely
call judges “immoral” for adhering in such cases?
I shall grant arguendo that one rarely hears an observer call a judge
“immoral” if the observer believes the judge’s decision to be required by
law. Calling decisions immoral, however, is commonplace. Consider Tama-
naha’s characterization of Brown. That decision, he asserts, was “undoubt-
edly correct from a moral standpoint,” but “arguably was not supported by
legal principles or reasoned elaboration or neutral principle.” It was
“impossible to justify in purely legal terms.”27 If the Court had upheld
school segregation, then Tamanaha and many others would have con-
demned the decision as immoral, albeit legally mandated.
A supporter of the undermining principle could respond as follows.
The fact that a result upholding school segregation is immoral, all things
considered, does not entail that the judge acts immorally, all things consid-
ered, by reaching that result if the law requires it. This may be true. How-
ever, you must remember that in part I, I am claiming only that judges
have pro tanto moral reasons to avoid suboptimal results. I need not and
shall not claim that these reasons prevail against all opposition. In part II,
I shall explain why adhering in some suboptimal-result cases is not just
morally permissible, but morally required.
Therefore, do not confuse my modest claim—that adhering in suboptimal-
result cases is pro tanto immoral—with the untenably strong claim that any
judge who adheres in such a case performs an all-things-considered immoral

25. Schauer, Playing by the Rules, p. 200. Similar ideas appear in Coleman, The Practice
of Principle, pp. 167–68; Joseph Raz, The Authority of Law (Oxford: Clarendon Press, 1979),
chs. 4, 10; David O. Brink, “Legal Positivism and Natural Law Reconsidered,” Monist 68
(1985): 364–87.
26. See, e.g., Coleman, The Practice of Principle, pp. 167–68; Joseph Raz, “Law and Value
in Adjudication,” in The Authority of Law (Oxford: Clarendon Press, 1979).
27. Brian Z. Tamanaha, On the Rule of Law: History, Politics, Theory (Cambridge:
Cambridge University Press, 2004), p. 80.
100 Limits of Legality: Part I

action. In ordinary discourse, to claim that someone acts “immorally” or

“violates a moral duty” is to claim that his actions are morally impermissible,
all things considered. So you should not be surprised if you rarely hear an
observer call a judge “immoral” for adhering in what the observer believes
to be a suboptimal-result case. One cannot infer from the fact that she
avoids this language that she believes that the judge has no reason
whatsoever to deviate. She might believe that the judge has a reason to
deviate but also a stronger reason to adhere. Furthermore, although some
people have a higher tolerance for unjust decisions than others, most
people can imagine a decision so unjust that they would call the judge’s
own decisional conduct “immoral,” even if they were convinced that her
decision was legally mandated.

6.6 FORMAL LEGALITY

The next argument for undermining appeals to the rule of law—often

called one of our preeminent political values.28 Rule of law is used to rep-
resent many different ideals.29 It denotes a concept with multiple
conceptions.30 I shall primarily concern myself with formal legality—the
most fundamental conception. A system exhibits formal legality to the

28. Not everyone supports it. See, e.g., Christine Sypnowich, “Utopia and the Rule of
Law,” in Recrafting the Rule of Law, ed. David Dyzenhaus (Oxford: Hart, 1999); John Hasnas,
“The Myth of the Rule of Law,” Wisconsin Law Review 1995 (1995): 199–233; Allan
C. Hutchinson, Dwelling on the Threshold: Critical Essays on Modern Legal Thought (Toronto:
Carswell, 1988), p. 40 (“The Rule of Law is a sham; the esoteric and convoluted nature of
legal doctrine is an accommodating screen to obscure its indeterminacy and the inescapable
element of judicial choice”); Morton J. Horwitz, “The Rule of Law: An Unqualified Human
Good?” Yale Law Journal 86 (1977): 561–66; Roberto Mangabeira Unger, Law in Modern
Society (New York: Free Press, 1976), pp. 238–42. Liberals and conservatives often dismiss
leftist and communitarian criticism of the rule of law as naïve. Tamanaha, On the Rule of Law,
p. 73 (“Such severe criticism [of the rule of law] could only be produced in a country in
which lengthy acquaintance with the rule of law confers intimate familiarity of its limitations,
and also leads to a sense of security that encourages forgetfulness about its benefits”); Judith
N. Shklar, “Political Theory and the Rule of Law,” in The Rule of Law: Ideal or Ideology, eds.
Allan C. Hutchinson and Patrick Monahan (Toronto: Carswell, 1987), p. 12 (“[D]estabilizing
the existing system of civil liberties and rights, and the individualistic ethos that sustains them
in the hope of building a truly fraternal order does not make sense. It shows little grasp of the
fragilities of personal freedom which is the true and only province of the Rule of Law”).
29. See, e.g., Richard H. Fallon, Jr., “‘The Rule of Law’ as a Concept in Constitutional
Discourse,” Columbia Law Review 97 (1997): 1–56. Andrei Marmor notes that “the various
ideas associated with the rule of law are often conflicting and not infrequently rather con-
fused.” Andrei Marmor, “The Rule of Law and Its Limits,” Law and Philosophy 23 (2004):
1–43, p. 1. As early as 1987, Judith Shklar could observe that “[i]t would not be very difficult
to show that the phrase ‘the Rule of Law’ has become meaningless thanks to ideological
abuse and general over-use.” Shklar, “Political Theory and the Rule of Law,” p. 1.
30. On the concept/conception distinction see John Rawls, A Theory of Justice
(Cambridge, Mass.: Harvard University Press, 1971), p. 5; Ronald Dworkin, Taking Rights
Seriously (Cambridge, Mass.: Harvard University Press, 1977), pp. 134–36. The basic division
Reasons to Deviate 101

degree that its laws are general, prospective, clear, publicized, stable, non-
contradictory, correctly applied, and realistically possible for subjects to
obey.31 Each branch of government has a role to play in upholding formal
legality. The legislature must take care to draft legislation that is general,
prospective, clear, possible to obey, and noncontradictory. Legislators must
not revise the law too often or enact contradictory provisions. The
executive, similarly, must not enact or enforce laws that violate these
conditions. The judiciary generally has authority to determine whether
the political branches have complied with these conditions. In addition,
judges sometimes act in a legislative capacity. When they do so they must
meet the same conditions as formal legislators: generality, prospectivity, et
cetera. Many rich debates have arisen concerning the conditions of formal
legality.32 Scholars have asked, for example, to what extent the law should
consist of rules, rather than other types of legal standards.33 These debates
relate directly to the prospectivity, clarity, and obedience conditions.
The legal systems that interest me exhibit formal legality to a sub-
stantial degree. Legal subjects cannot complain that they were unaware
of the legal consequences of their actions. Therefore, someone might
argue that a judge has no moral reason to refrain from imposing those
consequences. Imagine a law that, for the purpose of discouraging
Christianity, imposes a special tax on citizens who attend church. Assume
that singling out Christians in this way unjustly disadvantages them. This

within conceptions of the rule of law separates formal from substantive conceptions. Accord-
ing to Tamanaha, formal theories concern the “proper sources and form of legality,” whereas
substantive theories impose conditions on the content of the law. Tamanaha divides formal
theories into three subcategories: rule-by-law, which refers to the government using law as
an instrument of action; formal legality; and “democracy plus legality,” which holds that
consent determines the content of law. His substantive conceptions of the rule of law include
individual rights, the right of dignity and/or justice, and social welfare. Tamanaha, On the
Rule of Law, p. 91–92.
31. Lon L. Fuller, The Morality of Law, 2nd rev. ed. (New Haven: Yale University Press,
1969), pp. 32–38. Some earlier writers identify the rule of law with a subset of these condi-
tions. Montesquieu’s rule of law, for example, requires only procedural safeguards in the
criminal law, for the purpose of protecting individuals from violence at the hands of the
executive. Shklar, “Political Theory and the Rule of Law,” p. 5. Other theorists add conditions
of formal legality to Tamanaha’s basic list, such as the existence of a politically independent
and impartial court system; the right of every legal subject to seek legal redress and to
defend himself in courts of law; the right to appellate review of court decisions; and the right
to have a defense attorney provided when charged with a crime. See Robert S. Summers,
“The Principles of the Rule of Law,” Notre Dame Law Review 74 (1999): 1691–712, pp.
1694–95.
32. See, e.g., Ian Shapiro, ed., The Rule of Law: NOMOS XXXVI (New York: New York
University Press, 1994).
33. See, e.g., Louis Kaplow, “Rules Versus Standards: An Economic Analysis,” Duke Law
Journal 42 (1992): 557–629; Kathleen M. Sullivan, “The Justices of Rules and Standards,”
Harvard Law Review 106 (1992): 22–123; Duncan Kennedy, “Form and Substance in Private
Law Adjudication,” Harvard Law Review 89 (1976): 1685–1778.
102 Limits of Legality: Part I

unjust disadvantage would ordinarily constitute a pro tanto moral reason

not to confiscate the money. Likewise, the injustice of the law gives leg-
islators moral reasons to repeal it. However, let us assume that the law
also has the formal rule-of-law virtues: it is publicized in advance, intel-
ligible, prospective, possible to obey, et cetera. Everyone has a reasonable
opportunity to learn that attending church will subject him to a special
tax. One could argue that this fact undermines the judge’s moral reasons
not to confiscate the money—that formal legality entails the undermin-
ing principle.
This argument mistakes a necessary condition for a sufficient one. It
might indeed be wrong to enforce a law that lacked one or more of the
formal rule-of-law virtues. But the fact that an announced rule has all of
these virtues does not undermine whatever moral reasons one might have
to refrain from enforcing the rule. Here is an analogy. Pedro orders Ana,
his neighbor, never to attend church. He threatens to take ten dollars
from her purse every time she disobeys him. His order is intelligible to
her, given in advance, prospective, and possible to obey. These facts do not
undermine Pedro’s moral reasons to keep his hands off Ana’s money after
she disobeys his appalling command and attends church.
Similarly, the fact that a law, and the legal system to which it belongs,
has the formal rule-of-law virtues does not undermine a judge’s moral
reasons to deviate in a suboptimal-result case. In the words of Joseph Raz,
“[a] non-democratic legal system, based on the denial of human rights, on
extensive poverty, on racial segregation, sexual inequalities, and religious
persecution may, in principle, conform to the requirements of the rule of
law better than any of the legal systems of the more enlightened Western
democracies.”34
I will return to the subject of the rule of law in chapter 7. Before mov-
ing on I must emphasize that this section and the rest of this chapter are
directed exclusively against the undermining principle. My arguments
challenge neither the idea that judges have pro tanto reasons for adher-
ence nor the idea that these reasons can be decisive.

6.7 SUBMITTING TO THE STATE

Plato believes that individuals have an absolute moral duty to accept even
an unjust punishment from their state.35 In this spirit one could defend
the undermining principle as follows. If individuals have an absolute moral

34. Joseph Raz, “The Rule of Law and Its Virtue,” in The Authority of Law (Oxford:
Clarendon Press, 1979) p. 211; See also Jeremy Waldron, “The Rule of Law in Contemporary
Liberal Theory,” Ratio Juris 2 (1989): 79–96.
35. Cf. Plato, Crito 51c: “[I]f you cannot persuade your country you must do whatever
it orders, and patiently submit to any punishment that it imposes, whether it be flogging or
imprisonment. . . .”
Reasons to Deviate 103

duty to accept the legal consequences of their behavior, then judges have
no moral reason to protect them from these consequences. In politically
legitimate and reasonably just legal systems, individuals have an absolute
moral duty to accept legal consequences. Therefore, the argument con-
cludes, judges in such systems have no moral reason to protect individuals
from these consequences. They have no moral reason to deviate in order
to avoid suboptimal results.
There are two flaws in this argument. First, the major premise is sus-
pect. One cannot infer from the fact that one party has a duty not to resist
a certain threat to the conclusion that a second party has no moral reason
against imposing the threat. Nor can one infer that a third party has a
reason not to protect the first party from such a threat at the hands of the
second.
Second, few today (pace Plato) defend a categorical moral duty to
accept punishment or other legal consequences in suboptimal-result
cases.36 Even the much weaker principle that individuals have a pro tanto
moral duty to obey the law, as such, has fallen into disrepute with
philosophers.37 If there is no universal pro tanto duty to obey the law, then
there is no universal pro tanto duty to accept punishment.38 One who
would defend the undermining principle on the basis of the latter duty
must rebut many objections that have been raised against the duty to
obey.39 That may yet be possible, but defenders of the duty to obey pres-
ently labor under a heavy argumentative burden.

6.8 LEGAL AUTHORITY

Another argument for the undermining principle appeals to the legal

authority of judges: if the law authorizes a judge to reach a certain result,
then any contrary reasons are undermined and he should disregard them.
Consider the following hypothetical. Bud vandalizes Lou’s rosebush. Stip-
ulate that, as a moral matter, Bud owes Lou restitution. In civil society,
Lou has moral reasons not to use force against Bud’s person or property,

36. Even those who endorse a moral obligation to submit to punishment do not extend
the obligation to defendants who are factually innocent or who have been charged under
unjust laws. See, e.g., Gerard V. Bradley, “Plea Bargaining and the Criminal Defendant’s
Obligation to Plead Guilty,” South Texas Law Review 40 (1999): 65–82.
37. Joseph Raz, “The Obligation to Obey: Revision and Tradition,” in Ethics in the Public
Domain (Oxford: Clarendon Press, 1994); A. John Simmons, Moral Principles and Political
Obligations (Princeton, N.J.: Princeton University Press, 1979); M. B. E. Smith, “Is There a
Prima Facie Obligation to Obey the Law?” Yale Law Journal 82 (1973): 950–76.
38. This is not to say that wrongfully convicted defendants never have moral reasons to
accept punishment. Attempting to evade punishment might, for example, express disrespect
for the legal system or threaten the rule of law.
39. On these objections see William A. Edmundson, “State of the Art: The Duty to
Obey the Law,” Legal Theory 10 (2004): 215–59.
104 Limits of Legality: Part I

even for the purpose of extracting morally justified restitution. These

reasons also apply to Lou’s neighbor, Sol, and to all other private parties.
Lou sues Bud for damages. Sol, who happens to be a trial judge, is assigned
to the case, hears the evidence, and decides to award damages to Lou. As
the assigned judge, Sol is legally authorized to do so. His legal authoriza-
tion undermines some of his moral reasons not to use force against Bud—
reasons he had before hearing the case. Sol now has no moral reason not
to use force against Bud to the extent necessary for extracting morally
justified restitution under law.
Sol hears his next case. In this case, ruling against the defendant, Jon, is
a suboptimal result. Jon has done nothing to justify, in moral terms, the
use of force against him. However, Sol is legally authorized to rule against
Jon: Sol is the assigned judge and the applicable law, properly construed,
permits him to rule against Jon. The argument concludes that, just as Sol’s
legal authorization to use force against Bud undermines Sol’s moral rea-
sons not to use it, so does Sol’s legal authorization to use force against Jon
undermine Sol’s moral reasons not to use it.
The flaw in this argument is that Sol’s reasons not to use force against
Jon are natural reasons not to use morally unjustified force. By contrast,
Sol’s moral reasons not to use force against Bud were created by the
legal system. They are the same reasons that Lou has not to use force
against Bud in civil society, even when restitution is morally justified.
Sol has these reasons, too, but they are undermined for Sol when he is
assigned to Bud’s case and hears the evidence. But the fact that the law
undermines these reasons, which the legal system itself created, does not
imply that the law can undermine natural reasons. Sol’s legal authoriza-
tion to use morally unjustified force against Jon does not undermine his
natural reasons not to do so. At least if the law gives him discretion to
rule in Jon’s favor, then he should take these natural reasons into ac-
count. He should not ignore them simply because the law authorizes
him to do so.

6.9 INTENTION

In this section and the next, I turn to what may be the most interesting
arguments for the undermining principle. They use principles that have
been heavily debated in contemporary normative ethics. The first argu-
ment claims that a judge has no objective, pro tanto reason to deviate in a
suboptimal-result case, provided that she adheres with the proper inten-
tions. This is the argument:
1. A foreseeable, harmful effect of an action, Φ, on an innocent vic-
tim does not provide an agent with an objective pro tanto reason
to refrain from Φ-ing, if Φ-ing serves a greater good and the agent
does not intend the harmful effect.
Reasons to Deviate 105

2. Adhering to the law in a suboptimal-result case has foreseeable,

harmful effects on the losing party.
3. A judge who adheres to the law in a suboptimal-result case serves
a greater good.
4. A judge who adheres to the law in order to uphold the rule of law
does not intend to inflict harm upon the losing party.
5. Therefore, the harmful effects do not provide the judge with an
objective pro tanto reason to deviate from the law.
This argument raises several questions. Its first premise will remind
some readers of the doctrine of double effect. I shall avoid that terminol-
ogy because there are too many divergent formulations of double effect in
the literature. Also, the first premise is stronger than most formulations of
double effect, which concern all-things-considered duties. Many scholars
have rejected double effect.40 They would surely reject the first premise
as well. If they are correct, then my larger argument in this book has a
stronger foundation. But double effect has defenders.41 Rather than enter-
ing the debate, I shall temporarily assume arguendo that the first premise
is true.
The second question is, what greater good is served by adhering in
suboptimal-result cases? In such cases, adhering usually harms the legally
disfavored party more than deviating would harm the legally favored
party. So adhering does not contribute to any greater good if we consider
only the interests of the parties. But if we look beyond the interests of the
parties, then some goods are, arguably, served by adhering. Fulfilling
the expectations of a litigant is good, and the adhering judge fulfills the
expectations of the legally favored party, if the latter expected to win.
Keeping one’s promise is good, and one might argue that judges promise
to adhere to the law.42 One might also argue that obeying the law is intrin-
sically good and that judges obey the law when they adhere.43 Finally, one
might argue that maintaining the rule of law is good, and that adhering
maintains the rule of law.44
These are goods, certainly, but is any a greater good—good enough to
outweigh the harm inflicted by adhering in a suboptimal-result case? In

40. See, e.g., F. M. Kamm, Intricate Ethics: Rights, Responsibilities and Permissible Harm
(New York: Oxford University Press, 2007); Alison McIntyre, “Doing Away with Double
Effect,” Ethics 111 (2001): 219–55; Sophia Reibetanz (now Moreau), “A Problem for
the Doctrine of Double Effect,” Proceedings of the Aristotelian Society 98 (1998): 217–23;
Jonathan Bennett, The Act Itself (Oxford: Oxford University Press, 1995); Samuel Scheffler,
The Rejection of Consequentialism, revised ed. (Oxford: Oxford University Press, 1994);
Shelly Kagan, The Limits of Morality (Oxford: Clarendon Press, 1989).
41. See, e.g., Warren Quinn, “Actions, Intentions, and Consequences: The Doctrine of
Double Effect,” Philosophy and Public Affairs 18 (1989): 334–51; Thomas Nagel, The View
from Nowhere (Oxford: Oxford University Press, 1986).
42. See chapter 9.
43. See chapter 10.
44. See part II.
106 Limits of Legality: Part I

chapters 9 and 10, I shall argue that fulfilling expectations, keeping one’s
word, and obeying the law do not always outweigh this harm. In fact, in
suboptimal-result cases I think these considerations are undermined as
reasons to adhere.
The good that remains, therefore, is the rule of law. In part II, I explain
how adhering serves the rule of law. The rule of law is, arguably, a greater
good—good enough to outweigh the harmful effects of adhering in
a suboptimal-result case. So I shall assume for now that adhering to the
law in a suboptimal-result case serves a greater good.
Let us turn, then, to the third question raised by the intention argu-
ment: does a judge who adheres to the law for moral reasons in a subop-
timal-result case intend to inflict harm upon the legally disfavored party?
The debate surrounding double effect has shown that interpreting intent
for the purpose of applying moral principles is a notoriously complicated
matter.45 How might a judge deny that he intends to harm the legally
disfavored party? First, he could plausibly deny that he bears any ill will
toward the losing party. The judges who interest me are those who adhere
to the law for more innocuous reasons: to protect their reputations, to do
their jobs, to uphold their oaths of office, to support the rule of law, et
cetera. Judges who adhere to the law in suboptimal-result cases for such
reasons may feel genuine sympathy for the losing parties and regret that
adhering to the law in these cases inflicts undeserved harm. If they could
adhere to the law without inflicting the harm, then they would do so.
These judges contrast with those who enforce laws selectively against
individuals whom they dislike as individuals or as members of an ethnic
group. Such judges really do intend to harm the losing parties.46
However, the contrast between antipathic judges and ordinary adher-
ent judges should not obscure the fact that even ordinary judges intend to
harm the losing parties. Intent to harm does not entail ill will. It does not
entail that the harm is the agent’s final end. Otherwise intent to harm
would be the special province of vengeance-seekers, sadists, and few
others. Even the old villain of the double effect debates, the terror bomber,
seeks only to advance his just military campaign.47 If he could do so with-
out killing innocent civilians, then he would. He bears them no ill will, but
he intends to kill them nevertheless.
Different problems arise if we accept a theory of intention that is con-
sistent with the proposition that judges who adhere in order to uphold the
rule of law do not intend but only foresee the harm that they inflict upon

45. For a good recent treatment see William J. Fitzpatrick, “The Intend/Foresee Distinc-
tion and the Problem of ‘Closeness,’” Philosophical Studies 128 (2006): 585–617.
46. Recall the bigoted judge in §4.4.2.
47. The terror bomber, fighting a just war, targets civilians in order to terrorize his en-
emy into surrender. The strategic bomber, by contrast, bombs military targets knowing that
civilian casualties will result but not intending them. See, e.g., McIntyre, “Doing Away with
Double Effect,” p. 219.
Reasons to Deviate 107

legally disfavored parties. Such a theory of intention allows agents who

cause harm to disavow intent quite easily. Consider how such a theory of
intention would allow a deviating judge to defend herself. If she deviates
for moral reasons, as my judge does in suboptimal-result cases, then this
theory allows her to claim that she does not intend but only foresees
whatever harmful consequences her deviation may produce. The theory
still treats her as intending harm if vengeance or sadism motivates her or
if she deviates for the purpose of undermining the rule of law. But judges
who deviate in order to avoid suboptimal results have good intentions,
whatever their methods, on this theory of intention. So, on this theory,
whatever harmful effects deviating might have do not count against devi-
ating. I conclude that, whatever theory of intention we employ, the first
premise can exculpate adherent judges only at the price of exculpating
deviant judges, too, rendering it useless to someone trying to make the
case that judges have a moral duty to adhere in suboptimal-result cases.

6.10 MEANS

Instead of appealing to judicial intentions, one could assign moral rele-

vance to the distinction between harm caused by an agent as a means to
an end and harm caused as a side effect or aspect of action:
1. A harmful effect of an action, Φ, does not provide agents with an
objective pro tanto reason to refrain from Φ-ing, if Φ-ing serves
a greater good and the harmful effect is merely a side effect or
aspect of Φ-ing, rather than a means to the agent’s end in Φ-ing.
2. Adhering to the law in a suboptimal-result case has harmful effects
on the losing party.
3. A judge who adheres to the law in a suboptimal-result case serves
a greater good.
4. When a judge adheres to the law in a suboptimal-result case, the
harmful effects of his decision are not a means to his end, but
merely side effects or aspects of his decision.
5. Therefore, the harmful effects do not provide the judge with an
objective pro tanto reason to deviate from the law.
The first premise of this argument resembles F. M. Kamm’s principle of
permissible harm closely enough that I shall refer to it as such.48 I am not
sure that the principle is true, but I shall set aside this worry. The first thing
to notice about this argument is that if harming as a means or an end is
impermissible, then the argument is sound only if a judge who adheres in a
suboptimal-result case for moral reasons causes harm as an effect or aspect,

48. F. M. Kamm, Morality, Mortality: Rights, Duties, and Status, vol. 2 (New York: Oxford
University Press, 1996); Kamm presents her latest views in Intricate Ethics.
108 Limits of Legality: Part I

not as a means or end. This is somewhat plausible, although it is equally

plausible that the judge causes the harm as her means or even as her end.
More important, the principle of permissible harm does not permit
harming, even as an effect or aspect, unless the harm is an effect or aspect
of a greater good. In the previous section I suggested that adhering to the
law may, indeed, serve a greater good. I shall discuss this suggestion at
length in part II. But is the harm caused by adhering itself an effect or
aspect of that good or a means to it?
Suppose we adopt a theory of means according to which adhering harms
the legally disfavored party only as an effect or aspect, not as a means.
Judge Jack adheres to the law in a suboptimal-result case, unjustly sentenc-
ing Ivan to prison. Ivan suffers the harms of incarceration. But Jack’s reason
for inflicting these harms is to maintain the rule of law, not to torment Ivan.
Jack can deny that he has used Ivan as a means to an end. Ivan suffers harm
as a side effect or aspect of Jack’s decision to adhere, not as a means to
Jack’s end. Jack’s end of upholding the rule of law would not be compro-
mised if Ivan were to escape, miraculously, on the way to prison.
I agree that a plausible theory of means supports the conclusion that
judges who adhere for rule-of-law reasons do not use the legally disfa-
vored parties as means. This theory is not implausible, although it is not
obviously correct, either. However, this theory also allows the deviating
judge to make a similar claim. She can admit that deviating damages the
rule of law, but insist that it does so only as an effect or aspect, not as a
means to her end. Her reason for deviating is not to damage the rule of
law, but to protect the legally disfavored party from unjustified harm. This
end of hers would not be compromised if the rule of law were, miracu-
lously, to avoid damage. If anything, her argument based on the principle
of permissible harm seems stronger than the corresponding argument of
the adhering judge. Unlike litigants, the rule of law is not even a sentient
being with interests and rights.49 Therefore, although judges have moral
reasons not to damage the rule of law, such reasons do not undermine
their reasons not to inflict harm upon human beings. The principle of
permissible harm can exculpate adhering judges only at the price of
exculpating deviant judges.

6.11 RELIGIOUS OBJECTIONS

So far I have found no sound argument for the undermining principle. As

David Lyons notes, “a judge can be placed in a difficult moral predica-
ment, even in easy cases.”50 It is interesting to compare my conclusion
thus far with the views of writers in some religious traditions who have

49. Of course human beings suffer when the rule of law is damaged.
50. David Lyons, “Derivability, Defensibility, and the Justification of Judicial Decisions,”
in Moral Aspects of Legal Theory (Cambridge: Cambridge University Press, 1993), p. 120.
Reasons to Deviate 109

also addressed the plight of the judge whose personal convictions

condemn laws that he is asked to enforce, or actions that are lawful in
his jurisdiction. Although my position is rooted in secular ethical theory,
these studies are informative and some concur with me on the basic point
that judges have nonundermined, pro tanto moral reasons to avoid being
the agent of suboptimal results. Consider a devout Roman Catholic judge
facing one of the following decisions:

1. Whether to dismiss the case against a defendant who is being

prosecuted for sodomy under a law that has been invalidated on
constitutional grounds by the U.S. Supreme Court.
2. Whether to dismiss a criminal homicide case against a physician
who has lawfully assisted a suicide (e.g., in the State of Oregon).
3. Whether to uphold a sentence for criminal battery against a nurse
who forced life-sustaining treatment on a competent adult patient,
against his will and the law.
4. Whether to uphold a death sentence imposed by a lower court.
5. Whether to grant a waiver of parental consent to an abortion
when the petitioner, a minor child, is legally entitled to such a
waiver.
6. Whether to grant a properly filed divorce petition.

Sodomy, abortion, assisting suicide, failing to preserve the life of a

patient in one’s care, executing people, and getting divorced are evil
actions, according to the Catholic Church. There is a literature on
Catholics and other religiously observant judges facing such cases.51
Catholic judges and scholars hold that a judge who applies positive law in
these cases assists the evil actor, thereby engaging in “material coopera-
tion.” But material cooperation can be permissible in Catholic ethics.
Relevant factors include (1) whether there is a good reason for cooperat-
ing, such as avoiding a worse harm; (2) whether the cooperation is remote,

51. See, e.g., William H. Pryor, Jr., “The Religious Faith and Judicial Duty of an American
Catholic Judge,” Yale Law and Policy Review 24 (2006): 347–62; Rebekah L. Osborn,
“Beliefs on the Bench: Recusal for Religious Reasons and the Model Code of Judicial Con-
duct,” Georgetown Journal of Legal Ethics 19 (2006): 895–905; Mark C. Modak-Truran,
“Reenchanting the Law: The Religious Dimension of Judicial Decision Making,” Catholic
University Law Review 53 (2004): 709–816; Wendell L. Griffen, “The Case for Religious
Values in Judicial Decision-Making,” Marquette Law Review 81 (1998): 513–21; Daniel O.
Conkle, “Religiously Devout Judges: Issues of Personal Integrity and Public Benefit,” Mar-
quette Law Review 81 (1998): 523–32; Richard B. Saphire, “Religion and Recusal,” Marquette
Law Review 81 (1998): 351–63; John H. Garvey and Amy V. Coney, “Catholic Judges in
Capital Cases,” Marquette Law Review 81 (1998): 303–50; Scott C. Idleman, “The Role of
Religious Values in Judicial Decision Making,” Indiana Law Journal 68 (1993): 433–87;
Sanford Levinson, “The Confrontation of Religious Faith and Civil Religion: Catholics Be-
coming Justices,” DePaul Law Review 39 (1990): 1047–81; Lawrence B. Solum, “Faith and
Justice,” DePaul Law Review 39 (1990): 1083–106; Stephen L. Carter, “The Religiously
Devout Judge,” Notre Dame Law Review 64 (1989): 932–44.
110 Limits of Legality: Part I

rather than proximate; and (3) whether the cooperator avoids the danger
of a scandal, which the catechism defines as “an attitude or behavior which
leads another to do evil.”52
Judge William Pryor, a Roman Catholic, claims that the first and sec-
ond factors are present with respect to judges. First, “a judge has more
than a good reason to apply the law impartially in every case, because the
performance of that duty in a constitutional republic is a fundamental
safeguard for the protection of human liberty. The resources of the judi-
ciary are also scarce, so a judge is ordinarily obliged to perform his share
of the work of the judiciary.”53
Second, Pryor claims that “the performance of the judicial function is
likely to be remote from the intended evil act of the party before the
court; the typical scenario is where the judge determines that the law
does not empower the government to interfere with a third party’s choice
to commit an immoral act.”54 However, Pryor recognizes that adhering to
the law may constitute scandal, in the sense defined above.
Pryor believes that on these criteria a judge does not act immorally by
granting a divorce petition, but he notes that the abortion and death pen-
alty cases are more difficult for reasons of proximity and scandal. In some
suboptimal-result cases such as these, Pryor believes that judges have at
least pro tanto moral reasons to recuse themselves.55 Some Catholic
scholars believe that these are all-things-considered reasons.56 I shall take
up Pryor’s position again in §10.8 and compare it to my own.

52. Pryor, “The Religious Faith and Judicial Duty of an American Catholic Judge,” p. 361.
53. Ibid., pp. 360–61.
54. Ibid.
55. Ibid., p. 361.
56. “Catholic judges . . . are morally precluded from enforcing the death penalty. This
means that they can neither themselves sentence criminals to death nor enforce jury recom-
mendations of death. Whether they may affirm lower court orders of either kind is a ques-
tion we have the most difficulty in resolving.” Garvey and Coney, “Catholic Judges in Capital
Cases,” p. 305.
7

Adherence Rules

In this chapter and the next I return to the subject of adherence rules.1
I pose five questions. First, are adherence rules serious rules or pseudo-
rules (§7.1)? Second, what reasons do various adherence rules provide to
judges (§7.2)? Third, what reasons do various adherence rules exclude
from consideration (§7.3)? Fourth, which adherence rules, if any, do law-
makers have reasons to promulgate and why (§§7.4–7.11)? Fifth, how
should judges take adherence rules (promulgated or not) into account in
their decisions (chapter 8)?
The fourth and fifth questions are related. When deciding which adher-
ence rules to promulgate, a rational lawmaker2 considers the anticipated
effects of his decision on judicial behavior. So the answer to the fourth
question partly depends upon how judges, themselves, answer the fifth. In
this chapter I ask what reasons lawmakers have to promulgate adherence
rules, assuming that promulgation increases judicial conformity to pro-
mulgated rules.

7.1 ARE ADHERENCE RULES SERIOUS RULES?

Many writers split the universe of posited norms into two categories,
calling the first serious rules (or genuine rules, or simply rules) and the
second pseudorules or mere standards. Some writers classify a norm as a
serious rule only if it contains both a prescription and a factual predicate

1. Introduced in §4.1.
2. I use lawmaker broadly, referring to anyone who creates valid legal standards. Legis-
lative bodies are lawmakers, of course, but the executive and judicial branches also perform
some lawmaking functions. Presidents and governors sign executive orders. Administrative
agencies promulgate regulations. Judges create and revise the doctrines of the common law
and doctrines of constitutional implementation and statutory construction. See Brian Z.
Tamanaha, On the Rule of Law: History, Politics, Theory (Cambridge: Cambridge University
Press, 2004), p. 124 (“Once any degree of indeterminacy is recognized, it follows that the
claim that judges merely speak the law is implausible”). Judges also play a role in creating
standards of professional conduct for the judiciary itself. See, e.g., Final Draft Report of the
ABA Joint Commission to Evaluate the Model Code of Judicial Conduct, December 15,
2005 (listing members of Commission, including judges).

111
112 Limits of Legality: Part I

or hypothesis.3 On this theory, serious rules but not pseudorules can be

applied without using normative judgment.4 It is also sometimes
claimed that serious rules contain only precise and uncontroversial
terms in their formulations whereas pseudorules contain vague or con-
troversial terms.5 I agree with Alexander and Sherwin that the degree of
seriousness that can be attributed to a rule is an epistemological matter:
“a norm becomes a rule when most people understand it in a similar
way.” 6 I think the fundamental issue is how controversial the required
judgment is, not whether it is normative or factual. Some normative
judgments are, of course, more controversial than some factual judg-
ments. Whether a driver has swerved out of her lane will be less contro-
versial than whether she has driven responsibly. But some factual
judgments are more controversial than some normative judgments.7
Degree of controversy is what matters in classifying a norm as a serious
rule or a pseudorule.
Consider a simple adherence rule that requires judges to adhere in
every case. This rule is restated as follows: if adhering requires Φ-ing,
then you must Φ. The predicate is as follows: adhering requires Φ-ing.
The predicate of every adherence rule includes, inter alia, this proposi-
tion. Therefore, the greater the extent to which people disagree about
what the law requires, the lower the degree of seriousness that we can
attribute to adherence rules. Adherence rules are not serious rules unless
there are at least some cases in which people agree about what the law
requires.
I have assumed that the law is at least partially determinate.8 It pre-
cludes certain results in certain cases and there is a fact of the matter as to
which results it precludes. The determinacy assumption might seem to
entail that people agree, at least partially, about what the law requires. As
a logical matter it does not. The determinacy assumption is metaphysical,
not epistemological. Determinacy merely makes it possible for people to
agree, correctly, about what the law requires. It does not guarantee
agreement. Even complete legal determinacy would not entail that adher-
ence rules were serious rules. Nevertheless, I shall assume that there are at
least some cases in real legal systems in which sufficient agreement pre-
vails regarding what the law requires. Adherence rules can function as
serious rules in such systems.

3. Larry Alexander and Emily Sherwin, The Rule of Rules (Durham, N.C.: Duke Univer-
sity Press, 2001), p. 27.
4. Alan H. Goldman, Practical Rules: When We Need Them and When We Don’t
(Cambridge: Cambridge University Press, 2002), pp. 16–17, 107.
5. Alexander and Sherwin, The Rule of Rules, p. 29.
6. Ibid., p. 30.
7. Compare “There is life on other planets” with “Torturing people for fun is wrong.”
8. See §5.5.1.
Adherence Rules 113

7.2 PROTECTED REASONS AND CONTENT-INDEPENDENT

REASONS

Joseph Raz defines a mandatory rule as a “protected reason,” which is “a

systematic combination of a reason to perform the act . . . required by the
rule, and an exclusionary reason not to act for certain reasons (for or
against that act).” 9 Adherence rules are mandatory rules. If Raz is correct,
then an adherence rule provides a protected reason to adhere—a reason
to adhere plus an exclusionary reason that excludes certain competing
reasons (i.e., reasons to deviate from the law, to recuse oneself, or to resign
from the bench). We can understand adherence rules as what Jonathan
Dancy calls enablers.10 An authoritative adherence rule enables a certain
fact—that the law requires a certain result—to constitute a reason for the
presiding judge to reach that result when she decides the case.
According to Raz, exclusionary reasons can be cancelled, but they
always prevail within their uncancelled scope. His position has been
challenged—persuasively, I think.11 My arguments in this book become
easier to make if Raz is mistaken. However, my arguments are also com-
patible with Raz’s position. For the sake of argument, I shall assume him
to be correct hereafter. If he is wrong, then so much the better for my
argument.
H. L. A. Hart distinguishes between content-dependent and content-
independent reasons.12 As Raz puts it, “a reason is content-independent if
there is no direct connection between the reason and the action for which
it is a reason.” 13 A judge’s reasons for adhering can be content dependent,
content independent, or both. Consider a criminal defendant who is both
morally and legally innocent. The fact that he does not deserve punish-
ment, from a moral standpoint, constitutes a content-dependent, moral
reason for the judge to acquit. If the law permits her to acquit and the
defendant deserves acquittal, then the case is an optimal-result case. Judg-
es have content-dependent reasons to adhere in optimal-result cases.

9. Joseph Raz, Practical Reason and Norms, 2nd ed. (New York: Oxford University Press,
1990), p. 191.
10. Jonathan Dancy, Ethics without Principles (Oxford: Oxford University Press, 2004),
pp. 38–43.
11. Emran Mian, “The Curious Case of Exclusionary Reasons,” Canadian Journal of Law
and Jurisprudence 15 (2002): 99–124; Frederick Schauer, Playing by the Rules: A Philosophical
Examination of Rule-Based Decision-Making in Law and in Life (Oxford: Oxford University
Press, 1991), pp. 88–91; Larry Alexander, “Law and Exclusionary Reasons,” Philosophical
Topics 18 (1990): 5–22; Stephen R. Perry, “Second-Order Reasons, Uncertainty, and Legal
Theory,” Southern California Law Review 62 (1989): 913–94; Michael S. Moore, “Authority,
Law, and Razian Reasons,” Southern California Law Review 62 (1989): 827–96; Donald H.
Regan, “Authority and Value: Reflections on Raz’s Morality of Freedom,” Southern California
Law Review 62 (1989): 995–1095.
12. H. L. A. Hart, “Commands and Authoritative Legal Reasons,” in Essays on Bentham
(Oxford: Oxford University Press, 1982), p. 254.
13. Joseph Raz, The Morality of Freedom (Oxford: Clarendon Press, 1986), p. 35.
114 Limits of Legality: Part I

By contrast, authoritative adherence rules provide content-independent

reasons for judges to adhere to the law. These reasons apply in both
optimal-result and suboptimal-result cases. I shall examine arguments
with the following structure:

1. If a certain adherence rule has practical authority, then it provides

exclusionary reasons that exclude all reasons to deviate in suboptimal-
result cases.
2. The aforementioned adherence rule has practical authority for
judges (i.e., judges have all-things-considered reasons to obey it).
3. Therefore, judges have no reason to deviate in any suboptimal-
result case that they choose to decide.

In order to evaluate such arguments we must examine various adher-

ence rules, ascertaining if any has the requisite characteristics. In the next
section I shall discuss the reasons that adherence rules provide to judges.
Then I shall discuss judicial reasons to obey adherence rules and ask
whether any adherence rule has practical authority.

7.3 REASONS PROVIDED AND EXCLUDED BY

ADHERENCE RULES

Adherence rules divide into several overlapping categories based on the

reasons that they supply. We can distinguish between different rules
based on the range of cases in which they provide content-independent
reasons to adhere. The strongest rules provide reasons to adhere in all
cases. Weaker rules provide reasons to adhere in some but not all cases.
Even the weakest rules provide content-independent reasons to adhere
in required-result and optimal-but-not-required-result cases. Stronger
adherence rules provide reasons to adhere in suboptimal-result cases
as well.
We can also distinguish between adherence rules based on which rea-
sons to deviate, if any, they exclude. Restrictive rule provides reasons to
adhere in all cases and excludes all reasons to deviate. The excluded
reasons include moral reasons, some of which are reasons derived from
the negative effects that adherence has on losing parties. If restrictive
rule has practical authority, then it provides reasons to adhere and
excludes all reasons to deviate, even in suboptimal-result cases. In such
cases, however, the justification lying behind the exclusionary reason
provided by restrictive rule is inapplicable: there is a reason to deviate,
and excluding it does not serve the purpose of the exclusionary reason.
To this extent restrictive rule has characteristics that Raz’s theory does
not capture.
In order to capture this feature of restrictive rule we need Frederick
Schauer’s addendum to Raz. Schauer agrees with Raz that mandatory rules
Adherence Rules 115

include first-order reasons to act and exclusionary reasons not to act for
certain other reasons. However, Schauer finds Raz’s theory of rules incom-
plete because it misses the role of entrenchment. Rules, Schauer writes,
include “second-order reasons whose generality is entrenched even in those
circumstances in which the justification lying behind the second-order rea-
son is inapplicable.”14 On his view, “[a] rule exists (for some agent . . .)
insofar as an instantiation of a justification is treated (by that agent . . .) as
entrenched, having the power to provide a reason for decision even when
that instantiation does not serve its generating justification.”15
The weakest adherence rule that I shall discuss is permissive rule, which
has three features. First, it provides reasons to adhere in required-result
and optimal-but-not-required-result cases, but not in suboptimal-result
cases. Second, it excludes all private reasons to deviate.16 Third, it never
excludes moral reasons to deviate.17
In between permissive rule and restrictive rule lie various moderate
rules. They provide reasons to adhere in some but not all suboptimal-
result cases. They exclude all private reasons to deviate, plus some but not
all impartial reasons to deviate.18 A judge who obeys a moderate rule and
not restrictive rule sometimes uses force when not authorized to do so by
law and sometimes does not use force when required to do so by law.19
Restrictive rule is what Alexander and Sherwin call a pure rule: a pos-
ited norm that settles all practical questions that fall within its scope.20
Permissive rule and moderate rules are instances of what they call an
impure rule: a posited norm that settles some but not all unsettled ques-
tions that fall within its scope.
Hereafter, it will be convenient to refer to moderate and restrictive
rules collectively as nonpermissive rules. A nonpermissive rule excludes
private reasons to deviate in all cases, provides reasons to adhere in at least
some suboptimal-result cases, and excludes impartial reasons to deviate in
at least some suboptimal-result cases.

14. Schauer, Playing by the Rules, p. 93.

15. Ibid., p. 76. Schauer subsequently notes (pp. 190–91) that serious rules must be
treated as entrenched with respect to all moral reasons, not just their own justifications.
16. See §3.13.
17. One could also describe permissive rule as a rule that directs judges to treat restric-
tive rule in the most particularistic fashion possible, rather than treating restrictive rule as a
serious rule. Permissive rule resembles what Dworkin calls “legal pragmatism.” Ronald Dwor-
kin, Law’s Empire (Cambridge, Mass.: Harvard University Press, 1986), pp. 95, 151–75.
18. One could say that moderate rules instruct judges to treat restrictive rule in a par-
tially particularistic fashion.
19. See §2.9. One might say that moderate and permissive rules permit judges some-
times to use what Alec Walen calls “reasonable illegal force.” Walen, contra Rawls, endorses
the right of private parties in liberal societies to use it. I do not know if he would extend this
right to judges. Alec Walen, “Reasonable Illegal Force: Justice and Legitimacy in a Pluralistic,
Liberal Society,” Ethics 111 (2001): 344–73.
20. Alexander and Sherwin, The Rule of Rules, p. 30.
116 Limits of Legality: Part I

7.4 ADHERENCE RULES AND FORMAL LEGALITY

In sections 7.5–7.11, I shall consider reasons for lawmakers to promulgate

various adherence rules. Before doing so I wish to highlight a close con-
nection between adherence rules and the rule of law, specifically its di-
mension of formal legality.21 Many would argue that lawmakers should
promulgate adherence rules in order to promote and maintain formal le-
gality. The most important condition of formal legality for my purposes is
the congruence condition, which is precisely the requirement that the law
be applied “as written.” 22 Responsibility for maintaining congruence falls
primarily upon law enforcement officers, prosecutors, and adjudicators.
Police officers, for example, may not arrest individuals whose conduct
falls outside the scope of the criminal law, nor may prosecutors prosecute
the latter.
Congruence also has implications for the design of adjudicative institu-
tions. The main argument for judicial independence assumes the value of
congruence. Judicial independence is prized because it prevents legisla-
tors and executives from pressuring judges to deviate from the law as
might suit the political branches. Tamanaha notes that the “independent,
neutral judiciary” is the “final preserve of the rule of law.” 23 At the same
time, he emphasizes the importance of a robust judicial commitment to
fidelity. He speaks for many when he describes judicial commitment to
following the law as a “crucial feature” of the rule of law.24 He states that
“[t]he sina qua non of the rule of law is striving to decide cases according
to the law” 25 and claims that “the rule of law could not conceivably func-
tion without this group committed to the value of legality.” 26 Timothy
Endicott agrees that a community “lacks the rule of law to the extent that
its officials ignore the law.” 27
Because formal legality demands judicial fidelity, we need canons of
judicial ethics for neutralizing potential sources of deviation such as
conflicts of interest, corruption, and bribery.28 However, I shall focus on
judges who are not corrupt and who hear cases in which bribes and
conflicts of interest are not involved. I am primarily interested in what
honest judges, without conflicts of interest, can and must do to uphold
congruence and the other conditions of formal legality.

21. See §6.6.

22. As I have noted before, this requirement should not be understood as a requirement
to apply rules instead of other legal standards. Congruence requires judges to apply the law,
whatever that means.
23. Tamanaha, On the Rule of Law, p. 59.
24. Ibid., pp. 59, 125.
25. Brian Z. Tamanaha, Law as a Means to an End: Threat to the Rule of Law (New York:
Cambridge University Press, 2006), p. 244.
26. Tamanaha, On the Rule of Law, p. 59.
27. Timothy A. O. Endicott, “The Impossibility of the Rule of Law,” Oxford Journal of
Legal Studies 19 (1999): 1–18, p. 7.
28. See chapter 4.
Adherence Rules 117

The value of congruence also has implications for the standards and
procedures by which judges are chosen. Because judges are “uniquely
situated to undermine the rule of law,” 29 they

must be selected with the utmost care, not just focusing on their legal knowl-
edge and acumen, but with at least as much attention to their commitment to
fidelity to the law (not inclined to manipulate the law’s latent indeterminacy),
to their willingness to defer to the proper authority for the making of law
(accepting legislative decisions even when the judge disagrees).30

I shall focus primarily on judges who have already been appointed to

the bench by whatever procedure the law provides. I am interested in
Tamanaha’s suggestion that they “must be imbued with the sense that
their special task and obligation is fidelity to the law.”31 What ideas about
fidelity should lawmakers urge upon judges? What, exactly, should judges
believe about it?
If one supports formal legality and believes that judicial fidelity is
necessary to maintain it, then one is likely to support adherence rules and
to care about which adherence rules lawmakers promulgate. Different
conceptions of formal legality correspond to different adherence rules.
Many writers seem to interpret codified adherence rules as expressing
restrictive rule, which corresponds to a simple, intuitive conception of
formal legality. Restrictive rule provides a reason to adhere in all cases and
an exclusionary reason that excludes all reasons to deviate in all cases,
yielding an all-things-considered reason to adhere in all cases. A judge
who obeys restrictive rule never uses force when not authorized to do so
by law, sometimes uses force when authorized to do so by law, and always
uses force when required to do so by law.
In the rest of this chapter I shall consider a series of reasons for law-
makers to promulgate adherence rules. I draw upon arguments that sev-
eral authors have made for the creation of rules generally. I conclude that
lawmakers may have all-things-considered reasons to promulgate restric-
tive rule. In subsequent chapters, I turn to the reasons judges have to obey
adherence rules. I argue that most of what are usually proposed as reasons
to obey restrictive rule are not, in fact, good reasons. For example, the fact
that the judge took an oath to uphold the law is not a good reason to obey
restrictive rule. Other reasons I find are pro tanto reasons to obey restric-
tive rule, but they are often overridden by other moral considerations. In
other words, judges have no all-things-considered duty to obey restrictive
rule rather than a moderate rule. Codified adherence rules, if interpreted
as expressing restrictive rule, lack practical authority for judges. In the
remainder of this book I shall articulate and defend a moderate rule and

29. Tamanaha, On the Rule of Law, p. 59.

30. Ibid., p. 125.
31. Ibid., p. 59.
118 Limits of Legality: Part I

defend its practical authority. An alternative conception of formal legality

will gradually emerge as I do so.

7.5 SETTLEMENT AND PREDICTABILITY

Individuals who intend to act morally often disagree or are uncertain

about the morally correct course of action. These disagreements and
uncertainties generate problems of coordination, efficiency, and exper-
tise. Parties who disagree can end up in collisions, destructive competi-
tions, and violent conflicts. They suffer from frustrated expectations
and incur costs to avoid these frustrations. They forgo many benefits.
One function of conduct rules is to reduce these problems by provid-
ing authoritative settlement of concrete moral disagreements and
uncertainties.32
The problems of coordination, expertise, and efficiency carry over to
the judiciary. Judges need authoritative rules, too. If authoritative settle-
ment is sufficiently important, then a state of affairs in which all judges
obey adherence rules is superior to one in which no judge obeys adher-
ence rules. In other words, lawmakers have impartial pro tanto reasons to
want their judges to obey adherence rules. These arguments rest upon
empirical premises, but plausible ones in my view. I shall now examine
some of these arguments in greater detail.

7.6 COORDINATION AND RELIANCE

Imagine a jurisdiction with no promulgated adherence rules. In such a

jurisdiction, judges continue to adhere to the law when they have con-
tent-dependent reasons to do so. Such reasons could be moral, as when
the judge believes that the legally required result is optimal, or they could
be prudential, as when someone bribes the judge to adhere. But these
judges lack content-independent reasons to adhere, so they often deviate
when they have reasons to do so. They may deviate when they believe
that the legally required result is suboptimal, when they can profit finan-
cially by deviating, et cetera.
As compared to parties in jurisdictions with adherence rules, a party
to a dispute in this jurisdiction can more often successfully bribe the
judge or persuade him that her (the party’s) preferred result is optimal.
Accordingly, subjects in this jurisdiction have less confidence that judges
will adhere to the law. This lack of confidence has effects such as the
following. Motorists are less likely to obey traffic laws and to be more
worried about other motorists violating traffic laws. As a result, fewer

32. Alexander and Sherwin, The Rule of Rules, pp. 11–36.

Adherence Rules 119

people use the roadways, but the collision rate (per motorist) rises. Indi-
viduals are less likely to enter into mutually beneficial contracts. They are
more likely to litigate and less likely to settle disputes out of court.
Insurance rates rise. These are undesirable social consequences. Imagine
similar consequences across the entire legal system and you have a dys-
functional society.
Lawmakers can control these problems by promulgating adherence
rules. Judges who obey permissive rule do not deviate for reasons of par-
tiality such as self-interest. They will, accordingly, deviate in fewer cases
than if no adherence rule were promulgated, assuming that promulgation
has some positive effect on the obedience rate. As legal subjects come to
recognize this they will be more likely to obey traffic laws, enter contracts,
settle lawsuits, et cetera. Insurance rates will fall. Therefore, lawmakers
have a pro tanto reason to promulgate adherence rules at least as strong as
permissive rule.
However, judges obeying permissive rule still deviate when they believe
themselves to have impartial reasons for doing so. They deviate in what
they believe to be suboptimal-result cases. Although bribes are ineffective
with these judges, a party can still win by persuading her judge that her
favored result is normatively optimal. By promulgating restrictive rule,
instead, lawmakers can reduce the frequency of such victories, thereby
further increasing the extent to which subjects obey traffic laws, enter
contracts, settle lawsuits, et cetera. However, as Schauer acknowledges,
reaching the optimal result is sometimes more important than fulfilling
the expectations of others. Schauer concludes that “the force of the argu-
ment from reliance will vary across decision-making environments.”33
Nevertheless, in many environments lawmakers have a pro tanto reason to
promulgate restrictive rule.

7.7 EFFICIENCY

A judge who is motivated to reach optimal results will spend more time,
energy, and resources researching and deliberating if he obeys only per-
missive rule rather than restrictive rule. There is no limit to the range of
moral factors and evidence that he might choose to consider before decid-
ing each issue presented if he obeys permissive rule. By contrast, judges
who obey restrictive rule exclude all nonlegal reasons from their decision
making. This simplifies the decision process, thereby conserving decisional

33. Schauer, Playing by the Rules, p. 140. Lyons notes that, in unjust systems, “the rele-
vant advantages flowing from factors like certainty and predictability are likely to be primar-
ily conferred on those who profit from injustice.” David Lyons, “Derivability, Defensibility,
and the Justification of Judicial Decisions,” in Moral Aspects of Legal Theory (Cambridge:
Cambridge University Press, 1993), p. 132.
120 Limits of Legality: Part I

resources.34 So the value of efficiency gives lawmakers another pro tanto

reason to promulgate restrictive rule.

7.8 ERROR AND EXPERTISE

Rules in general promote coordination and efficiency at the cost of pro-

ducing suboptimal results in certain cases.35 In some environments this is
desirable overall, because coordination and efficiency are more important
than reaching optimal results. The same reasoning applies to adherence
rules. In suboptimal-result cases, by definition, a judge who adheres to the
law reaches a suboptimal result. Promulgating restrictive rule leads judges
to adhere in some such cases, yielding suboptimal results that would oth-
erwise not have occurred. Nevertheless, promulgating restrictive rule
could reduce the total number of suboptimal results reached in the
system. Judges who deviate do not always reach optimal results even
when they try to do so. Judges who do not obey restrictive rule are pre-
pared to take nonlegal considerations into account more often. Some of
them will make moral mistakes, reaching suboptimal results in optimal-
result cases. In such cases they would have reached optimal results had
they simply adhered.36
This is the error argument for promulgating adherence rules, also
known as the argument from expertise. It depends on the premise that
judges make more moral mistakes than “the law” makes. If this premise
holds, then lawmakers have another pro tanto reason to promulgate restric-
tive rule.

7.9 STABILITY

Schauer also presents an argument from stability: “[R]ule-based decision-

making narrows the range of potential decisions, and in doing so makes
changes from the status quo, both for better and for worse, more difficult
than would be the case were decision-makers freer to depart from the
categories and prescriptions of yesterday.” 37 This argument applies
straightforwardly to restrictive rule. A judge who obeys restrictive rule
will find it more difficult to change the law than one who obeys permis-
sive rule. Promulgating restrictive rule encourages judges to obey restric-
tive rule, thereby stabilizing the law. If the status quo is good, then
lawmakers have another pro tanto reason to promulgate restrictive rule.

34. Schauer, Playing by the Rules, pp. 145–49.

35. Alexander and Sherwin, The Rule of Rules, pp. 34–36; Goldman, Practical Rules,
pp. 32–33; Schauer, Playing by the Rules, pp. 128–34.
36. See Goldman, Practical Rules, pp. 4–5.
37. Schauer, Playing by the Rules, p. 157.
Adherence Rules 121

7.10 LEGITIMACY, AUTONOMY, AND RESPECT

Another class of arguments for restrictive rule invokes political legitimacy.

Consider Schauer’s claim that

[t]he traditional theory of judicial authority, under which legislatures make

the rules and judges apply them, is not based nearly so much on a fear that
judges will make errors in the process of engaging in open-ended decision-
making as it is based on the assumption that determination of questions of
substantive value should be for popularly responsible and responsive insti-
tutions such as legislatures, and not for non-majoritarian institutions such as
the judiciary.38

We should generalize this “traditional theory” so that it holds that law-

makers create the law (including but not limited to rules of law) and
judges adhere to it. Now we can ask why someone might believe that
a system is more politically legitimate if lawmakers (including judges
in their lawmaking capacity), rather than judges in their law-applying
capacity, determine “questions of substantive value.” Plausible answers
link legitimacy to values such as autonomy and respect. Many citizens
who participate in the political process in representative democracies
are attempting to exercise their autonomy. So are lawmakers, in many
instances, when they make law. By adhering to the law, judges facili-
tate the exercise of autonomy by citizens and lawmakers. By deviating,
judges interfere with this exercise. One might also argue that deviat-
ing expresses disrespect for lawmakers and citizens. Consider this
argument:
1. Lawmakers have a pro tanto reason to discourage judges from
interfering with the exercise of autonomy by lawmakers and
citizens, and from disrespecting lawmakers and citizens.
2. When a judge deviates from the law, she interferes with the
exercise of autonomy by lawmakers and citizens and she
disrespects them.
3. Promulgating restrictive rule discourages judges from deviating.
4. Therefore, lawmakers have a pro tanto reason to promulgate
restrictive rule.

Even if the second premise of the previous argument is false, a similar

argument could be advanced, substituting the following premises:
1s. Lawmakers have a pro tanto reason to discourage judges from causing
lawmakers or citizens to feel disappointment or disrespect.
2s. Acquiring the belief that a judge has deviated from the law may
cause a lawmaker to feel disappointed that a product of his lawmaking

38. Ibid., p. 159.

122 Limits of Legality: Part I

efforts has been ignored. He may also feel that the judge has treated him
disrespectfully. Citizens may have similar feelings insofar as they con-
sider the lawmaker to represent them.

7.11 FAIRNESS

Finally, the argument from fairness or comparative justice is worth men-

tioning, although it is a fallacious argument. The idea is that lawmakers
should promulgate restrictive rule in order to ensure that similar cases are
treated similarly. The fallacy here is that restrictive rule merely ensures
that legally similar cases are treated similarly. It does so, however, at the
cost of causing some morally similar cases to be treated differently and
some morally different cases to be treated similarly.39 So restrictive rule
does not implement the principle of comparative justice any more effec-
tively than permissive rule does.40 I will reintroduce the argument from
comparative justice in chapter 15 as an objection to my own recommend-
ed adherence rule.41
My conclusion in this chapter is that lawmakers have reasons to pro-
mulgate adherence rules, possibly even restrictive rule, on the assumption
that promulgation increases judicial conformity.

39. Ibid., pp. 135–37.

40. For a detailed refutation of the argument from formal justice, see David Lyons, “On
Formal Justice,” in Moral Aspects of Legal Theory (Cambridge: Cambridge University Press,
1993).
41. See §15.2.
8

Obeying Adherence Rules

8.1 PROMULGATION VERSUS OBEDIENCE

It is easy to imagine jurisdictions in which arguments from coordination,

reliance, efficiency, error, stability, and legitimacy are collectively strong
enough to support promulgating restrictive rule. In what follows I shall
accept arguendo that modern legal systems are among these jurisdictions.
I turn now to the reasons for judges to obey adherence rules. An agent
obeys a rule if and only if she is guided by it and complies with its dic-
tates.1 As Schauer notes, “the way in which and the extent to which, if at
all, rules become a part of the decisional process is ultimately determined
by the decision-maker alone.”2 How, if at all, should judges take adherence
rules into account? Are there any adherence rules that have practical
authority for judges?
We can distinguish between conduct rules promulgated by rule makers,
guidance rules followed by rule subjects, and appraisal rules used for eval-
uating the conduct of rule subjects. I shall also refer to decision rules used
by public officials deciding cases.3 The best conduct rule for lawmakers to
promulgate for legal subjects may be stricter than the best guidance and
appraisal rules.4 The best conduct rule to promulgate regarding highway
speed limits may be “Drive no more than sixty-five miles per hour,”
whereas a better rule for motorists to obey may be “Drive no more than
seventy-five.” Likewise, the better appraisal rule may be “Drivers who
exceed seventy-five miles per hour act wrongly.” The better decision rule
for highway patrol officers and trial judges to use may be “Ticket/convict
drivers who exceed seventy-five miles per hour.”
Parallel distinctions apply to adherence rules. “Judges,” Schauer writes,
“must necessarily make their own decisions whether to treat the rules
written in law books as the ones they will employ in reaching a decision.”5

1. Frederick Schauer, Playing by the Rules: A Philosophical Examination of Rule-Based

Decision-Making in Law and in Life (Oxford: Oxford University Press, 1991), p. 113.
2. Ibid., p. 128.
3. Confusingly, these decision rules are guidance rules for the officials in their official
capacities.
4. See, e.g., Meir Dan-Cohen, “Decision Rules and Conduct Rules: On Acoustic Sepa-
ration in Criminal Law,” Harvard Law Review 97 (1984): 625–77.
5. Schauer, Playing by the Rules, p. 146 (paraphrasing Duncan Kennedy).

123
124 Limits of Legality: Part I

Therefore, we should keep reasons for lawmakers to promulgate adherence

rules in a separate analytical category from any reasons that judges might
have to obey these rules. We can distinguish between first-order decision
rules codified and promulgated by lawmakers, guidance rules obeyed by
judges, appraisal rules used by critics of the bench, and second-order decision
rules used by decision makers (often other judges) during judicial disciplin-
ary proceedings. The best first-order decision rules may be stricter than the
best guidance rules, appraisal rules, and second-order decision rules.
Agents can take rules into account in several different ways. They can
treat rules as mere rules of thumb. Rules of thumb lack independent prac-
tical authority. They merely direct agents to an action that is usually
appropriate under the circumstances. When atypical conditions obtain—
conditions not contemplated by the rules—they lose all force.6
Another approach to rules is rule-sensitive particularism. A particularist
reasons directly from moral principles to particular decisions.7 A rule-
sensitive particularist also takes into account the value of having and fol-
lowing rules, so he will disobey less readily than a pure particularist, but
he still does not regard himself as rule-bound. He does not treat rules as
serious rules.
Should judges treat adherence rules as serious rules? Why not treat
them, instead, as rules of thumb, or use the rule-sensitive particularist
approach? The previous chapter rehearsed some reasons for lawmakers
to promulgate adherence rules. Similar considerations might constitute
reasons for judges to obey adherence rules, but this is not necessarily so.
Schauer recognizes the difficulty of providing reasons for agents to treat
rules as serious rules, as opposed to reasons for lawmakers to promulgate
them:

From the perspective of the agent deciding what to do, anything more rule-
bound than rule-sensitive particularism is difficult (though I think not impos-
sible) to defend. But from the perspective of some society or environment
deciding what institutions to establish, and whom to empower to do what,
stronger commitments to rules become considerably more appealing.8

8.2 OBEYING PERMISSIVE RULE

In this section I shall present an argument—drawing on Raz’s theory of

authority—that permissive rule has practical authority for some judges.
The authority of a rule for an agent, according to Raz, derives from

6. See Ibid., pp. 104–11.

7. Larry Alexander and Emily Sherwin, The Rule of Rules (Durham, N.C.: Duke Univer-
sity Press, 2001), p. 28.
8. Schauer, Playing by the Rules, p. 98, n. 26.
Obeying Adherence Rules 125

the fact that the agent conforms more closely to his actual reasons if he
complies with the rule than if he attempts to comply directly with those
reasons.
Here is a Razian argument for the conclusion that permissive rule has
authority for some judges. Judge Jerry is preparing to decide a case. He
believes that he has pro tanto prudential reasons to deviate. Perhaps he
has been offered a bribe or can otherwise profit financially by deviating.
Perhaps he is in love with the legally disfavored party. He also perceives
prudential reasons to adhere, including reasons of reputation, prestige,
occupational status, remuneration, and influence. Taking into account all
of these reasons, he concludes that his prudential reasons to deviate
outweigh his prudential reasons to adhere. However, he might be mis-
taken. Deviating might be a prudential error, notwithstanding Jerry’s
deliberation: perhaps his misconduct will be detected and his career
ruined.
If Jerry is so error-prone about when deviation is prudentially justified
that he makes fewer prudential mistakes when he obeys permissive rule,
then his prudential reasons to internalize permissive rule are stronger
than his opposing prudential reasons. However, if Jerry is not so error-
prone, then permissive rule lacks prudential authority for him. So the
scope of this Razian argument is limited. Nevertheless, I shall accept argu-
endo that permissive rule has prudential authority for some judges in
modern legal systems.
Judges also have moral reasons to obey permissive rule, corresponding
to the reasons for system designers to promulgate adherence rules, dis-
cussed in the previous chapter. Neil MacCormick mentions several rea-
sons that public officials have to accept the rule of recognition of their
system:
1. “[I]t is good that judicial decisions be predictable and contribute
to certainty of law, which they are and do when they apply known
rules in accordance with commonly shared and understood criteria
of recognition”;
2. “[I]t is good that judges stay within their assigned place in the
constitutional order, applying established law rather than inventing
new law”;
3. “[I]t is good that law-making be entrusted to the elected representa-
tives of the people, not usurped by non-elected and non-removable
judges”;9
4. “[T]he existing and accepted constitutional order is a fair and just
system, and accordingly the criteria of recognition of laws which it
institutes are good and just criteria which ought to be observed.”10

9. In the United States most state judges stand for reelection.

10. Neil MacCormick, Legal Reasoning and Legal Theory (Oxford: Clarendon Press,
1978), pp. 63–64.
126 Limits of Legality: Part I

These reasons correspond roughly to reasons identified in the previous

chapter. The first reflects coordination and reliance, the second and third
reflect legitimacy, and the fourth reflects error and stability. These are
reasons for system designers to promulgate adherence rules. They may
also be reasons for judges to obey permissive rule.

8.3 OBEYING RESTRICTIVE RULE

Along with many others, however, MacCormick appears to believe that

these are reasons for judges to obey restrictive rule, not just permissive
rule. In contrast to permissive rule, restrictive rule forbids judges from de-
viating even for moral reasons. In Schauer’s terms, restrictive rule “en-
trenches the generality” of the exclusionary reason. Restrictive rule excludes
moral reasons derived from the fact that adhering in a suboptimal-result
case imposes undeserved disadvantages on the losing party. Restrictive
rule provides a general reason to adhere and a general exclusionary reason
not to deviate for certain reasons, even if the first-order and exclusionary
reasons do not serve the instantiating justification of restrictive rule.
First, consider prudential reasons to obey restrictive rule. An agent can-
not have prudential reasons to obey a rule that excludes moral reasons, as
restrictive rule does. A prudential reason to obey restrictive rule can come
into conflict with moral reasons to deviate. In such conflicts the moral
reason typically wins. If a rule excludes moral reasons, then it has practical
authority for an agent only if he has moral reasons to obey it. Prudential
reasons cannot typically override or undermine moral reasons. A judge
might have a conclusive prudential reason to obey restrictive rule, but if
she also has moral reasons to deviate in suboptimal-result cases, then these
moral reasons would undermine or override her prudential reason to obey
restrictive rule. Restrictive rule excludes moral reasons, so restrictive
rule cannot have practical authority for judges unless they have moral
reasons, not just prudential reasons, to obey restrictive rule. So we
cannot yet claim that judges have an all-things-considered reason to obey
restrictive rule.

8.4 MORAL REASONS TO OBEY RESTRICTIVE RULE

My main question, then, is whether judges have moral reasons to obey

restrictive rule. This question has received little attention in the legal
and philosophical literatures. Philip Soper observed in 1984 that there
was “virtually no literature on the question of the judge’s obligation to
apply the law.”11 The literature has not grown much in the decades

11. Philip Soper, A Theory of Law (Cambridge, Mass.: Harvard University Press, 1984),
p. 41.
Obeying Adherence Rules 127

since.12 Robin West wonders, “Why are lawyers and legal academics so
oddly inattentive to the problem of the lawless adjudicator?”13 Despite
this dearth of scholarship, commentators have long insisted that judges
have strong moral reasons to adhere to the law in most cases, if not all.
According to Steven Smith, “virtually everyone assumes that courts nor-
mally have a duty to follow duly-enacted statutes.”14 Bentham appears to
have agreed.15 Kent Greenawalt states, “Judicial power to nullify the sub-
stantive law has never been suggested as a desirable feature of trial before
judges. . . .”16 Robert Summers writes that “the legal conclusions and any
reasons for action or decision on the part of the law’s addressees which . . .
arise under valid law, duly interpreted or applied, generally remain pe-
remptory for the law’s addressees, including courts and other tribunals.”17
Ronald Dworkin repeatedly asserts that the law binds judges: “The law we
have, the actual concrete law for us, is fixed by inclusive integrity. This is
law for the judge, the law he is obliged to declare and enforce.”18 A judge,
Dworkin insists, must not deviate from the law in order to advance her
moral or political beliefs (although Dworkin’s conception of “the law” is,
of course, more expansive than many). For example, even if a judge favors
extensive wealth redistribution, as does Dworkin,19 he
cannot appeal to the Constitution to order Congress or state legislatures to
adopt the economic and redistributive programs that equality of resources
demands. Nor, given the various constraints he accepts about how far he is free

12. But see M. B. E. Smith, “May Judges Ever Nullify the Law?” Notre Dame Law Review
74 (1999): 1657–71; Evan H. Caminker, “Why Must Inferior Courts Obey Superior Court
Precedents?” Stanford Law Review 46 (1994): 817–73; Michael Stokes Paulsen, “Accusing
Justice: Some Variations on the Themes of Robert M. Cover’s Justice Accused,” Journal of
Law and Religion 7 (1989): 33–97; Steven D. Smith, “Why Should Courts Obey the Law?”
Georgetown Law Journal 77 (1988): 113–64.
13. Robin West, “The Lawless Adjudicator,” Cardozo Law Review 26 (2005): 2253–61,
p. 2256.
14. Smith, “Why Should Courts Obey the Law?” p. 113. Smith proceeds to defend the
assumption.
15. Postema writes that Bentham’s doctrine of stare decisis requires obedience to estab-
lished precedent not only when, the rule being beneficial on the whole, nonetheless utility
calls for an exception in a particular case, but also when the rule itself seems arbitrary or
unreasonable. Gerald J. Postema, Bentham and the Common Law Tradition (Oxford: Oxford
University Press, 1986), p. 197.
16. Kent Greenawalt, Conflicts of Law and Morality (New York: Oxford University
Press, 1987), p. 367.
17. Robert S. Summers, “The Principles of the Rule of Law,” Notre Dame Law Review 74
(1999): 1691–712, p. 1694.
18. Ronald Dworkin, Law’s Empire (Cambridge, Mass.: Harvard University Press, 1986),
p. 406. See also Jeremy Waldron, “Kant’s Legal Positivism,” Harvard Law Review 109 (1996):
1535–66, pp. 1538–39 (“The official’s failure to implement the law because he believes that
it is unjust . . . is tantamount to abandoning the very idea of law”). Waldron presents the
quoted claim as Kantian and appears to endorse it.
19. Ronald Dworkin, Sovereign Virtue: The Theory and Practice of Equality (Cambridge,
Mass.: Harvard University Press, 2000).
128 Limits of Legality: Part I

to read statutes to promote his view of justice, can he read into welfare and
taxation schemes provisions equality of resources would approve.20

Although these programs would serve “[p]olitical integrity and justice,”

Dworkin thinks, the judge “would violate integrity himself” if he were to
read existing law as favoring them. Judges
may not read the abstract moral clauses [of the Constitution] as expressions of
any particular moral judgment, no matter how much that judgment appeals to
them, unless they find it consistent in principle with the structural design of the
Constitution as a whole, and also with the dominant lines of past constitu-
tional interpretation by other judges.21

Therefore, he continues,
[e]ven a judge who believes that abstract justice requires economic equality
cannot interpret the equal protection clause as making equality of wealth, or
collective ownership of productive resources, a constitutional requirement,
because that interpretation simply does not fit American history or practice, or
the rest of the Constitution.22

The manifest consensus, restated in my terminology, is that judges have

strong pro tanto reasons to obey restrictive rule. Is there also consensus
that judges have all-things-considered reasons to obey restrictive rule?
Some language used by commentators suggests this, although it is some-
times qualified with “normally,” “generally,” et cetera. The U.S. Supreme
Court states that “a court is to apply the law in effect at the time it renders
its decision, unless doing so would result in manifest injustice or there is
statutory direction or legislative history to the contrary.”23 These qualifica-
tions seem to anticipate cases in which the law requires suboptimal results,
but writers who make such exceptions usually limit them to cases in
which the law requires an extreme injustice—what I call “extreme imper-
missible-result cases.” Greenawalt, for example, finds it “conceivable that
some convictions would be so abhorrent that judicial defiance of the law
would be defensible, and this conclusion may be true even if such an
action is considered to be outside the law in every sense.”24 Justice Antonin
Scalia admits that he would invalidate a statute providing for public

20. Dworkin, Law’s Empire, p. 404.

21. Ronald Dworkin, Freedom’s Law: The Moral Reading of the American Constitution
(Cambridge, Mass.: Harvard University Press, 1996), p. 10.
22. Ibid., p. 11. Elsewhere he asserts that “[n]either a Marxist nor a fascist would find
enough present law distinctively explained by his political philosophy” to claim that his
views are consistent with the law of the United States. Dworkin, Law’s Empire, p. 408.
23. Bradley v. Richmond School Board, 416 U.S. 696, 711 (1974).
24. Greenawalt, Conflicts of Law and Morality, p. 368. Compare David Dyzenhaus,
Hard Cases in Wicked Legal Systems: South African Law in the Perspective of Legal Philosophy
(Oxford: Clarendon Press, 1991), p. 60 (South African judges were legally obligated to
ignore or overturn racist laws).
Obeying Adherence Rules 129

flogging even though he believes it to be constitutional.25 And some

writers accept the so-called Radbruch formula, which denies legal validity
to extremely unjust positive laws26 in the spirit of the Augustinian-
Thomistic slogan: lex iniusta non est lex.27 One who endorses the Rad-
bruch formula can accept restrictive rule while denying that judges are
morally obligated to obey extremely unjust “laws.”
David Lyons is the rare writer who actually emphasizes the justifiabil-
ity of deviation, stating that “it seems possible for a judicial decision to be
justified, all things considered, even when it is contrary to a decision that is
required by law.”28 He even challenges what he calls “the doctrine of legal-
istic justification,” which holds that “a judicial decision is at least ‘prima
facie’ justified if it is required by law.”29 Lyons appeals to the fact that one
who attempts to argue that the law merits respect under certain condi-
tions has implicitly conceded that those conditions are not necessarily
satisfied. He concludes by denying “that legal considerations . . . them-
selves provide any measure of justification for judicial decisions.”30
The proposition that all judges in all legal systems have an all-
things-considered reason to obey restrictive rule entails the doctrine of legal-
istic justification. If Lyons is correct, then it is not the case that all judges in
all legal systems have all-things-considered reasons to obey restrictive rule.
But consider the adjectives that Lyons uses to describe systems in which
judicial decisions required by law lack “prima facie” justification: “unfair and
unjust, undemocratic and oppressive, exploitative and inhumane.”31 Lyons’
point appears to be that judges in very bad regimes have no reason to
adhere, a point on which Greenawalt and others concur.
What about judges in reasonably just and effective legal systems? Lyons
leaves open the possibility that judges in such systems have moral reasons

25. Antonin Scalia, “Originalism: The Lesser Evil,” University of Cincinnati Law Review
57 (1989): 849–65, p. 864. From the context of this remark I infer that Scalia means that
flogging is too inhumane for him to uphold, regardless of the fact that originalism entails
that it is constitutional. I do not understand him to mean that he would invalidate a flogging
statute in deference to nonoriginalist precedent that mandates invalidation.
26. Gustav Radbruch, “Gesetzliches Unrecht und übergesetzliches Recht,” Süddeutsche
Juristen-Zeitung 1 (1946): 105–8; translated by Bonnie Litschewski Paulson and Stanley L.
Paulson, “Statutory Lawlessness and Supra-Statutory Law,” Oxford Journal of Legal Studies
26 (2006): 1–11; Robert Alexy, “A Defence of Radbruch’s Formula,” in Recrafting the Rule of
Law: The Limits of Legal Order, ed. David Dyzenhaus (Oxford: Hart, 1999).
27. See, e.g., Norman Kretzmann, “Lex Iniusta Non Est Lex: Laws on Trial in Aquinas’
Court of Conscience,” American Journal of Jurisprudence 33 (1988): 99–122.
28. David Lyons, “Derivability, Defensibility, and the Justification of Judicial Decisions,”
in Moral Aspects of Legal Theory (Cambridge: Cambridge University Press, 1993), p. 120
(emphasis in original). For a qualified defense of nonacquiesence, limited to administrative
agencies, see Samuel Estreicher and Richard L. Revesz, “Nonacquiescence by Federal Ad-
ministrative Agencies,” Yale Law Journal 98 (1989): 679–772.
29. Lyons, “Derivability, Defensibility, and the Justification of Judicial Decisions,” p. 120.
30. Ibid., p. 124.
31. Ibid., p. 131.
130 Limits of Legality: Part I

to adhere, even in suboptimal-result cases. Indeed, he agrees that “[j]udi-

cial decisions may be justified directly, on their merits, on the merits of
the laws that require them, or on the merits of the legal system as a
whole.”32 To reject the doctrine of legalistic justification is simply to deny
that such justifications always apply.
Despite the virtual consensus that judges have strong reasons to adhere,
few deny that judges do, in fact, deviate from the law, sometimes know-
ingly. A judge who deviates from the law even once does not obey restric-
tive rule. The Kadishes observe that it is “striking . . . the extent to which
courts in their judicial opinions proclaim obedience to a principle far
more restrictive than the principle they employ in practice.”33 They sug-
gest that judicial deviation from legal rules is now seen as a normal and
expected practice:
[J]udicial departures from the obligation to decide in accordance with the
established rules has [sic] become a deeply ingrained and characteristic feature
of the judicial process, a feature sustained by the milieu in which judges operate.
So government officials take judicial lawmaking into account when acting in
their own roles; legal analysis and argumentation rest on it; initiates into the legal
system encounter the tension between analytical and result-oriented thinking as
one of the central features of a process they are expected to master.34

Given the consensus that judges have moral reasons to obey restrictive
rule, which reasons are these? Consider some of the reasons discussed in
the previous section as reasons for lawmakers to promulgate restrictive
rule: coordination, reliance, efficiency, stability, error reduction, and allo-
cation of authority. One might hope that these reasons also serve as rea-
sons for judges to obey restrictive rule. We shall see.

8.4.1 Predictability
It is better, ceteris paribus, for people to be able to predict court decisions.
Predictability facilitates the honoring of legitimate expectations, enables
subjects to have advance warning when sanctions may be applied, and makes
available information required to coordinate actions over time.35 Judges
serve these predictability values when they obey restrictive rule. However, a
particular deviant decision compromises predictability values only to a min-
imal extent. This is significant because in suboptimal-result cases there are
opposing values, such as the value of not subjecting parties to undeserved
disadvantages. In many suboptimal-result cases the damage that deviation
does to predictability is less important from a moral standpoint than the

32. Ibid., p. 128.

33. Mortimer R. Kadish and Sanford H. Kadish, Discretion to Disobey: A Study of Lawful
Departures from Legal Rules (Stanford, Calif.: Stanford University Press, 1973), p. 88.
34. Ibid., p. 91.
35. Alan H. Goldman, Practical Rules: When We Need Them and When We Don’t (Cam-
bridge: Cambridge University Press, 2002), p. 34.
Obeying Adherence Rules 131

undeserved disadvantages imposed by adherence. Judges who obey

restrictive rule give predictability values lexical priority over all other
values. They could instead treat restrictive rule as a rule of thumb, deviat-
ing whenever the disvalue of adherence outweighs the disvalue of devia-
tion. This approach allows judges to serve the values of predictability
along with other values.36
There are, of course, cases in which the parties have relied upon an
expectation that the law would be enforced and disappointing these
expectations would be bad enough to outweigh any moral reason to devi-
ate. The judge has an all-things-considered reason to adhere in such cases,
but these are not suboptimal-result cases by definition, because judges
would have no all-things-considered reason to reach a different result in
such cases even if the law permitted it. Therefore, the expectations of
parties do not constitute reasons for judges to obey restrictive rule.

8.4.2 Efficient Use of Decisional Resources

Judges have at their disposal a finite supply of resources for making deci-
sions—decisional resources. These include time, mental energy, money,
assistance from judicial clerks, et cetera. Figuring out what the law requires
consumes some of these resources—more of them in more complex cases.
A judge consumes additional decisional resources if, having determined
the legally correct result, she chooses to ascertain whether it is suboptimal
and, if so, to decide whether to deviate. She conserves resources if, having
determined the legally correct result, she simply reaches it. Obeying re-
strictive rule conserves more than any other reasonable adherence rule.37
Legal systems are conceivable in which judges should obey restrictive
rule. Restrictive rule is optimal if, unrealistically, there are no suboptimal-
result cases. It might also be optimal for the odd judge whose mere choice
to ask herself whether the legally correct result is suboptimal causes her
to consume excessive decisional resources. Imagine an “obsessive” judge
who spends 100 hours agonizing over any moral question she asks herself.
She should obey restrictive rule (if she remains on the bench), because
she cannot be trusted to resolve moral questions in a reasonable amount
of time.
However, few real judges are so obsessive. Most are, and know them-
selves to be, capable of budgeting their time and resources in reasonable
ways. A reasonable judge can contemplate whether a legally correct result
is suboptimal without consuming more than a small portion of decisional
resources. In many cases she will quickly conclude whether or not the
legally correct result is suboptimal in her opinion. Some decisions will

36. Goldman acknowledges that the need for predictability does not by itself justify
treating restrictive rule as a serious rule. Ibid., p. 35.
37. Equally efficient rules include ones that prescribe decision by coin flip or (per-
versely) consistent deviation.
132 Limits of Legality: Part I

take longer, but when evaluating optimality is taking too long most judges
are capable of cutting short deliberation and simply deciding the case, at
which point adhering probably makes sense. This is a reasonable personal
policy: attempt to evaluate optimality, but if the process becomes too
costly, then adhere by default. This is a variation on permissive rule. Obey-
ing restrictive rule, by contrast, gives conservation of decisional resources
lexical priority over all other values. The question should be whether the
total cost in extra decisional resources consumed by judges who reject
restrictive rule is outweighed by the value of avoiding the undeserved
disadvantages that restrictive rule would require them to inflict. Until we
answer that question we cannot conclude that judges have reasons to
obey restrictive rule. The fact that doing so conserves resources does not
settle the matter.

8.4.3 Error/Fallibility
Next, consider the argument from error, which recalls Raz’s theory of
authority.38 Raz writes that a directive has authority over someone if “the
alleged subject is likely better to comply with reasons which apply to him
(other than the alleged authoritative directive) if he accepts the directives
of the alleged authority as authoritatively binding and tries to follow
them, rather than by trying to follow the reasons which apply to him
directly.”39
Everyone has many laws the disobedience of which would directly and
tangibly benefit him if he could avoid the legal consequences. In only a
small fraction of these cases would disobedience be just or serve the gen-
eral welfare. In all other cases, individuals will conform more closely to
their actual reasons if they obey the law than if they disobey, assuming
that they have strong reasons to act morally, whether or not doing so ben-
efits them. However, individuals are tempted to disobey whenever they
believe that disobedience will benefit them and that they can avoid the
legal consequences. In such cases they will sometimes conclude, although
usually incorrectly, that they have moral reasons to disobey. Therefore, if
they often disobey whenever they believe themselves to have moral rea-
sons to disobey, then they will disobey too often and will fail to conform
to their actual reasons. However, if they exclude moral reasons to disobey,
then their degree of aggregate conformity to their actual reasons will be
higher (so the argument assumes). So some legal conduct rules have prac-
tical authority.
One could argue, analogously, that restrictive rule has authority for
judges. This would involve claiming that judges conform more closely to
their actual reasons, in the aggregate, if they obey restrictive rule than if

38. Schauer, Playing by the Rules, pp. 149–55; Joseph Raz, The Morality of Freedom
(Oxford: Clarendon Press, 1986), chs. 2–4.
39. Raz, The Morality of Freedom, p. 53.
Obeying Adherence Rules 133

they obey moderate or permissive rules. Suppose a judge is motivated to

avoid suboptimal results. However, he is often mistaken about which
results are suboptimal. He is mistaken so often that he reaches fewer sub-
optimal results during his career if he obeys restrictive rule than if he tries,
case by case, to avoid suboptimal results.40
I am not sure how many judges are such unreliable identifiers of sub-
optimal-result cases or how many would accept this fact about them-
selves. One’s estimates would seem to depend in part on the degree of
overlap between one’s own moral opinions and those of the judges in
question. Although most judges are prepared to agree that certain other
judges have worse moral judgment than “the law,” how many judges
believe this of themselves?
I can imagine a judge who has learned from experience that his own
moral judgment is less reliable than the law. He remembers many occa-
sions when he was very confident that a legally required result was mor-
ally suboptimal, but later concluded that the law had been correct, after
all. Or he might believe, on the basis of his knowledge of many other
judges throughout history, that the law is morally correct more often than
judges. A rational, morally motivated judge will obey restrictive rule if he
believes, for whatever reason, that the law has better moral judgment
than he.
Perhaps all judges are very bad at distinguishing suboptimal-result
cases from optimal-result cases. A judge who disregards moral reasons to
deviate will end up adhering in every case. Jack is such a judge. Over the
course of his career he adheres in every case, including 100 suboptimal-
result cases. But suppose Jack is so prone to misidentify suboptimal-result
cases that a commitment to deviate in every case that he believed to be a
suboptimal-result case would lead him to adhere in 50 suboptimal-result
cases and to deviate in 51 optimal-result cases. In that case, ceteris pari-
bus, Jack’s overall track record would be better if he adhered consistently
than if he tried to deviate in every perceived suboptimal-result case. This
is a Razian argument for the authority of restrictive rule.
However, judges have plenty of reasons to believe that their own moral
judgment is at least as good as that of the law. Because gap cases are inev-
itable, it is always possible that the case at bar is one of these.41 Did the
lawmaker really anticipate this case? Are lawmakers really more reliable
than judges on moral matters? Are the salient facts of the case really more
distorting than informative?42 As long as judges on average identify subop-
timal-result cases correctly more often than not, judges who deviate in
these cases will conform more closely to the reasons they have than if
they adhere consistently. I see no reason to anticipate that judges will do

40. See also Schauer, Playing by the Rules, p. 126.

41. See §5.4.
42. For an affirmative answer, see Adrian Vermeule, Judging under Uncertainty: An Insti-
tutional Theory of Legal Interpretation (Cambridge, Mass.: Harvard University Press, 2006).
134 Limits of Legality: Part I

better if they adhere consistently than if they deviate in every case that
they believe to be a suboptimal-result case.
Judges are situated very differently from private individuals. In legal
systems that effectively police judicial conflicts of interest, bribery, and
cases of undue influence, judges almost never hear cases in which devia-
tion would directly and materially benefit them. Judges do not, therefore,
have material incentives to conclude that they have moral reasons to de-
viate when they do not. At least their incentives for self-delusion are
nowhere near as strong as those of private individuals contemplating their
own cases. So it is not as evident as in the case of ordinary folk that judges’
degree of aggregate conformity to their actual reasons is higher if they
always disregard what they see as moral reasons to deviate. I do not claim
that a judge who knows himself well would be unreasonable to conclude
that his own moral judgment is inferior to that of the law. But I also sub-
mit that it is not unreasonable for a judge to conclude that his moral
judgment is at least as good as that of the law. In other words, I do not think
there is a Razian argument, applicable to all judges, for obeying restrictive
rule. As Goldman notes, “in the absence of special reasons to suspect bias,
ignorance, or some other special source of fallibility, the assumption by rule
makers that they know better is . . . presumptuous. . . .”43
More important, even a judge who accepts that he is “worse than the
law” at identifying suboptimal results as a general matter will hear cases in
which he is especially confident that the legally required result is subop-
timal. The Razian offers no reason for him to obey restrictive rule in such
cases. Only a judge who never concludes that the legally required result is
suboptimal has a reason to adhere in every case. Such a judge is not actu-
ally obeying restrictive rule.
My general point is that in suboptimal-result cases adhering to the law
does not serve the justifications for promulgating restrictive rule to a
degree that is sufficient to justify obeying it. That is part of the definition
of a suboptimal-result case. Suboptimal-result cases open a gap between
the lawmaker’s reasons to promulgate restrictive rule and a judge’s rea-
sons to obey it.44 The lawmaker has all-things-considered moral reasons to
promulgate it, but the judge lacks all-things-considered reasons to obey
(at least when considering the reasons examined thus far).45
Alexander and Sherwin argue that the nature of rules entails that gaps
cannot be closed by any means.46 They argue that rulemakers cannot
eliminate the moral flaws of rules by improving the content of the rules.47
If they are correct, then lawmakers cannot close the gap that judges face

43. Goldman, Practical Rules, p. 35.

44. See §5.4 on gaps.
45. This is an instance of what Schauer calls the “asymmetry of authority.” Schauer,
Playing by the Rules, p. 128.
46. They consider, inter alia, rule-sensitive particularism, presumptive positivism, exclu-
sionary reasons, sanctions, and deception. Alexander and Sherwin, The Rule of Rules, ch. 4.
47. Ibid., p. 35.
Obeying Adherence Rules 135

in suboptimal-result cases by improving either legal rules or adherence

rules. There will always be cases in which an adherence rule, however
ideal, mandates suboptimal results.
Alexander and Sherwin’s conclusion that the gap is ineliminable seems
to entail that judges have no reason to obey restrictive rule rather than per-
missive rule. Yet these coauthors appear to believe that judges should none-
theless obey restrictive rule.48 This is an awkward position to maintain.49
I shall describe in part II an adherence rule that Alexander and Sherwin do
not consider. I think it superior to both strong and permissive rules.

8.4.4 Indirect Consequentialism

Readers who favor indirect consequentialism might argue that a version
of this moral theory supports obeying restrictive rule. According to Brad
Hooker’s recent formulation of rule consequentialism, an act is wrong if
and only if it is forbidden by the code of rules the internalization of which
by the overwhelming majority of the population has maximum expected
value.50 Hooker’s principle supports obeying restrictive rule only if every
value-maximizing code of rules forbids deviating in every suboptimal-
result case. Is this true?
There are, to be sure, some codes that permit deviation in some subop-
timal-result cases and that do not maximize value. Consider a code that
includes a subjective version of permissive rule that permits the judge to
deviate in any case that she believes to be a suboptimal-result case. Such
a code would probably not maximize expected value if internalized by
the overwhelming majority of judges. If too many judges have bad moral
judgment, then they would deviate in more optimal-result cases than the
suboptimal-result cases in which the judges with better judgment deviate.
But permissive rule is not the only rule that permits deviation in some
suboptimal-result cases. Suppose Jack’s moral judgment is good enough
that if he deviates whenever he thinks the case is a suboptimal-result case,
then he actually deviates in more suboptimal-result cases than optimal-
result cases. If the overwhelming majority of judges were to internalize a
code that included both permissive rule and Jack’s particular criteria of
optimality, then value would be maximized. So there is, after all, a value-
maximizing code that permits Jack to deviate. Therefore, judges are per-
mitted to obey permissive rule and not required to obey restrictive rule.
Of course, this conclusion depends on the highly unrealistic stipulation
that an overwhelming majority of judges correctly identify suboptimal-result
cases and deviate only in these. But Hooker’s formulation warrants this

48. See, e.g., ibid., pp. 145–50; Larry Alexander, “‘With Me, It’s All er Nuthin’: Formal-
ism in Law and Morality,” University of Chicago Law Review 66 (1999): 530–65.
49. Schauer differs from Alexander and Sherwin in that he seems prepared to have
lawmakers inculcate irrational “rule-worship” in judges. Schauer, Playing by the Rules, p. 132.
50. Brad Hooker, Ideal Code, Real World: A Rule-Consequentialist Theory of Morality
(Oxford: Oxford University Press, 2000), p. 32.
136 Limits of Legality: Part I

stipulation. In part II, I shall argue that a defense of a nonpermissive rule

requires rejecting this stipulation and plunging even further into the realm
of nonideal theory than Hooker ventures.51

8.4.5 Contractualism
An alternative to rule consequentialism that has generated great interest
in recent years is the contractualism of T. M. Scanlon. I shall argue that
contractualism does not support obeying restrictive rule, either. Contrac-
tualism holds that “an act is wrong if its performance under the circum-
stances would be disallowed by any set of principles for the general
regulation of behavior that no one could reasonably reject as a basis for
informed, unforced general agreement.”52 Contractualism could be
applied to adjudication theory in many different ways, few of which have
been examined in the literature. I explore just one possibility. A contrac-
tualist theory of adjudication might hold that a decision is wrong if reach-
ing it under the circumstances would be disallowed by any set of
adjudication rules that no one could reasonably reject as a basis for
informed, unforced general agreement.
A set of adjudication rules that includes no nonpermissive rule permits
deviation in every suboptimal-result case. Who could reasonably reject
this set? In order to answer this question, we must determine on whom
the burden of such rules falls. We must look at each individual separately
and ask, what are the effects on him if all judges obey these rules?
A defeated party usually wants to reject whatever rule permitted his
defeat. Whether such rejection is reasonable depends on whether an al-
ternative rule would have given anyone as strong reason to object as the
present rule gives him. So we need to determine who is burdened when
judges obey permissive rule and who is burdened when they obey restric-
tive rule. Then we must compare these burdens.
The fact that a decision is deviant does not entail that it makes the
defeated party worse off than an adherent decision would have made his
adversary. It all depends on the facts of the case. Whether a set of adjudi-
cation rules containing permissive rule can be reasonably rejected depends
entirely on the criteria used to identify suboptimal-result cases. Consider
a sexist criterion that classifies as suboptimal any result that facilitates the
ownership of real estate by women. A judge who accepts permissive rule
and this criterion will deviate whenever the law awards real property to a
female litigant. People (especially women!) can reasonably reject a set of
adjudication rules containing permissive rule and this criterion of
optimality. Therefore, according to the contractualist theory under con-
sideration, the judge should obey restrictive rule.
51. Nonideal theory is discussed in chapter 11.
52. T. M. Scanlon, What We Owe to Each Other (Cambridge: Belknap, 1998), p. 153. I
challenge the capacity of contractualism to capture conventional morality in Jeffrey Brand-
Ballard, “Contractualism and Deontic Restrictions,” Ethics 114 (2004): 269–300.
Obeying Adherence Rules 137

This is not, however, the right kind of argument against permissive

rule. Its conclusion for judges is conditional: if your criteria for identifying
suboptimal-result cases can be reasonably rejected, then you should obey
a nonpermissive rule. This conclusion is not helpful to us because we are
trying to give judges reasons to adhere even when they are confident that
the law requires a suboptimal result. If contractualism is true and a certain
result is actually suboptimal, then it cannot be reasonable to reject criteria
simply because they classify that result as suboptimal. So a judge who is
rational, accepts contractualism, and believes that a certain result is sub-
optimal will conclude that no one can reasonably reject criteria that clas-
sify that result as suboptimal.
Consider the following combination: a statute that forbids women to
own real estate and a set of optimality criteria that classify as suboptimal
any result that prevents women from acquiring property when a similarly
situated man would receive title. According to these criteria, this statute
typically requires results that make the legally disfavored party worse off
than deviating makes the legally favored party. The legally favored parties
are men. If judges deviate from the statute, then the losing men still enjoy
other opportunities to own real estate. However, if judges adhere, then the
losing women enjoy no such opportunities, on account of the statute. No
one can reasonably reject criteria that classify the legally required result as
suboptimal. Therefore, no one can reasonably reject a set of adjudication
rules containing permissive rule and these optimality criteria. However,
women can reasonably reject the package of restrictive rule plus the sexist
statute. More generally, legally disfavored parties in suboptimal-result cases
can reasonably reject any set of adherence rules that includes restrictive
rule. A superior set of rules is always available, consisting of permissive rule
and optimality criteria that cannot be reasonably rejected. If contractual-
ism condemns any adherence rules, it appears to condemn restrictive rule.
Thus, I am still seeking all-things-considered reasons to obey restrictive
rule. I have found that these would have to be reasons that one could offer
to a judge who has at least as much confidence in his own moral judgment
as in that of the law. What could we say to persuade him to obey restric-
tive rule, without asking him to doubt his own judgment? In the rest of
this chapter and in the next two, I shall discuss arguments for obeying
restrictive rule that do not rely on the dubious premise that the judgment
of a judge who has access to both the law and the facts of the case is usu-
ally inferior to that of the law by itself.

8.5 LEGITIMACY, AUTONOMY, AND RESPECT

I have agreed that legitimacy arguments support the conclusion that law-
makers have pro tanto reasons to promulgate restrictive rule.53 Someone

53. See §7.10.

138 Limits of Legality: Part I

might claim that such arguments also support the conclusion that judges
have moral reasons to obey restrictive rule, even when they are confident
that the legally required result is suboptimal. I shall explain why I dis-
agree.
The first legitimacy argument appeals to respect:
1. Judges have a pro tanto reason to express respect for lawmakers
and citizens.
2. A judge who obeys restrictive rule expresses respect for lawmakers
and citizens.
3. Therefore, judges have a pro tanto reason to obey restrictive rule.
This argument represents the idea that reaching the morally correct
decision is not the only goal that matters. It is also important that the
right official make the decision. There are, after all, many situations in
which it is important that the correct person make the decision, not just
that the decision be correct.54 The same zirconium earrings mean more
when chosen by a lover than when received as a “gift” for opening a bank
account or when chosen by the lover’s office assistant. Part of the mean-
ing of a gift is that it was chosen by the giver. Someone who intercedes
and replaces the chosen gift with another gift expresses disrespect for the
giver, even if the recipient actually prefers the replacement gift to the
original. Analogously, when a judge adheres to the law, he expresses
respect for lawmakers and citizens, even if the law requires a suboptimal
result.
Striking a similar note, Philip Soper argues that deference to the views
of others fosters a community of shared values.55 Soper devotes only a few
pages to the judicial obligation to apply the law,56 but the argument
applies to judges as well as citizens. Judges can, perhaps, express defer-
ence and foster communities of shared values by adhering in suboptimal-
result cases. To this extent, they might have a reason to defer to rules that
were made in a good faith attempt to promote the common good. Per-
haps obeying restrictive rule expresses respect for lawmakers and citizens,
as the second premise states.
I suggest, however, that adherence in suboptimal-result cases is only
one of many ways in which judges can express respect for others, and it is
not an especially desirable way. In the expressive realm, appearances mat-
ter. Judges can express respect by pretending to defer and by adopting
deferential rhetoric without actually deferring.
Moreover, even if the second premise is true, the conclusion of the
respect argument does not follow. Adhering in a suboptimal-result case
expresses disrespect for the losing party at the least. In fact, adhering does

54. Schauer, Playing by the Rules, p. 159.

55. Philip Soper, The Ethics of Deference: Learning from Law’s Morals (Cambridge: Cam-
bridge University Press, 2002), p. 161. Soper does not ultimately appeal to this argument.
56. Ibid., pp. 89–99.
Obeying Adherence Rules 139

something worse to him: it infringes his moral rights, depriving him of

liberty or property. It is never permissible to infringe someone’s moral
rights in order to avoid expressing disrespect for someone else. This prin-
ciple may seem unfamiliar because ordinary people so rarely face
dilemmas in which failure to infringe someone’s moral rights will express
disrespect for someone else. I do not believe that I have ever been in such
a situation. That judges often face such dilemmas reflects the very moral
uniqueness of their job that motivates this book.
Imagine a mythical state in which judges are required by law to pay
periodic visits to legislators. The tradition of this state dictates that on
these visits the judge should purchase and present to the legislator an
expensive plaque that reads, “You make the law. I apply it. Respectfully,
Judge _____.” It is considered highly disrespectful for a judge to arrive at
a legislator’s office without plaque in hand.
Judge Jane is preparing to visit a legislator. Unfortunately, she has
no money, having recently made some bad investments. She needs $1,000
to buy an acceptable plaque. What to do? She notices an impoverished
bystander, net worth $10,000, who happens to be carrying $1,000 in
cash. He sets it down momentarily, where Jane could easily take it.
Assume that taking the cash would otherwise be impermissible, all
things considered. Given that assumption, the fact that Jane needs the
cash in order to buy the plaque and show respect for the legislator does
not constitute a reason for her to take it. It might be permissible to take
the cash in order to protect life, limb, or liberty, but the need to avoid
disrespect provides not even a pro tanto reason to take it.
Judges deciding suboptimal-result cases in the real world face an anal-
ogous dilemma. Deviating disrespects lawmakers and citizens, but adher-
ing violates the losing party’s moral rights. I conclude that the need to
avoid disrespecting lawmakers provides no reason for judges to adhere.
Notice that I could have required Jane to choose between disrespect-
ing the legislator and using physical force against the bystander—killing,
maiming, assaulting, or physically confining him. I even left the bystander
with $9,000, rather than having Jane reduce his net worth to zero. I can
make my point without resorting to such extremes. Judges cannot invoke
“respect for lawmakers” to justify adhering to a law that requires unjustly
infringing someone’s rights, even if it is a matter of only $1,000. A fortiori,
such reasoning cannot support unjustly executing or incarcerating defen-
dants under law.
The second legitimacy argument for obedience to restrictive rule
appeals to autonomy:

1. Judges have a pro tanto reason to facilitate the exercise of auton-

omy by lawmakers and citizens.
2. When a judge obeys restrictive rule, she facilitates the exercise of
autonomy by lawmakers and citizens.
3. Therefore, judges have a pro tanto reason to obey restrictive rule.
140 Limits of Legality: Part I

This argument is no stronger than the respect argument. We can revise

the story of Judge Jane. Tradition now dictates that, rather than purchas-
ing a plaque, the judge must donate $1,000 to the legislator’s favorite
charity or one chosen by the legislator’s constituents. Jane again finds her-
self penniless on her way to visit the legislator. If she arrives without
$1,000, then she fails to facilitate the efforts of the legislator and her con-
stituents to advance their autonomously chosen ends. Again, this need
does not provide Jane with a pro tanto reason to take $1,000 from the
poor bystander.
Someone might object that individuals who do not wish to suffer the
disadvantages of lawbreaking should simply obey the law, just or unjust.
But this objection cannot bolster legitimacy arguments. After all, a law-
maker can always resign from office if he does not wish to be disrespected
by judges who deviate in suboptimal-result cases. And a lawmaker who
wants to avoid interference from deviating judges can find other ways to
exercise his autonomy and promote his personal objectives.
It is also worth noting that judges who reject restrictive rule in favor of
permissive rule still adhere to the law in all permissible-result cases. Every
time they do so they express respect for lawmakers and citizens and facil-
itate the exercise of autonomy by lawmakers and citizens. This is so even
in cases in which the judge would have reached the legally required result,
had the law not required it, as when the judge punishes a defendant for
committing a crime malum in se. Consider a judge who would sentence a
certain rapist to five years in prison, if the law permitted but did not
require him to do so. In fact, the law requires him to do so. Even if this
judge rejects restrictive rule, he can treat this legal requirement as a rea-
son to impose the five-year sentence. He can sentence the rapist to five
years because the law requires him to do so. His decision in such a case is
rationally overdetermined. His decision constitutes deference to prior
lawmakers despite the fact that he would have made the same decision
absent any legal mandate. I can give you flowers because it is your birthday
even if I bought them for you before I knew about your birthday.

8.6 SEPARATION OF POWERS

Someone might object that a judge who deviates even once is “doing the
lawmaker’s job,” and that doing so is always wrong, all things considered.
I reject this argument. First, defining the job of the judge is precisely what
is in dispute in this book. Second, I do not believe that a deviating judge
necessarily even attempts to change the law. He simply fails, albeit know-
ingly, to apply the law. Judges can even make their intentions explicit by
announcing that they do not intend to apply the law.57

57. See Justice Kline’s dissent quoted in §3.12.

Obeying Adherence Rules 141

It is easy to overlook the possibility of deviating without changing the

law. Judges in Anglo-American systems play both law-applying and law-
making roles.58 When a judge announces and applies a new legal rule or
other standard that conflicts with preexisting law, she necessarily departs
from preexisting law and plays a lawmaking role. Such decisions do not
constitute deviation unless the judge lacks the legal authority to make law
in this particular area. In such cases the judge does, indeed, both deviate
and usurp lawmaking authority. She asserts and exercises an authority
that she does not legally possess. In ordinary cases of deviation, by con-
trast, the judge makes no new law.
Even if we concede that deviation usurps some lawmaking authority,
this fact does not necessarily provide a pro tanto reason to adhere. Law-
making authority is a shared, divisible resource. Deviating in a single case
usurps relatively little lawmaking authority. I can imagine an unrealistic
scenario in which a single deviant decision by any given judge would have
the effect of usurping all lawmaking authority in the system, leaving none
for anyone else. In such a scenario each judge would, indeed, have moral
reasons to obey restrictive rule. But in real legal systems a single deviant
decision virtually never has such an effect.
I can, however, imagine a judge deviating in a way that really would
constitute total usurpation. Imagine a judge making the following
announcement:
Hereafter, I shall treat the text of statutes and the opinions of superior courts
much as I treat legal scholarship. These texts may contain good arguments and
good legal standards to apply. I shall adopt these when I agree with them. But
the fact that a proposition appears in a statute or in the ratio decidendi of a case,
as opposed to appearing in a law review or a newspaper editorial, shall have no
effect whatsoever on my deliberations. Moreover, I hereby assert jurisdiction
over all cases.
This judge is declaring himself to be, effectively, both the legislative and
judicial branches of government. There is a world of difference between
this judge and one who, having rejected restrictive rule, deviates in subop-
timal-result cases.59
In this chapter I examined several arguments for the conclusion that
judges have pro tanto reasons to obey adherence rules. I concluded that
judges have strong reasons to obey permissive rule, but I raised objections
to the arguments for obeying restrictive rule. Next I shall consider two
other important arguments for obeying restrictive rule: an argument from
the judicial oath and an argument from political obligation. Each of these
arguments requires its own chapter.

58. This is a thesis of Douglas E. Edlin, Judges and Unjust Laws: Common Law Consti-
tutionalism and the Foundations of Judicial Review (Ann Arbor: University of Michigan Press,
2008).
59. It is, of course, unlikely that other officials will follow this judge’s decrees. At some
point, he will almost certainly be removed from office.
9

The Judicial Oath

Judges swear an oath to perform their official duties. These duties surely
include obeying adherence rules.1 Therefore, I must consider the content
and moral significance of oaths. An oath traditionally involves a solemn
appeal to a deity or to some revered person or thing, to witness one’s
determination to act: to speak the truth, to keep a promise, et cetera.2
Federal law, for example, states the following:

Each justice or judge of the United States shall take the following oath or
affirmation before performing the duties of his office: “I, ______, do solemnly
swear (or affirm) that I will administer justice without respect to persons, and
do equal right to the poor and to the rich, and that I will faithfully and impar-
tially discharge and perform all the duties incumbent upon me as _______
under the Constitution and laws of the United States. So help me God.”3

All judges, state and federal, swear oaths of office. The wording varies at
the state level,4 but the federal oath is representative. How, if at all, does
swearing an oath affect one’s reasons for action after one becomes a judge?
Does the oath-taker acquire moral obligations that others lack?

1. “[A]n official who seeks or consents to serve in a position of public trust has made a
commitment, a voluntary undertaking, to follow the law. . . .” David Lyons, “Derivability,
Defensibility, and the Justification of Judicial Decisions,” in Moral Aspects of Legal Theory
(Cambridge: Cambridge University Press, 1993), p. 137.
2. In our age of religious pluralism it can be a formally affirmed statement or promise
that is accepted as an equivalent to divine appeal.
3. 28 U.S.C. § 453.
4. Some typical oaths of judicial office:

I, ______, do solemnly swear that I will support the Constitution of this State and the
Constitution of the United States, and will perform the duties of my office, faithfully,
impartially and justly, to the best of my ability. So help me God. N.J. Stat. § 41:2A-6
(2009).

I, the undersigned, who have been elected (or appointed) to the office of. . . ., but have not
yet entered upon the duties thereof, do solemnly swear that I will support the constitution
of the United States and the constitution of the state of Wisconsin; that I will administer
justice without respect to persons and will faithfully and impartially discharge the duties
of said office to the best of my ability. So help me God. Wis. Stat. § 757.02 (2008)

142
The Judicial Oath 143

Unfortunately, there is virtually no philosophical literature on oaths of

office.5 There is, however, a vast literature on promissory obligation that
is pertinent because judicial oaths incorporate promises. Oaths differ
from ordinary promises in several respects that I shall discuss shortly.
However, none of these differences diminishes the strength of the prom-
issory obligations generated by oaths. If anything, they strengthen these
obligations. So I can begin by stating a simplified version of the oath
argument:
Oath Argument 1
1. If an agent promises to Φ, then he has a pro tanto moral duty to Φ.
2. By swearing the oath of office, judges promise to fulfill the
duties of the judicial office.
3. Therefore, judges have a pro tanto moral duty to fulfill the duties
of the judicial office.
Oath arguments raise the following questions, which I shall answer in
order. To whom does a judge make promises when he swears his oath?
What does he promise to do? What reasons for action does he acquire
when he makes these promises?

9.1 PROMISES

When a judge swears his oath, he promises that he will fulfill his judicial
duties in every case that he decides. Every federal judge swears, “I will
faithfully and impartially discharge and perform all the duties incum-
bent upon me as a [federal judge] under the Constitution and laws of
the United States.” However the oath is worded, judges promise to
perform the “duties incumbent” upon them. I conclude that a judge
who swears the oath of office has promised to obey the adjudication
rules that specify those duties. Which duties these are we have yet to
determine.

I do solemnly swear (affirm) that I will support the Constitution of the United States,
and the Constitution of the State of Illinois, and that I will faithfully discharge the duties
of the office of. . . . to the best of my ability. Illinois Const., Art. XIII, § 3 (2009)

The Arizona Constitution states that “Each justice, judge and justice of the peace shall,
before entering upon the duties of his office, take and subscribe an oath that he will support
the Constitution of the United States and the Constitution of the State of Arizona, and that
he will faithfully and impartially discharge the duties of his office to the best of his ability.”
A.R.S. Const. Art. VI, § 26 (2008).
5. But see Daniel P. Sulmasy, “What Is an Oath and Why Should a Physician Swear
One?” Theoretical Medicine and Bioethics 20 (1999): 329–46. T. M. Scanlon discusses oaths
in What We Owe to Each Other (Cambridge: Belknap, 1998), pp. 323–26.
144 Limits of Legality: Part I

Swearing the oath gives the judge promissory duties to the general
public, to his fellow judges, and to other public officials. As I shall discuss
in the next section, he promises everyone that he will endeavor to do his
job well for as long as he holds judicial office. However, his promise is not
addressed to any particular promisee. It does not yet give him special
duties to anyone as opposed to general duties to everyone in his legal
system. I suggest that we understand him as making a conditional promise
that enables his subsequent actions to generate special duties. When a
case is assigned to him, he always has the option of requesting that it be
reassigned to another judge. He can even resign from the bench before
hearing the case. If he chooses to hear the case, then he voluntarily fulfills
the condition of his promise with respect to the parties, thereby complet-
ing an unconditional promise to them that he will fulfill his judicial duties
in their case. Only at that point does he acquire special promissory duties
to them.

9.2 FIDELITY

It is widely agreed that individuals have a natural duty of fidelity or

promise keeping.6 Promises give promisors pro tanto moral reasons to
perform. The judicial oath incorporates a promise to obey adjudication
rules, so it gives judges pro tanto moral reasons to obey these rules. Prom-
ises are subject to familiar defeating conditions at formation, but these
conditions are absent when judges swear oaths of office. There is no coer-
cion or duress, for example. Judges always have reasonable alternative
occupations.
Which adherence rule or rules does the oath obligate judges to obey?
The easiest case is made for permissive rule. If the oath means anything,
then it includes a promise to adhere to the law in required-result and
optimal-but-not-required-result cases. So judges have a reason to adhere
in such cases. They also have an exclusionary reason that excludes all rea-
sons to deviate in required-result and optimal-but-not-required-result
cases. These are private reasons because there are, by definition, no impar-
tial reasons to deviate in required-result or optimal-but-not-required-
result cases. Therefore, the oath gives permissive rule practical authority
for judges. To obey permissive rule is to accept a reason to adhere in per-
missible-result cases and an exclusionary reason that excludes private rea-
sons to deviate in all cases. I endorse the following:

6. See, e.g., Bernard Gert, Morality: Its Nature and Justification (New York: Oxford Uni-
versity Press, 1998), pp. 188–90. Some writers deny that promises as such normally generate
moral obligations, but even they agree that promisors have moral obligations to do as they
have promised. These writers simply deny that the obligation derives from the fact that the
promisor makes the promise. See P. S. Atiyah, Promises, Morals, and Law (Oxford: Clarendon
Press, 1981), pp. 123–29.
The Judicial Oath 145

Oath Argument 2
1. If an agent promises to Φ, then he has a pro tanto moral duty to Φ.
2. Judges promise to obey permissive rule.
3. Therefore, judges have a pro tanto moral duty to obey permis-
sive rule.
Matters are more complicated with respect to restrictive rule. Restric-
tive rule gives judges reasons to adhere in certain suboptimal-result
cases, as well. It also excludes certain moral reasons, not just private rea-
sons, to deviate. Does the oath give restrictive rule practical authority for
judges?
First, we must examine more closely what judges promise to do when
they take the oath. Do they promise to adhere in suboptimal-result cases?
The language of the oath does not refer to such cases explicitly. I am not
sure how much weight to place on this omission, but it could be signifi-
cant. After all, the oath could easily have included a clause stating, “I will
apply the law faithfully, no matter how unjust I may consider its demands
to be.” The oath contains no such language. In fact, federal judges swear
that they “will administer justice without respect to persons,” (emphasis
added) although I do not think this language can be honestly read as sup-
porting the thesis that judges are morally permitted to deviate in order to
avoid unjust results.7
I have located no code of judicial conduct that so much as acknowl-
edges the existence of suboptimal-result cases. The codes do not mention
cases in which the law requires results that are unjust, bad, immoral, ineq-
uitable, or wrong. So there is no explicit textual support for the proposi-
tion that the judicial oath includes a promise to adhere in suboptimal-result
cases. It is not unreasonable, however, to read such a promise into the
judicial oath. After all, textual adjudication rules do not explicitly exclude
suboptimal-result cases from the scope of the oath, either. The text is
silent, but in the absence of an explicit textual exception it is reasonable
to infer that the general promise to adhere to the law extends to subopti-
mal-result cases. Let us assume arguendo that the oath incorporates by
implication a promise to adhere in suboptimal-result cases, despite the
absence of explicit language in either the oath or the codes. If this as-
sumption is mistaken, then so much the better for my main argument.
It is not surprising that the oath and other textual adjudication rules
lack explicit language promising to adhere in suboptimal-result cases. I
think such language would be otiose because such a promise would not,
in fact, create moral reasons to adhere in suboptimal-result cases. I believe
that the oath would not generate a moral reason to adhere in suboptimal-
result cases, even if it contained an explicit promise to do so. This is to say

7. I do not read this language as contemplating cases of conflict between justice and
positive law. Rather, the phrase expresses a commitment to impartial decision. “Administer
justice” in this context is merely a sonorous way of saying, “render decisions under law.”
146 Limits of Legality: Part I

that the oath does not give any nonpermissive rule practical authority for
judges. The next sections explain my position.

9.3 IMMORAL PROMISES

My argument rests on two complementary premises. First, promises do

not attenuate one’s other moral reasons. Second, other moral reasons can
undermine or override promissory obligations. These principles are widely
recognized.8 The first is reflected in the well-established legal principle
that contracts to commit crimes or torts are unenforceable.9 Others have
noted that an agent cannot acquire a moral obligation to follow a rule by
consenting, committing himself, or promising to follow it, if the rule
requires an action that is morally wrong.10 I shall now describe some
hypothetical scenarios in which promisors promise to perform immoral
actions. Common moral convictions hold that the immorality of these
actions undermines promissory obligation. Then I shall examine how
popular theories of promising support these principles.
Bruce gets on the radio, identifies himself, and swears a solemn oath
before his fellow citizens that he will kick a squirrel in his yard on Satur-
day.11 Taking this oath gives Bruce no reason whatsoever to kick a squirrel
come Saturday. Nor does it attenuate his reasons to refrain from doing so.
His natural duty of nonmaleficence forbids inflicting gratuitous pain on
animals.12 This duty undermines whatever reasons his oath might other-
wise have given him.13 Taking an oath does not give one a reason that
overrides or attenuates one’s natural duty of nonmaleficence.14
Juanita swears a public oath that she will not call the police if she
ever happens to witness a kidnapping in progress. Taking this oath gives
Juanita no reason whatsoever to refrain from calling the police in such an

8. Joseph Raz, The Morality of Freedom (Oxford: Clarendon Press, 1986), p. 173: “one’s
right to promise does not include the right to promise to perform immoral acts.” See also
Heidi M. Hurd, “Justifiably Punishing the Justified,” Michigan Law Review 90 (1992): 2203–
324, p. 2242; Joseph Raz, “Promises in Morality and Law,” Harvard Law Review 95 (1982):
916–38, p. 926; Kurt Baier, “The Justification of Governmental Authority,” Journal of Philos-
ophy 69 (1972): 700–16, p. 712.
9. Restatement First of Contracts §§ 598–609 (1932).
10. Larry Alexander and Emily Sherwin, The Rule of Rules (Durham, N.C.: Duke Uni-
versity Press, 2001), p. 75.
11. Actually, kicking a squirrel is not as easy as Bruce thinks it is.
12. If you find squirrel kicking too trivial to trigger the duty of nonmaleficence, then
substitute an act of violence that does, in your opinion, trigger that duty.
13. In 2005, two college students started a website, SaveToby.com, on which they
promised to kill their adorable pet rabbit, Toby, unless they received donations totaling
$50,000 before June 30. Stephen E. Sachs, “Saving Toby: Extortion, Blackmail, and the Right
to Destroy,” Yale Law and Policy Review 24 (2006): 251–61.
14. It is probably wrong for Bruce to swear such an oath in the first place, but that is not
important here.
The Judicial Oath 147

event. Natural samaritan duties require taking such steps to prevent kid-
nappings. Juanita’s oath does not override or undermine her samaritan
duties. Nor does it seem to me, intuitively, that Juanita can alienate her
natural samaritan rights by taking an oath.
A point about the function of promises emerges from these hypothet-
icals: the function of a promise, including the promise imbedded in an
oath, is to create a special obligation in the promisor. The reply, “Because
I promised to do it,” answers the question, “Why should I do that, rather
than something else that I would now prefer to do?” It does not answer
the question, “Why am I permitted to do that, which would otherwise be
wrong?” Promises make morally optional actions morally obligatory. They
do not make morally impermissible actions morally permissible, much
less obligatory.
The preceding arguments entail that the first premise in Oath Argu-
ment 2 needs a qualification. Corrected, the argument reads as follows:
Oath Argument 3
1. If an agent promises to Φ and has no otherwise undefeated pro tanto
moral duty not to Φ, then he has a pro tanto moral duty to Φ.
2. Judges promise to obey a permissive rule.
3. Judges have no otherwise undefeated pro tanto moral duty not to obey
a permissive rule.
4. Therefore, judges have a pro tanto moral duty to obey a permissive rule.

However, we have also learned that substituting restrictive rule for per-
missive rule falsifies the third premise. Restrictive rule requires judges to
adhere in suboptimal-result cases, which they have an undefeated pro
tanto moral duty not to do. Therefore, oath arguments do not support a
pro tanto moral duty to obey restrictive rule.

9.4 THEORIES OF PROMISING

My preceding argument rests upon the premise that a promise to Φ gives

the promisor no reason to Φ if he has an otherwise undefeated, pro tanto
moral duty not to Φ. I have defended this premise by appeal to hypothet-
icals, but it finds additional support in philosophical theories of promis-
ing. Each of these theories supports the principle that promises to perform
otherwise immoral actions are void.

9.4.1 Deflationary and Act-Consequentialist Theories

According to a deflationary theory, promises add nothing to a promisor’s
reasons.15 If this thesis applies to the incorporated promise in the judicial

15. See, e.g., Elinor Mason, “We Make No Promises,” Philosophical Studies 123 (2005):
33–46.
148 Limits of Legality: Part I

oath, then a deflationary theory entails that the oath adds nothing to the
judge’s reasons to adhere. Such theories obviously cannot help defend the
conclusion that judges should obey restrictive rule.
Act-consequentialist theories of promissory obligation offer little more.
According to these theories, the fact that a particular practice of oath keep-
ing promotes good consequences gives judges a pro tanto reason to keep
their oaths. However, deviant decisions in suboptimal-result cases often
promote better consequences than adherent decisions would, even
accounting for damage done to the practice of judicial oath taking, the
rule of law, et cetera. So act-consequentialist theories cannot help us
defend the conclusion that judges can have all-things-considered reasons
to obey restrictive rule.

9.4.2 Rule Consequentialism

Rule consequentialists hold that promisors have a duty to keep promises
because a widespread practice of promise breaking has suboptimal conse-
quences. However, rule consequentialists permit promise breaking under
certain conditions. They hold that promisors may break a promise if and
only if it would promote the good for people generally to internalize a
rule permitting the breaking of promises in such situations.
Consider a rule that permits promisors to break promises to Φ if Φ-ing
is immoral. For a rule consequentialist, Φ-ing is immoral if and only if it
would promote the good for people to internalize a rule that forbids
Φ-ing. So the question is, would it promote the good for people to inter-
nalize a rule that requires promisors to keep promises to Φ, if it would
promote the good for people to internalize a rule that forbids Φ-ing?
If it promotes the good for people to internalize a rule that forbids
Φ-ing, then Φ-ing must, in the aggregate, diminish the good. Now I need
not deny that the fact that someone has promised to Φ can make the
promisor’s act of Φ-ing somewhat less good-diminishing. But Φ-ing is still
good-diminishing, overall. At least, the combination of promising to Φ
and then proceeding to Φ does not promote the good if Φ-ing is otherwise
immoral. The world would contain more good if one simply did not
promise to Φ in the first place. If this conclusion were false, then a bad
agent could simply promise someone that he would Φ, proceed to Φ, and
correctly claim thereby that he was promoting the good, no matter how
immoral Φ-ing might be.
Therefore, it would not promote the good for people to internalize a
rule that requires promisors to keep promises to Φ even if Φ-ing is other-
wise immoral. Rule consequentialism supports the principle that if Φ-ing
is immoral, then a promise to Φ gives the promisor no reason to Φ.
At this point, a defender of restrictive rule might interject. Suppose
people internalize a rule that allows promisors to break promises to per-
form otherwise immoral actions. This rule does not promote the good if
people too often misidentify immoral actions. Such people will break the
The Judicial Oath 149

wrong promises and keep the wrong ones, despite sincerely trying to obey
the rule. A rule that requires keeping all promises, moral or immoral,
might lead to fewer broken promises that should be kept than would a
rule that permits breaking immoral promises. The exceptionless rule, if
internalized, might better promote the good.
This objection exaggerates moral incompetence. The basic fidelity
principle holds that a promise to Φ gives the promisor a pro tanto moral
reason to Φ. The relevant population comprises individuals who are capa-
ble of internalizing this principle. By definition, these people recognize
the immorality of promise breaking and act upon this moral knowledge.
There is no reason to doubt their ability and motivation to identify other
immoral actions and to use this knowledge pursuant to the principle that
if Φ-ing is otherwise immoral, then a promise to Φ gives the promisor no
pro tanto moral reason to Φ.

9.4.3 Free Riding

According to free-rider theories, if breaking a promise involves free riding
on a just practice, as it often does, then the promisor has a pro tanto reason
to keep the promise.16 What do free-rider theories imply about immoral
promises? If a promisor does not benefit from breaking an immoral prom-
ise, then he does not ride free on any practice. But what about immoral
promises that the promisor benefits by breaking? In such cases it matters
why the promise is broken. Breaking an immoral promise is still objec-
tively permissible, although it may be blameworthy if the promisor breaks
it for a private reason. Lazy Assassin promises to assassinate a senator but
breaks his promise in order to watch television. Breaking this promise is
objectively permissible but blameworthy. Although he does the objec-
tively obligatory thing, he rides free on the general practice of not break-
ing promises for nonmoral reasons. This is a just practice from which he
benefits.17
Another question is whether the practice upon which the promisor
rides free is a just practice. To answer this question, we must identify the
practice or practices upon which a promisor rides free when he breaks an
immoral promise. Let us assume that the promisor (unlike Lazy Assassin)
breaks the promise because keeping it would be immoral. There are two
possibilities. First, we might conclude that he rides free on the general
practice of promise keeping. That practice is just, and most promisors

16. John Rawls, A Theory of Justice (Cambridge, Mass.: Harvard University Press, 1971),
pp. 344–48; H. A. Prichard, “The Obligation to Keep a Promise,” in Moral Obligation (Oxford:
Clarendon Press, 1949).
17. One might also argue that breaking a promise is blameworthy if the promisor has
confidence level c that she is breaking it for a moral reason and there is a general practice,
from which she benefits, of keeping promises unless one has at least confidence level d that
one is breaking one’s promise for a moral reason, and c < d.
150 Limits of Legality: Part I

benefit from it. If breaking an immoral promise constitutes free riding on

the general practice, then promisors have a pro tanto reason to keep such
promises.
Alternatively, we might conclude that he rides free on the practice of
keeping immoral promises. Is there such a practice? The fact that many
writers have treated as common knowledge the principle that immoral
promises do not obligate suggests that there is no such practice. However,
because I am not sure how many immoral promises are kept in modern
societies, I shall assume arguendo that such a practice exists. The next
question is whether the promisor benefits from the practice. Most prom-
isors do not benefit from a general practice of keeping immoral promises.
They benefit from the practice of keeping morally permissible promises.
When one breaks an immoral promise, one violates the first practice, not
the second. Furthermore, the practice of keeping immoral promises is not
just. So breaking an immoral promise does not constitute free riding at all,
much less free riding on a just practice. If breaking an immoral promise
does not constitute free riding on the general practice of promise keeping,
then promisors have no pro tanto reason to keep immoral promises,
according to free-rider theories.

9.4.4 Nonpractice Theories and Reliance

T. M. Scanlon and others object that practice theories, including free-rider
theories, fail to capture the sense in which breaking a promise wrongs the
promisee—a particular individual—as opposed to society at large.18 This
objection might seem inapplicable to the judicial oath, which really is
made to society at large. Actually, the objection still applies because the
judge wrongs a particular litigant if she (the judge) violates her judicial
duties to his detriment. When the judge swears her oath she makes a con-
ditional promise to all parties who may come before her, promising each
of them that she will fulfill her judicial duties in his case. When she devi-
ates she breaks this promise to the litigant whose legal rights she thereby
disregards. Practice theories fail to capture the way in which deviant
decisions putatively wrong defeated litigants.
Contemporary theories of promising such as Scanlon’s are designed to
reflect the idea that breaking a promise wrongs the promisee. They base
promissory obligation on promisees’ expectations and reliance. In the
process of discussing Scanlon’s theory I shall also consider the principle
that agents have moral reasons not to frustrate expectations that they
have led others to form. This principle, if true, could provide a nonpromis-
sory basis for the claim that judges have reasons to obey adherence rules.
Scanlon argues that promises generate reasons independently of any
social practice. He defends

18. Scanlon, What We Owe to Each Other, ch. 7; Neil MacCormick, “Voluntary Obliga-
tions and Normative Powers I,” Proceedings of the Aristotelian Society 46 (1972): 59–78.
The Judicial Oath 151

Principle D: One must exercise due care not to lead others to form reason-
able but false expectations about what one will do when one has good rea-
son to believe that they would suffer significant loss as a result of relying on
these expectations.19
and
Principle F: If (1) A voluntarily and intentionally leads B to expect that A will
do X (unless B consents to A’s not doing so); (2) A knows that B wants to be
assured of this; (3) A acts with the aim of providing this assurance, and has
good reason to believe that he or she has done so; (4) B knows that A has the
beliefs and intentions just described; (5) A intends for B to know this, and
knows that B does know it; and (6) B knows that A has this knowledge and
intent; then, in the absence of special justification, A must do X unless B con-
sents to X’s not being done.20

According to principles D and F, a promise to Φ gives the promisor a

reason to Φ only if the promisor has led the promisee to form a reasonable
expectation that the promisor will Φ. We must determine the conditions
under which an expectation is reasonable. We can distinguish epistemically
reasonable from normatively reasonable.21 Relying on an expectation that
an event will occur is epistemically unreasonable for someone in a speci-
fied epistemic state if a person in that state possessing theoretical rational-
ity would not predict the event. By contrast, relying on an expectation
that someone will Φ is normatively unreasonable if Φ-ing is otherwise
immoral.
Suppose that every year for the past five years, Leo’s father has hired a
new assistant. Leo has tried to seduce each of these assistants. Each woman
has rejected Leo’s advances, whereupon Leo has asked his father to fire
her and his father has complied. Leo’s father has hired a new office
assistant this year. She rejects Leo’s advances. By firing the assistants in
previous years, Leo’s father has led Leo to expect that this assistant will
be fired. Leo has an epistemically reasonable expectation that his father
will fire her. However, it is not normatively reasonable for Leo to expect
the assistant to be fired (sexual harassment is wrong). Even if Leo relies to
his detriment on his expectation, these facts give his father no reason
whatsoever to fire the assistant.
My first conclusion follows. Suppose one has led another to form an
epistemically reasonable expectation that one will Φ, inducing detrimen-
tal reliance. These facts do not constitute a reason to Φ if Φ-ing is im-
moral. This principle holds whether or not the relying party reasonably
believes Φ-ing to be immoral.
What might Scanlon’s theory say about promises to act immorally?
The “special justification” clause in principle F provides a hint that Scanlon

19. Scanlon, What We Owe to Each Other, p. 300.

20. Ibid., p. 304.
21. A similar distinction is drawn in David Lefkowitz, “A Contractualist Defense of
Democratic Authority,” Ratio Juris 18 (2005): 346–64.
152 Limits of Legality: Part I

does not believe that immoral promises generate all-things-considered

obligations. But is his theory consistent with the proposition that immoral
promises create pro tanto obligations? I suggest that a promise to Φ gives
the promisor a reason to Φ only if the promisor has led the promisee to
form an expectation that the promisor will Φ and the expectation is both
epistemically and normatively reasonable.
Here is a case in which a promisor has no reason to keep her promise
because the promisee’s reliance is normatively unreasonable. Rachel
promises to install a computer in her grandmother’s bedroom. Her grand-
mother relies upon this promise, planning everything around it. It is sub-
jectively epistemically reasonable for her to do so. However, Rachel
overhears her grandmother’s live-in boyfriend talking about how he plans
to use the computer to plan a terrorist attack. Rachel’s grandmother
knows nothing about these plans. Neither Rachel’s promise nor her grand-
mother’s detrimental reliance gives Rachel any reason whatsoever to
install the computer.22
If someone promises to Φ, then it may be epistemically reasonable for
the promisee to expect him to Φ even if Φ-ing is immoral. However, if
Φ-ing is immoral, then it is not normatively reasonable for the promisee
to expect the promisor to Φ. Therefore, on Scanlon’s theory, a promise to
Φ does not give the promisor a pro tanto reason to Φ if Φ-ing is immoral.
The reasoning applies whether or not the promisee reasonably believes
that Φ-ing is immoral.
I have discussed two related principles: (1) that one has no reason to Φ
if Φ-ing is immoral, even if one has led someone else to form an epistemi-
cally reasonable expectation that one will Φ and that person has relied to
his detriment; and (2) that a promisor has no reason to keep a promise to
Φ if Φ-ing is immoral. These principles are widely accepted and find sup-
port in current theories of justified reliance and promissory obligation.
These principles have implications for my discussion. I have asserted
that relying upon a prediction that someone will Φ is objectively unrea-
sonable if Φ-ing is otherwise immoral. The act of Φ-ing counts as “other-
wise immoral” if and only if Φ-ing would be immoral if no one had relied
upon it. I argued in chapter 6 that a decision to adhere in a suboptimal-
result case is otherwise immoral in this sense. A judge who adheres to the
law in a suboptimal-result case performs actions that would be immoral,
all things considered, if the law did not require them. Even if the judge
has led parties to believe that he will adhere in suboptimal-result cases,
and they are epistemically reasonable to expect him to do so, and they
rely to their detriment on their expectations, their expectations give him
no pro tanto reason to adhere. It is objectively normatively unreasonable

22. However, these facts retain some practical significance. Both give Rachel a moral
reason to apologize for disappointing her grandmother. Perhaps she should also offer to
install the computer in her own home so her grandmother can use it when she visits. She
might want to mention the boyfriend’s proclivities, too.
The Judicial Oath 153

to rely on one’s expectation that a judge will adhere in a case if it is a

suboptimal-result case. Similarly, even if we interpret the judicial oath as
incorporating a promise to adhere in suboptimal-result cases, it gives
judges no reason to do so.
I shall illustrate by revisiting the eviction case.23 Suppose Yasmin has
missed three rent payments over the past twelve months. Rafael’s lawyer
tells him that he has a legal right to evict Yasmin. Therefore, it is subjec-
tively epistemically reasonable for Rafael to predict that a judge will grant
his eviction petition and for Rafael to rely upon that expectation. Unfor-
tunately for Rafael, his lawyer misread the code. In Rafael’s jurisdiction
the law does not grant a right to evict until the tenant misses three consec-
utive rent payments. If she misses three nonconsecutive payments, then
the law merely gives the court discretion to grant or deny an eviction
petition, without specifying legal considerations for the judge to take into
account. Granting Rafael’s petition will inflict a substantial undeserved
disadvantage on Yasmin. Denying it will inflict an undeserved disadvan-
tage on Rafael. However, let us stipulate that the magnitude of the wrong
against Yasmin is at least as great as the magnitude of the wrong against
Rafael, even taking into account Rafael’s reasonable reliance. Therefore,
the court has an all-things-considered moral reason to rule for Yasmin.
Suppose Judge Lucas grants Rafael’s petition. He thereby orders the
sheriff to force Yasmin from her apartment. No one, with the possible
exception of Judge Lucas, has an all-things-considered moral reason to
order anyone to do this to Yasmin. Indeed, everyone else has a strong pro
tanto moral reason not to do so. Because the law does not require it, even
Judge Lucas has an all-things-considered reason not to do it. I suggest that
just as Bruce’s oath gives him no reason that competes with his natural
duty to refrain from kicking squirrels, likewise Judge Lucas’s oath gives
him no reason that competes with his natural duty to deny the eviction
petition, notwithstanding Rafael’s expectations—which are reasonable as
a subjective epistemic matter.
Now imagine identical facts, except that the law grants a right to evict
after three nonconsecutive missed payments, just as Rafael was told. The
law requires the judge to grant Rafael’s petition. This is now a subopti-
mal-result case. If the law did not require the judge to grant Rafael’s peti-
tion, then the judge would have an all-things-considered moral reason
against doing so, even taking into account Rafael’s reliance. The judge’s
pro tanto reason to deviate has at least as much weight as the pro tanto
reason to adhere generated by the reliance of parties to the case.
My point is that Rafael’s reliance is no more subjectively reasonable or
detrimental in the second version of the case than it was in the first. The
fact that Rafael has reasonably relied upon the judge to grant his petition
does not give the judge a reason to grant it that is any stronger in the

23. Introduced in §5.3.

154 Limits of Legality: Part I

second version of the case than it was in the first. Therefore, if judges are
ever morally required to adhere in a suboptimal-result case, then they
must have some other reason to do so, in addition to the reasonable reli-
ance of parties.
This is not to deny that a judge who deviates in a suboptimal-result
case may have certain residual reasons. She may have reason to apologize
to parties whose expectations she has disappointed. But the reliance of
parties gives her no reason to adhere.
I should also emphasize that my point is not that the judge has no rea-
son to adhere in Rafael’s case. That would be a much stronger claim that
I shall not make. My point is that neither the judicial oath, as such, nor the
reliance of parties, as such, generates any reason to adhere to the law in
suboptimal-result cases. These facts generate reasons to adhere only in
conjunction with certain enabling conditions that I shall spell out in the
chapters ahead.

9.4.5 Contractual Duties

Some would say that it is the judge’s job to adhere to the law. He is con-
tractually obligated to do it. He is paid to do it. This might seem to give
him a moral reason to adhere. Again, we must be careful not to slide from
the indisputable thesis that judges are morally obligated to adhere in per-
missible-result cases to the stronger conclusion that they are so obligated
in all cases. If an action is otherwise morally impermissible, then one’s job
description cannot give one a moral reason to do it. Nor can forming a
contract, nor can receipt of compensation.

9.5 THE SPECIAL WEIGHT OF OATHS

One might be tempted to object as follows. My argument proceeds from

a premise about ordinary promises—that promises to act immorally are
void. But the judge’s promise to obey adherence rules is no ordinary
promise. It is a promise incorporated into an oath of office. One might
plausibly suggest that such promises have exceptional weight, whereas I
have treated them as though they were ordinary promises.
There are, indeed, several aspects of the promises incorporated into the
judicial oath that would seem to give them more weight than ordinary
promises. First, the judicial oath concerns a very important subject—the
judge’s professional duties—whereas some promises concern trivial sub-
jects. Second, judges swear their oaths in public, whereas promises can be
made to a single promisee in private. Third, with his oath the judge makes
a general commitment to the law as a whole, for an indefinite time period,
whereas promises can concern specific actions to be taken at specific times.
Fourth, when a judge swears her oath, she promises to be a certain kind of
person for the sake of the public, whereas a promise need not involve any
The Judicial Oath 155

such commitment to maintain a certain sort of character. Fifth, the judi-

cial oath appeals to God, whereas ordinary promises make no such appeal.
Finally, the force of the judicial oath may be seen as less likely to diminish
with changed circumstances than in the case of promises.24
All of these seem to be good reasons to agree that a judge’s promises,
undertaken by oath, have greater weight than ordinary promises. How-
ever, this concession does not affect my argument. The moral reasons to
deviate in a suboptimal-result case need not outweigh the reasons to
adhere that are provided by the judge’s promise. Rather, moral reasons to
deviate undermine promissory duties categorically. It is not a matter of
balancing. Recall that Bruce and Juanita do not merely make promises,
they swear oaths.25 It makes no difference.

9.6 OATH AS INUS CONDITION

I have argued that the judicial oath does not give judges a pro tanto moral
reason to obey restrictive rule—or any nonpermissive rule, for that matter.
However, I shall argue in part II that judges have reasons to obey a non-
permissive rule. So in one sense swearing the oath must generate a reason
to obey a nonpermissive rule: in modern legal systems, after one has been
appointed or elected to judicial office, taking the oath of office suffices to
transform one into a judge. However, anyone can take the judicial oath
without becoming a judge, in which case one does not acquire even a
right, much less a duty, to decide cases. Sincerely uttering the words of the
oath does not give one adherence reasons. Taking the oath is not concep-
tually necessary, either, because not all possible legal systems require judg-
es to swear oaths of office.
In modern legal systems, however, taking the oath is a condition for
becoming a judge. It is “an insufficient but nonredundant part of an
unnecessary but sufficient condition”—an INUS condition.26 Taking the
oath is a performative speech act that, given the right background condi-
tions, converts one into a judge. Taking the oath by itself is not sufficient.
You can swear the judicial oath right now without becoming a judge.
Nevertheless, the oath is part of a constellation of conditions that are
jointly sufficient for becoming a judge. Moreover, given that these other
conditions have occurred, rather than a different set of conditions that are
sufficient for becoming a judge, taking the oath is necessary. One cannot
become a judge in the United States today without taking the oath of
office. My point is that the oath by itself does not give one a reason to obey

24. This list was adopted from Sulmasy, “What Is an Oath and Why Should a Physician
Swear One?”
25. See §9.3.
26. J. L. Mackie, “Clauses and Conditions,” American Philosophical Quarterly 2 (1965):
245–64.
156 Limits of Legality: Part I

a nonpermissive rule, even if the oath constitutes a promise to obey such

a rule. The oath gives one a reason to obey a nonpermissive rule only inso-
far as the oath is an INUS condition on becoming a judge. Once one is a
judge, aspects of the judicial role give one reasons to obey a nonpermissive
rule, and these aspects of the role would do so even if one had taken no
oath. The law could instead provide that becoming a judge requires shout-
ing, “Open sesame!”27
It may seem unimportant to distinguish between the claim that the
oath gives the judge a reason to obey a certain adherence rule and the
claim that the judicial role gives him such a reason. In fact, the distinction
is important. Promises give promisors exclusionary reasons. The judicial
oath does not merely outweigh the judge’s private reasons to deviate: it
excludes them. That is what it means to say that the oath gives judges a
reason to obey permissive rule. If the oath also excluded moral reasons to
deviate, which I have denied in this chapter, then judges would have a
reason to obey restrictive rule. If judges’ reasons derive from other aspects
of the judicial role, not from the oath, then it does not follow that they
have any reason to obey restrictive rule. Whether they have such a reason
remains an open question.

27. A detour into science fiction reinforces the point. Dr. Sam Beckett, hero of the
television program Quantum Leap, finds himself suddenly and unexpectedly inhabiting the
bodies of other people, seriatim, without his consent or volition. If Sam is suddenly thrust
into the body of a judge who is about to hear a case, then I think he has moral reasons to
obey the same adjudication rules that “real” judges do, despite having never taken the judi-
cial oath. Of course, he can resign—which he should if he doubts his judicial competence
(although viewers learn that Sam is a super-genius and legally knowledgeable). But if Sam
chooses to hear cases, then he has as strong a moral reason to obey adjudication rules as a
judge who swore the oath would have. Quantum Leap is wholesome, educational entertain-
ment for young people.
10

Legal Duty and Political Obligation

This chapter explores another argument for the conclusion that judges
have moral reasons to obey restrictive rule. The reasoning is that judges
have a legal duty to obey restrictive rule and a moral duty to fulfill their
legal duties:

1. Judges in reasonably just, legitimate states have a pro tanto moral

duty to obey the law.
2. Obeying the law entails fulfilling one’s legal duties.
3. Judges have a legal duty to obey restrictive rule.
4. Therefore, judges have a pro tanto moral duty to obey restrictive
rule.

This argument is valid, but the first and third premises require
discussion.

10.1 POLITICAL OBLIGATION AND JUDICIAL OBLIGATION

Chapter 4 considered arguments for the third premise. I concluded that a

viable case, although not a conclusive one, can be made for it. If the third
premise is false, then my larger argument in the book is easier to make.
A reader who rejects the third premise can skip this chapter. In this
chapter I shall assume arguendo that the third premise is true: judges have
a legal duty to obey restrictive rule. The next question is whether they
have a moral duty to fulfill their legal duties. For convenience, we can ask
whether they have a moral duty to fulfill their judicial duties, provided we
remember that we are interested in a moral duty to fulfill judicial duties
qua legal duties.
There has been extensive debate on the general question whether
individuals have a moral duty to obey the laws of the jurisdiction in
which they reside. The debate during the past forty years has concen-
trated on the question whether we have a duty to obey the law that is
content independent as well as “comprehensively applicable [and] uni-
versally borne,” meaning that it obligates all residents to obey all laws of

157
158 Limits of Legality: Part I

the jurisdiction.1 Many different arguments have been offered in support

of such a duty.2 It is worth pausing to ask what logical relationship holds
between this debate and my question concerning the moral duty to fulfill
judicial duties.
One might assume that judges’ moral duty to fulfill their legal duties
stands or falls with the proposition that legal subjects have a “comprehen-
sively applicable, universally borne, content-independent” duty to obey
the law. After all, judges are legal subjects. If all legal subjects have a moral
duty to obey traffic laws, then judges have a moral duty to obey traffic
laws.
It would, however, be too quick to conclude that a successful defense
of a general duty to obey the law is either necessary or sufficient to sup-
port the premise that judges have a moral duty to fulfill their judicial
duties. Consider sufficiency. Although judges are, indeed, legal subjects,
the arguments for a general duty to obey the law are aimed at conduct
rules, not decision rules. Adjudication rules direct judges to apply decision
rules, not conduct rules. Therefore, the soundness of an argument for a
general duty to obey is not sufficient to support the conclusion that judges
are morally obligated to fulfill their judicial duties.
Next consider necessity. Every prominent argument for a duty to obey
defends a duty that is comprehensively applicable and universally borne.
A moral duty to fulfill judicial duties, by contrast, is neither universally
borne nor comprehensively applicable. Only judges have judicial duties,
and these comprise only some of a judge’s legal duties. Therefore, the fact
that the law requires judges to fulfill their judicial duties could give them
a moral reason to do so even if legal subjects have no duty to obey the law
that is either universally borne or comprehensively applicable.
So the success of a defense of a general duty to obey the law is neither
necessary nor sufficient to support the premise that judges have a moral
duty to fulfill their judicial duties. In the next sections, I consider four
arguments for the principle that judges are morally obligated to fulfill
their legally imposed judicial duties. My arguments do not support

1. For useful overview see William A. Edmundson, “State of the Art: The Duty to Obey
the Law,” Legal Theory 10 (2004): 215–59. The language quoted here is attributed to
Matthew Kramer on p. 215.
2. See, e.g., Mark C. Murphy, “Natural Law, Consent, and Political Obligation,”
Social Philosophy and Policy 18 (2001): 70–92; Christopher Heath Wellman, “Liberalism,
Samaritanism, and Political Obligation,” Philosophy and Public Affairs 25 (1996): 211–37;
David Miller, On Nationality (Oxford: Oxford University Press, 1995); Margaret Gilbert,
“Group Membership and Political Obligation,” The Monist 76 (1993): 119–31; George
Klosko, The Principle of Fairness and Political Obligation (Lanham, Md.: Rowman & Littlefield,
1992); A. D. M. Walker, “Political Obligation and the Argument from Gratitude,” Philosophy
and Public Affairs 17 (1988): 191–211; Ronald Dworkin, Law’s Empire (Cambridge, Mass.:
Harvard University Press, 1986), pp. 191–92; John Rawls, A Theory of Justice (Cambridge,
Mass.: Harvard University Press, 1971), pp. 114–17, 333–55; H. L. A. Hart, “Are There Any
Natural Rights?” Philosophical Review 64 (1955): 175–91.
Legal Duty and Political Obligation 159

a general duty to obey the law, but they adapt widely discussed arguments
for such a general duty: arguments from consent, fair play, natural duty,
and gratitude. If one of these arguments succeeds, then judges have at
least a pro tanto moral duty to fulfill their judicial duties. I must empha-
size that I am not committed to the claim that there are sound arguments
for a general duty to obey the law. Such arguments have been subjected
to powerful criticisms.3
In order to reflect the different reasons for which legal subjects can act
as the law requires, philosophers often draw a distinction between conforming
to the law and obeying the law, also known as complying.4 An agent con-
forms to the law if and only if she acts as the law requires, regardless of her
reasons for doing so. An agent obeys or complies with the law if and only
if she treats the fact that the law requires conformity as a reason to con-
form. To obey the law is to conform for content-independent reasons. The
conform/obey distinction is important because historical legal systems
contain mandatory rules that make illegal much conduct that we have
moral reasons to avoid anyway, independent of the law, such as assault,
destruction of property, defamation of character, et cetera. Individuals
have content-dependent moral reasons to conform to these laws. Individ-
uals who conform to these laws for exclusively moral reasons do not
obey.
By contrast, subjects obey the law if they conform for content-
independent reasons. For example, subjects sometimes obey because they
fear sanctions. They act from prudential, content-independent reasons.
One might also assume that the law also provides its subjects with
content-independent moral reasons to obey. In the twentieth century, the
existence of such reasons came into question with the work of Joseph
Raz, Leslie Green, A. J. Simmons, M. B. E. Smith, Robert Paul Wolff, Rich-
ard Wasserstrom, and many others.5 These authors deny that legal subjects

3. See authors cited below.

4. Robert Paul Wolff, In Defense of Anarchism (New York: Harper and Row, 1970), p. 9;
Joseph Raz, Practical Reason and Norms, 2nd ed. (New York: Oxford University Press, 1990),
p. 178.
5. See, e.g., Meir Dan-Cohen, “In Defense of Defiance,” Philosophy and Public Affairs
23 (1994): 24–51; Joseph Raz, “The Obligation to Obey: Revision and Tradition,” in Ethics
in the Public Domain (Oxford: Clarendon Press, 1994); Leslie Green, The Authority of the
State (Oxford: Clarendon Press, 1988); Kent Greenawalt, Conflicts of Law and Morality
(New York: Oxford University Press, 1987); Rolf Sartorius, “Political Authority and Political
Obligation,” Virginia Law Review 67 (1981): 3–17; A. John Simmons, Moral Principles and
Political Obligations (Princeton, N.J.: Princeton University Press, 1979); M. B. E. Smith, “Is
There a Prima Facie Obligation to Obey the Law?” Yale Law Journal 82 (1973): 950–76;
Wolff, In Defense of Anarchism; Richard A. Wasserstrom, “The Obligation to Obey the Law,”
UCLA Law Review 10 (1963): 780–807. David Lyons even rejects the claim “that justifica-
tion is always required for disobedience to law.” David Lyons, “Derivability, Defensibility,
and the Justification of Judicial Decisions,” in Moral Aspects of Legal Theory (Cambridge:
Cambridge University Press, 1993), p. 135.
160 Limits of Legality: Part I

have a general, content-independent, pro tanto moral reason to obey the

law. A lively debate on “political obligation” continues.6
The question whether ordinary subjects have content-independent
moral reasons to obey the law might be dismissed as a merely scholastic
dispute. Whether or not this dismissal is warranted, the question has great
urgency where judges are concerned. There is some reason to predict that
at least one of the arguments for citizens’ moral duty to obey the law will
also support judges’ moral duty to apply it. Several major writers have
defended a pro tanto moral duty to obey at least some unjust laws.7 Their
arguments might be adapted to support a judge’s moral duty to apply
unjust laws or, more broadly, to adhere to the law in suboptimal-result
cases. Let us consider these arguments.
Someone might suggest that obedience to the law is an end in itself—an
intrinsic good—rather than a means to an end.8 If so, then the fact that the law
requires one to Φ provides one with a pro tanto, nonderivative, content-
independent moral reason to Φ. If a judge is legally required to use force, then
she has a pro tanto, nonderivative, content-independent moral reason to use it.
Is positive law an ultimate, nonderivative source of moral reasons?
I have found no living writer who thinks so. Some believe that the law
simpliciter incorporates parts of morality that are not mentioned in posi-
tive law. But they believe that morality, not positive law, provides the
reasons. Some deny that citizens have a general, pro tanto moral duty to
obey positive law,9 a position that conflicts with the claim that positive
law is an ultimate source of moral reasons. Even those who endorse such
a general duty to obey do not claim that obedience to positive law is an
end in itself—that positive law is the ultimate source of moral reasons.
Rather, they defend a general moral duty to obey the law on the basis of
other ultimate moral principles.
Even if positive law provides nonderivative, content-independent rea-
sons, such reasons can surely be undermined. Consider a duty to carry on
a family tradition. This duty could be understood as nonderivative and
content-independent. If, however, the tradition involves the use of unjust
force, then the duty is undermined. Similarly, I think a judge’s reason to
use force is undermined in suboptimal-result cases. The judge has a
content-dependent reason to avoid suboptimal results that undermines
any content-independent reason he might otherwise have to adhere to
the law.

6. See Edmundson, “State of the Art: The Duty to Obey the Law.”
7. See, e.g., J. L. Mackie, “Obligations to Obey the Law,” Virginia Law Review 67 (1981):
143–58; Tony Honoré, “Must We Obey? Necessity as a Ground of Obligation,” Virginia Law
Review 67 (1981): 39–61; Rawls, A Theory of Justice, pp. 350–55.
8. Mackie, “Obligations to Obey the Law,” p. 151. Note that Mackie advises us to
“invent” the obligation to obey.
9. See works of Raz, Green, Simmons, Wolff, and others cited above.
Legal Duty and Political Obligation 161

10.2 ACTUAL CONSENT

Philosophers have long sought to base a general duty to obey the law on
actual consent. This argument is famously problematic as applied to
natural-born citizens, most of whom never give actual consent to the
state.10 Unlike most natural-born citizens, however, judges definitely
consent to something: they swear an oath to uphold the law. As discussed
in chapter 9, the judicial oath gives judges a moral reason to fulfill their
judicial duties, but I concluded that it gives them no reason to obey
restrictive rule. A parallel objection applies to actual-consent arguments
in support of a duty to obey restrictive rule. Consenting to Φ does not
give the consenting party a moral reason to Φ if Φ-ing is otherwise
immoral, nor does consent attenuate his moral reasons not to Φ. There-
fore, a judge’s consent to obey restrictive rule does not give her a pro
tanto moral reason to obey it, insofar as it requires her to adhere in
suboptimal-result cases.
There are, however, many other theories of political obligation. Can
any of these theories succeed where consent theories fail, supporting rea-
sons for judges to adhere in suboptimal-result cases? One might jump to
the conclusion that these other theories cannot possibly succeed if, as
I have argued, consent theories fail. It is widely assumed that political
obligation would be unproblematic if all citizens expressly consented to
obey, absent fraud or duress. Consent theories fail to support political
obligations for natural-born citizens precisely because they do not consent.
That is the only reason that political philosophers turn to arguments from
fair play, gratitude, et cetera. The latter arguments are designed to do the
work of consent when consent is absent, not to do work that consent
when present fails to do.
It would, however, be a mistake to conclude that these arguments can-
not succeed if consent arguments fail. The question remains open as to
whether a theory of political obligation based on fair play, natural duty, or
gratitude can support a judicial duty to obey restrictive rule. Let us turn
to these theories.

10.3 FAIR PLAY

Consider first the fair play principle: “[W]hen a number of persons engage
in a mutually advantageous cooperative venture according to rules, and
thus restrict their liberty in ways necessary to yield advantages to all, those
who have submitted to these restrictions have a right to a similar

10. But see Murphy, “Natural Law, Consent, and Political Obligation”; Margaret Gilbert,
“Reconsidering the ‘Actual Contract’ Theory of Political Obligation,” Ethics 109 (1999):
236–60.
162 Limits of Legality: Part I

acquiescence on the part of those who have benefited from their

submission.”11 Could this principle give judges a pro tanto moral reason to
obey restrictive rule? When Jack becomes a judge, he joins the judicial
enterprise that constitutes a “mutually advantageous cooperative venture.”
This enterprise is undertaken “according to rules”—the rules of adjudica-
tion. The other judges in Jack’s system “restrict their liberty” by obeying
adjudication rules. Jack benefits from their obedience in several ways.
First, the legal system itself would not exist if judges did not obey cer-
tain adjudication rules. It is very probable that Jack benefits from the legal
system in his capacity as a private legal subject. Perhaps a lone survivalist
in the uncharted wilderness could deny that any legal system benefits
him, but judges cannot.
Second, being a judge probably benefits Jack. He is paid to do it. He
holds a position of honor and esteem in his community. His work may be
intrinsically fulfilling. These benefits would be unavailable to him without
a legal system that, again, requires judicial obedience to adjudication
rules.
Third, in addition to the general ways in which Jack benefits from the
legal system, he may obtain a special kind of benefit from a general prac-
tice of judicial adherence to the law. As a judge, Jack may want to influ-
ence the development of the law. The satisfaction of that desire, if rational,
benefits him. But his efforts to influence the path of the law would be
much less effective, if effective at all, if his fellow judges did not obey
rules such as stare decisis. So Jack benefits from general adherence in
another way if he wants to influence the law.
Although the elements appear to be in place to apply a fair play theory
to judicial duties, these theories are vulnerable to the “limiting argu-
ment,”12 pressed by Robert Nozick and A. J. Simmons.13 One cannot
impose obligations on another by foisting unwanted benefits upon him.14
In response to the limiting argument, George Klosko has rehabilitated fair
play theory with the qualification that the benefits provided must be
“(i) worth the recipients’ effort in providing them and (ii) presumptively
beneficial.”15 A presumptively beneficial good is something everyone is
presumed to want. The benefits of having a legal system and of being

11. Rawls, A Theory of Justice, p. 112. Rawls endorses a similar principle in “Legal
Obligation and the Duty of Fair Play,” in Law and Philosophy: A Symposium, ed. Sidney Hook
(New York: New York University Press, 1964). He follows Hart, “Are There Any Natural
Rights?”
12. So called by George Klosko, “Presumptive Benefit, Fairness, and Political Obliga-
tion,” in The Duty to Obey the Law: Selected Philosophical Readings, ed. William A. Edmundson
(Lanham, Md.: Rowman & Littlefield, 1999), p. 195.
13. Robert Nozick, Anarchy, State, and Utopia (New York: Basic Books, 1974),
pp. 90–95; Simmons, Moral Principles and Political Obligations, ch. 5.
14. Judges agree to take office, of course, but the consent argument was addressed in
§10.2.
15. Klosko, “Presumptive Benefit, Fairness, and Political Obligation,” p. 197.
Legal Duty and Political Obligation 163

a judge in one are certainly goods that every judge can be presumed to
want. Obedience to adjudication rules is necessary for a legal system to
exist, and for a judge to influence the law. If judges obeyed no adjudica-
tion rules, then no legal system could exist.
However, no particular set of adjudication rules is necessary. The advo-
cate of fair play needs a principle stating that an agent has a fair play duty
to obey an established rule if either that rule or some functional equiva-
lent thereof is necessary for the production of benefits enjoyed by the
agent. Therefore, a fair play argument cannot establish that other judges
have a right to demand that Jack obey every adjudication rule that they
obey. At most it establishes the right to demand that Jack obey a certain
adjudication rule only if that rule, or its functional equivalent, is necessary
for the existence of every legal system that would benefit Jack. We must
examine individual adjudication rules in order to ascertain which ones fall
in this category.
Consider adherence rules. Every legal system that would benefit Jack
probably includes such rules. No legal system could be indifferent to judi-
cial adherence to the law. So far I have established that a fair play argu-
ment could support Jack’s duty to obey certain adjudication rules and
that these will include certain adherence rules. But I have not established
which ones. I shall take for granted that general obedience to restrictive
rule benefits Jack. One might conclude, therefore, that the fair play prin-
ciple gives Jack a pro tanto reason to obey restrictive rule. This is because
the usual statements of the fair play principle make it appear to require
individuals who benefit from general obedience to a certain rule to obey
that same rule. However, I think this is an uncharitable reading of the
principle. The only rational statement of the principle requires individuals
to obey either the general rule or an alternative rule that yields at least the
same benefits as the general rule. On my reading the fair play principle
requires Jack to obey either restrictive rule or an alternative rule that
yields at least the same benefits as restrictive rule.
One might object that codes of judicial conduct require judges to be
“faithful to the law,” which one might read as a statement of restrictive
rule. How, in the name of fair play, could Jack justifiably disregard this
rule in favor of another? Let us suppose arguendo that the objector reads
the codes correctly as including restrictive rule. I suggest that restrictive
rule still controls Jack’s reasoning by setting a floor below which he must
not fall. Whatever rule Jack obeys instead of restrictive rule must offer at
least the benefits of restrictive rule.
The next question is whether there is a rule that yields at least the
benefits of restrictive rule, but that gives judges no pro tanto reason to
adhere in some suboptimal-result cases. One alternative is permissive
rule, which forbids deviation in all optimal-result cases but permits it in
all suboptimal-result cases. For the purposes of the fair play principle,
permissive rule seems to offer all the benefits of restrictive rule with none
of its disadvantages. So the fair play principle cannot support a rule that
164 Limits of Legality: Part I

requires adherence in suboptimal-result cases. The more selective rule is

more effective, overall. I conclude that the fair play principle does not
support a duty to obey restrictive rule.

10.4 GRATITUDE

Some writers have defended a duty to obey the law as a duty of grati-
tude.16 One owes such a duty when someone acts for one’s sake, whether
or not one requests or desires it. The state takes many actions for the sake
of its citizens, including judges. I shall accept arguendo that a judge is
obligated to do something that properly expresses his gratitude. Obeying
adjudication rules might be one way to do so. But discharging this duty
does not seem to require anything as specific as obeying adherence rules,
particularly in suboptimal-result cases. At most, we could perhaps say
that a judge has a conditional duty of gratitude to obey adherence rules if
he does not adequately express his gratitude in any other way.
Also note that some actions are inappropriate and impermissible
expressions of gratitude. Ron saves Daisy’s life at great cost to himself.
Daisy owes Ron a large debt of gratitude. Ron asks Daisy to shoplift some-
thing for him as an expression of her gratitude. Daisy has not even a pro
tanto reason to do so. Similarly, adhering to the law in suboptimal-result
cases seems a particularly inappropriate expression of gratitude because it
involves performing actions that are otherwise pro tanto wrongful.

10.5 NATURAL DUTY

Rawls argues that individuals have a “fundamental natural duty . . . of

justice.” We are obligated
to support and to comply with just institutions that exist and apply to us. [The
duty] also constrains us to further just arrangements not yet established, at least
when this can be done without too much cost to ourselves. Thus if the basic
structure of society is just, or as just as it is reasonable to expect in the circum-
stances, everyone has a natural duty to do his part in the existing scheme.17

If Rawls is correct that we have a duty to comply with and to do our share
in just institutions, and certain adjudication rules are part of a just institu-
tion, then judges have a natural pro tanto duty to obey those rules.
What might a Rawlsian say about adherence rules, specifically? Rawls
himself acknowledges that unjust laws will be enacted even if the consti-
tution and legislative process are just: “[T]here is no feasible political

16. Walker, “Political Obligation and the Argument from Gratitude.”

17. Rawls, A Theory of Justice, p. 115.
Legal Duty and Political Obligation 165

process which guarantees that the laws enacted in accordance with it will
be just.”18 He notes that “among the very limited number of procedures
that have any chance of being accepted at all, there are none that would
always decide in our favor . . . and . . . consenting to one of these proce-
dures is surely preferable to no agreement at all.”19 Procedures such as
majority rule will inevitably produce some unjust laws: “In choosing a
constitution, then, and in adopting some form of majority rule, the parties
accept the risks of suffering the defects of one another’s knowledge and
sense of justice in order to gain the advantages of an effective legislative
procedure.”20 Rawls holds that, at least in a state of near justice, “there is
normally a duty . . . to comply with unjust laws provided that they do not
exceed certain bounds of injustice.”21 If this principle applies directly to
judges and adherence rules, then we could conclude that, at least in a
reasonably just system, judges have a duty to obey whatever adherence
rules the lawmakers enact, provided that these do not exceed certain
bounds of injustice. If the lawmakers enact restrictive rule, then judges
have a moral duty to obey it, on this view.
I think Rawls’s discussion of disobeying unjust laws suffers from a
major limitation that is common to most discussions of the topic. His
position may be defensible regarding the most common kinds of situa-
tions in which private parties face unjust laws. But his perspective on
disobedience is fundamentally that of the private citizen. My question
concerns the duties of public officials. Judges have moral reasons to dis-
obey restrictive rule, yet Rawls does not reach the question of whether
moral reasons to disobey can undermine moral reasons to obey. This omis-
sion on his part is understandable. Unjust laws rarely require private indi-
viduals to perform immoral acts or to omit morally obligatory acts. Most
unjust laws merely require individuals either to perform certain morally
permissible acts or to omit certain morally optional acts. Typical unjust
laws deprive individuals of their rights. By contrast, restrictive rule requires
judges to perform actions that would otherwise be immoral.22
I think Rawls’s restriction of civil disobedience to “serious violations of
justice” is, therefore, inappropriate for judges. In my view, the natural duty
to obey the law does not give a judge a reason to obey an adherence rule
that requires him to perform actions that are even slightly immoral. At
least, if a judge is choosing between two adherence rules, one of which
requires him to perform an action that is slightly more immoral than any

18. Ibid., p. 353.

19. Ibid., p. 354.
20. Ibid., p. 355. Rawls does not distinguish between unjust laws and just laws that
dictate unjust results, or between suboptimal laws and optimal laws that dictate suboptimal
results. But presumably he would acknowledge the existence of negative-gap cases, as well
as suboptimal-rule cases.
21. Ibid., p. 355.
22. See further discussion in §10.7.
166 Limits of Legality: Part I

action that the second rule requires, then he may choose to obey the sec-
ond rule. Or if the second rule requires him to perform two slightly
immoral actions and the first rule requires him to perform only one such
action, then he may choose to obey the first rule.
Rawls claims that parties in the original position23 would choose a prin-
ciple that requires us “to comply with and to do our share in just institu-
tions when they exist and apply to us.”24 But he does not reach the question
of whether they would understand it to entail that judges have pro tanto
moral reasons to adhere in suboptimal-result cases. At most, his argu-
ments show only that parties in the original position would choose prin-
ciples that supply pro tanto moral reasons to obey the law and that these
reasons are not undermined by private reasons. His arguments do not
show that moral reasons to disobey cannot undermine moral reasons to
obey. In fact, Rawls permits limited civil disobedience even in a nearly just
society: even “forceful resistance may later be entertained.”25 However,
Rawls permits disobedience only for “serious violations of justice.”26 He
thinks we have a duty to obey laws that are only slightly unjust because
“[w]ithout some recognition of this duty mutual trust and confidence are
liable to break down.”27 The idea, which I accept, is that trust and confi-
dence break down if individuals suspect one another of disobeying for
private reasons. Disobeying for private reasons raises this suspicion. So
individuals should not disobey for private reasons. By extension, judges
must not deviate for private reasons, meaning that they must obey at least
permissive rule. This conclusion extends to disobedience of laws that are
only slightly unjust. The less obviously unjust a law is, the more likely it
becomes that disobedience will be perceived as privately, rather than
morally, motivated. So individuals should obey slightly unjust laws because
disobeying will be perceived as motivated by private reasons.
A Rawlsian might try to apply this argument to judges, producing an
argument in favor of obeying restrictive rule. I think such an extension
fails. Rawls envisions a typical case of disobedience in which a private in-
dividual breaks the law, benefits in a tangible way from doing so, and
believes that she has a good chance of evading the legal consequences.
Under these conditions, given what we know of human behavior, it is
reasonable to infer that she disobeys for private reasons. This inference is
reasonable even if the law happens to be seriously unjust and disobedi-
ence morally justifiable.
By contrast, in a legal system that assigns cases randomly and ade-
quately polices judges for conflicts of interest, corruption, bribery, and
other forms of undue influence, everyone should have confidence that

23. For description of the original position see Rawls, A Theory of Justice, ch. 3.
24. Ibid., p. 334.
25. Ibid., p. 366.
26. Ibid., p. 363.
27. Ibid., p. 355.
Legal Duty and Political Obligation 167

judges who deviate do not do so for the obvious private reasons: money,
sexual favors, promotion, revenge, et cetera. When a judge deviates he
does not benefit in any tangible way. If Posner is correct, then the deviat-
ing judge actually suffers displeasure at not “playing the judicial game”
according to the rules.28 It is not reasonable to infer that he deviates for
self-serving reasons, even if the legally required result is arguably optimal.
The reasonable inference is that the judge sincerely, although perhaps
mistakenly, believes the legally required result to be suboptimal. This
inference is even more reasonable if the legally required result is arguably
suboptimal, even if only slightly so. So judges who deviate in slightly sub-
optimal-result cases do not threaten mutual trust and confidence. This is
not, of course, to deny that patterns of deviation can have other negative
effects on the legal system. That is the subject of part II. These effects,
I shall argue, give practical authority to moderate rules over permis-
sive rule.

10.6 SAMARITANISM

Samaritan accounts of the duty to obey the law begin with the premise
that each of us has the right to use force (or coercion) when necessary to
rescue others.29 Likewise, the state has the right to use force in order to
rescue its residents from the perils of the state of nature. Each of us then
has a duty to obey the law, lest he interfere with the state’s performance
of its samaritan duties.
A samaritan account of the judge’s duty to adhere to the law is still
simpler. Everyone has a natural samaritan duty to rescue others, at least
when he can do so at little cost to himself. A judge occupies an unusually
good position to rescue others from the perils of the state of nature at little
cost to himself, merely by adhering to the law when he decides cases.
When a judge adheres to the law in a suboptimal-result case, however,
he does not rescue the losing party from the state of nature. On the con-
trary, he partially returns the party to the state of nature and/or prevents
others from rescuing her from a partial state of nature. When a judge
enters a judgment against a defendant who deserves better, he uses unjus-
tified coercion against her and forcibly prevents others from protecting or
rescuing her.
When a judge deviates from the law in a suboptimal-result case, he
does not interfere with the state’s performance of its samaritan duties.
Rather, he performs these duties when the state would otherwise fail to
do so. Therefore, samaritan accounts of the duty to obey the law do not
support the practical authority of restrictive rule.

28. See §4.6.

29. Wellman, “Liberalism, Samaritanism, and Political Obligation.”
168 Limits of Legality: Part I

10.7 OBEYING UNJUST LAWS

In this chapter I conceded arguendo that judges have a legal duty to adhere
in suboptimal-result cases, but I proceeded to deny that they have even a
pro tanto moral duty to do so. Therefore, a reader might assume that my
position presupposes that there are no legitimate states or at least that
there is no pro tanto moral duty to obey the law—certainly reputable
philosophical positions, but still controversial.30 I cannot predict where
the ongoing debates will lead, so I have actually tried to avoid assuming
either philosophical anarchism or the nonexistence of a pro tanto duty to
obey. In fact, I have shown how standard arguments for a duty to obey can
be adapted to a narrower purpose. Assuming that judges have a legal duty
to obey permissive rule, the standard arguments can be used to support a
corresponding moral duty.
I have, however, rejected every argument so far for the conclusion that
judges are morally obligated to obey a nonpermissive rule. I have found no
reason to deny that judges are, at least sometimes, morally permitted to
deviate. If judges have a legal duty to adhere, then my tentative conclu-
sion is inconsistent with an all-things-considered moral duty to obey the
law in every instance. But modern proponents of a duty to obey defend
only a pro tanto duty, so they need not disagree with me on this score.
In fact, my position is compatible with a surprisingly broad all-things-
considered duty to obey—one that encompasses even the majority of
unjust laws that exist in the modern world. I think the following views are
mutually consistent: (1) subjects have an all-things-considered moral
obligation to obey most unjust laws; (2) judges are legally obligated to
adhere; and (3) judges have no moral obligation to obey restrictive rule.
These views are consistent because the law rarely requires ordinary sub-
jects to use force at all, much less unjustified force. Sometimes laws
require this, but only rarely, even in the most unjust legal systems the
world has known. By contrast, restrictive rule requires judges to perform
pro tanto immoral actions. Within the set of injustices perpetrated by
states we can draw several distinctions. First, we can distinguish legally
authorized injustices from injustices perpetrated ultra vires—without
legal authority or in violation of the law. Within each of these categories
we can distinguish mandated injustices from nonmandated injustices. An
injustice is mandated if and only if either a law or a public official acting
under color of law requires a private party either to commit a morally
impermissible act or to omit a morally obligatory act.
It is easy to imagine mandated injustices. Laws could require individ-
uals to molest children, torment animals, break promises, et cetera. His-
torically, however, only a small fraction of the injustices perpetrated by

30. See, e.g., William A. Edmundson, Three Anarchical Fallacies (Cambridge: Cambridge
University Press, 1998).
Legal Duty and Political Obligation 169

states have involved laws requiring private parties to act immorally or to

omit morally obligatory acts. Examples include the following: conscript-
ing men to fight in unjust wars; requiring private parties to reveal the
location of fugitive slaves or to deliver other innocent victims to malevo-
lent authorities; requiring realtors to discriminate on the basis of race;
et cetera.31 But most unjust laws have not required private parties to act
immorally. Even less often have unjust laws required them to use morally
unjustified force. Consider typical examples of laws that most readers will
consider highly unjust:
1. Until the mid-nineteenth century, married women had no legal
right to own real property in the United States, but the law did
not forbid private parties from selling land to married women; it
merely gave title to their husbands.32 This was a grave injustice,
but the law did not require private parties to act immorally or to
omit any morally obligatory acts.
2. The Cuban constitution protects freedom of speech and press only
insofar as they “conform to the aims of socialist society.”33 Cuban
law authorizes public officials to punish dissidents, but it does not
require private parties to act immorally or to omit any morally
obligatory acts. Assuming that no Cuban is morally obligated to
engage in antisocialist speech (at least if the state has banned it),
these injustices are nonmandated.
3. The Thirteenth Regulation under the Nazi Reich Citizenship Law
stripped German citizenship from Jewish citizens living overseas.34
Other examples of extremely unjust laws that did not require private
parties to use unjustified force include laws requiring public officials to
protect the property rights of slave owners (e.g., by capturing fugitive
slaves), antimiscegenation laws, laws segregating public facilities, most of
the laws that sustained South African apartheid, and the federal laws that
forcibly relocated and interned Americans of Japanese descent during
World War II. In the overwhelming majority of cases in which a private
party is subjected to an unjust law or unjust treatment by public officials
acting ultra vires, he is himself a victim of the injustice. Rarely does the

31. Some injustices have involved public officials, acting ultra vires, requiring private
parties to act immorally or to omit morally obligatory acts. There are reports of public offi-
cials in the People’s Republic of China requiring parents to commit infanticide, although the
state denies these reports. See Xiaorong Li, “License to Coerce: Violence against Women,
State Responsibility, and Legal Failures in China’s Family-Planning Program,” Yale Journal of
Law and Feminism 8 (1996): 145–91.
32. See Marylynn Salmon, Women and the Law of Property in Early America (Chapel
Hill: University of North Carolina Press, 1986).
33. Cuba Const. (1992), art. 53.
34. See David Fraser, “‘This Is Not Like Any Other Legal Question’: A Brief History of
Nazi Law before U.K. and U.S. Courts,” Connecticut Journal of International Law 19 (2003):
59–125.
170 Limits of Legality: Part I

unjust law or treatment require him to act immorally or to omit morally

obligatory acts. The injustice is usually nonmandated.
The thesis held by Rawls and others35 that we have pro tanto reasons to
obey unjust laws makes sense if we limit our attention to private parties. The
typical unjust law requires private parties to act contrary to their preferences,
welfare, and values. It may prevent them from realizing their goals and fulfill-
ing their obligations (e.g., to provide for their children, to marry the mothers
of their children), but it does not require them to perform immoral positive
actions or to violate anyone else’s rights. So they have no countervailing moral
reason to undermine the putative pro tanto moral duty to obey the law.
Emily emigrates from a reasonably just state to another state, knowing
that in the new state the (presumably unjust) laws will require her to pay poll
taxes, forbid her to worship God, and forbid her to own real estate. Perhaps,
although I emphasize the uncertainty, Emily thereby acquires a pro tanto
duty to obey these laws if others are relying on her to do so. There is a ques-
tion of whether that reliance is normatively reasonable, but I shall ignore this
worry. I shall accept arguendo that with respect to most unjust laws, one who
gives actual consent to an unjust law can acquire a pro tanto duty to obey it.
In these respects the law imposes very different demands on public
officials than on private parties. In impermissible-result cases, the law
requires judges either to act in ways that would otherwise be immoral or
to omit acts that would otherwise be morally obligatory. These are man-
dated injustices. Even in suboptimal-but-permissible-result cases, judges
have moral reasons to deviate from the law. Being moral reasons them-
selves they can undermine moral reasons to obey. Therefore, even if con-
senting to obey a typical unjust law generates a pro tanto moral reason to
obey, a judge’s consent to obey restrictive rule would not give him a pro
tanto moral reason to obey it because doing so could involve using unjus-
tified force, which he has moral reasons not to use.36
Most of the actions that one agent consents to perform on behalf of
another, or to assist another to perform, are actions that the first agent has
a right to perform. Therefore, consent gives the first agent a duty to
perform or to assist in the performance of those actions. However, if one
has no moral right to Φ, then one cannot acquire a right to have someone
Φ on one’s behalf or assist one to Φ. Moreover, the legally favored parties
in suboptimal-result cases have no moral right to do what the law would
allow them to do, or to receive assistance to do it, or to have it done on
their behalf. Therefore, judges do not owe them adherence to the law in
these cases. The legally disfavored parties did not necessarily consent to
the law even if their judge did.

35. Rawls, A Theory of Justice, pp. 350–55; Honoré, “Must We Obey? Necessity as a
Ground of Obligation.”
36. Judges have pro tanto moral reasons not to consent to obey restrictive rule in the first
place. If you ask a judge, “Will you uphold the law when you believe that it requires an
unjust result?” I think her honest answer should be, “Not always.” See part II.
Legal Duty and Political Obligation 171

10.8 PUNISHMENT WITHOUT LAW

So far I have found no argument for obeying restrictive rule, but now
I must address an important special case. The suboptimal-result cases that
I have had in mind thus far fall in two categories: (1) civil cases in which
the law requires a judgment for the plaintiff; and (2) criminal cases in
which the law requires a guilty verdict and/or punishment. When judges
deviate in such cases, favoring defendants, no one can object that his moral
rights have been violated. In a suboptimal-result case, by definition, if the
law favors the plaintiff or prosecution, then the plaintiff or prosecution
had no moral right to win in the first place.
When the law favors a civil or criminal defendant, by contrast, special
questions arise about the moral permissibility of deviation. Deviation in
such cases imposes liability without legal authorization. Furthermore, defen-
dants in such cases face adverse judgments without having been duly noti-
fied that their actions could have such consequences. Deviation that results
in a criminal conviction is especially troublesome because it violates the
principle of nullum crimen, nulla poena, sine lege.37 Many would also hold
that deviation violates nullum crimen when it results in a sentence more
severe than the law permits. A supporter of nullum crimen could conclude
that judges must never deviate from laws favoring criminal defendants.
This depends, however, on what exactly nullum crimen forbids. At a
minimum it forbids convicting a defendant for an action that is neither
malum in se nor malum prohibitum. This prohibition is also entailed by
minimal retributivism,38 a principle that many readers will accept. Mini-
mal retributivism holds that the state is morally forbidden to punish
anyone in excess of what he deserves from a moral standpoint. No one
deserves to be punished for an action that is neither malum in se nor
malum prohibitum. If minimal retributivism is true, then it is never subop-
timal to acquit a defendant for such an action. If the law requires acquittal,
then the defendant’s action is not, by definition, malum prohibitum. There-
fore, there are no suboptimal-result cases in which the law requires

37. “No crime, no punishment, without law.” This maxim “dates from the ancient
Greeks.” Jerome Hall, General Principles of Criminal Law, 2nd ed. (Indianapolis, Ind.:
Bobbs-Merrill, 1960), p. 59. See also Joshua Dressler, Understanding Criminal Law, 3rd ed.
(New York: Lexis, 2001), pp. 39–40; Rogers v. Tennessee, 532 U.S. 451, 467–68 (2001);
Sparf v. United States, 156 U.S. 51, 88 (1895).
38. See J. L. Mackie, “Morality and the Retributive Emotions,” in Persons and Values
(Oxford: Clarendon Press, 1985). Minimal retributivism is also known as negative retributiv-
ism, weak retributivism, and limiting retributivism. See C. L. Ten, “Positive Retributivism,”
Social Philosophy and Policy 7 (1990): 194–208; Heidi M. Hurd, Moral Combat (Cambridge:
Cambridge University Press, 1999), p. 1; Paul H. Robinson, “The A.L.I.’s Proposed Distribu-
tive Principle of ‘Limiting Retributivism’: Does It Mean in Practice Anything Other than
Pure Desert?” Buffalo Criminal Law Review 7 (2003): 3–15. I challenge minimal retributiv-
ism elsewhere, but do not rely on those arguments here. See Jeffrey Brand-Ballard, “Inno-
cents Lost: Proportional Sentencing and the Paradox of Collateral Damage,” Legal Theory
15 (2009): 65–105.
172 Limits of Legality: Part I

acquittal and the defendant’s action is not malum in se. So far there are no
suboptimal-result cases in which deviation violates nullum crimen. Notice
that nullum crimen remains a strong principle of limited government, with
much work to do, even if we limit its scope to actions that are neither
malum in se nor malum prohibitum.
The interesting cases involve actions mala in se that are not prohibited
under criminal law. Taken literally, nullum crimen also extends to these
cases. If there are any suboptimal-result cases in which the law requires
acquittal, they are these. A hypothetical illustrates. Chester visits the
(fictional) nation of Vortulia. Vortulian officials tell Chester that Vortulian
law permits inducing minor children by fraudulent means to engage in
sexual activity.39 Relying upon their assurances, Chester fraudulently
induces a nine-year-old into a sexual act. Although he has broken no law,
Chester is charged with a sex crime. Is acquitting Chester a suboptimal
result?
To answer this question, we must ask whether a judge would have an
all-things-considered reason to convict Chester if the law actually permit-
ted convicting him, despite the fact that Chester believed his conduct to
be lawful. Imagine for the moment an alternative scenario. Lester visits
the nation of Mortulia. Mortulian officials tell Lester that Mortulian law
permits inducing children by fraudulent means to engage in sexual activ-
ity. However, the officials are misinformed about Mortulian law, which
actually gives judges discretion whether to convict such defendants. Les-
ter does not understand this either. Relying upon this misinformation, he
fraudulently induces a nine-year-old into a sexual act. He is charged with
a sex crime. Does a Mortulian judge who is legally authorized to convict
and punish Lester have an all-things-considered reason to do so? One’s
answer to this question will depend upon one’s theory of punishment,
among other factors. Consider one who accepts minimal retributivism.
She might accept in addition one of these three principles:
1. If someone believes his action to be lawful, then he does not
deserve punishment for it.
2. If someone reasonably believes his action to be lawful, then he
does not deserve punishment for it.
3. If someone reasonably believes his action to be lawful, then the
state must not punish him for it, even if he deserves punishment.
Because Lester reasonably believes that tricking children into sex acts is
lawful in Mortulia, any of these three principles entails that a Mortulian
judge has no all-things-considered reason to convict Lester, assuming min-
imal retributivism. Assuming that Chester’s and Lester’s cases are other-
wise identical, each principle also entails that acquitting Chester, back in

39. The fraud might involve telling the child that you will make her into a movie star if
she performs the sex act when you have no intention of doing so.
Legal Duty and Political Obligation 173

Vortulia, is not a suboptimal result. Each principle entails that there are
no suboptimal-result cases in which the law requires acquittal.
We can, however, imagine stricter principles of punishment that would
give the Mortulian judge a reason to convict Lester. Consider the follow-
ing retributive argument:
1R. If punishing a defendant for a seriously immoral action is
permitted by law, then this immorality gives the state a reason
to punish him.
2R. Lester’s actions are seriously immoral.
3. Punishing Lester is permitted by law.
4. Therefore, Mortulia has a reason to punish Lester.
There is also an act-consequentialist argument for this conclusion:
1C. If punishing a defendant is permitted by law and promotes the
good, then the fact that it promotes the good gives the state a
reason to punish him.
2C. Punishing Lester promotes the good.
3. Punishing Lester is permitted by law.
4. Therefore, Mortulia has a reason to punish Lester.
If both 1R and 1C are false, then there are no suboptimal-result cases
in which the law requires acquittal. If either 1R or 1C is true, then such
cases might exist, although they are probably rare. If 4 is true and Mortulia
has no stronger reason not to punish Lester, then Mortulia has an
all-things-considered reason to punish Lester. This entails that acquitting
Chester, back in Vortulia, is a suboptimal result. However, if we take
nullum crimen literally, then it applies to Chester’s case and forbids pun-
ishment. We appear to have found a case in which nullum crimen forbids
reaching the optimal result.
Here we reach a fork in the road. One could adopt an exceptionless
version of nullum crimen and treat cases in which the law requires a sub-
optimal acquittal as special cases in which judges are obligated to reach
suboptimal results simply because nullum crimen requires it. Readers who
are sufficiently committed to an exceptionless version of nullum crimen
will take this path.
Alternatively, one could qualify nullum crimen, arguing as follows. If
either argument for 4 is sound, then nullum crimen as applied to Chester’s
case becomes difficult to justify. This is so because the moral basis for nul-
lum crimen is the idea that individuals must not be convicted or punished
for actions that they reasonably believe to be lawful. However, both argu-
ments for 4 entail that it is sometimes morally permissible to punish
defendants who reasonably believe their actions to be lawful. The retrib-
utive argument entails that it is permissible in the case of seriously im-
moral actions. The consequentialist argument entails that it is permissible
if punishing the action promotes the good. If the retributive argument is
sound, then there is no moral justification for extending nullum crimen to
174 Limits of Legality: Part I

seriously immoral actions. If the consequentialist argument is sound, then

there is no moral justification for extending nullum crimen to actions the
punishment of which promotes the good.
Some special remarks apply to the consequentialist argument. Premise
2C might be false. Punishing Chester for his actions promotes the good
only if similar actions are regularly punished in Vortulia. Because Chester’s
actions are lawful in Vortulia it is unlikely that a regular practice of
punishing such actions exists. So it is unlikely that punishing Chester
promotes the good. Therefore, it is unlikely that acquitting Chester is a
suboptimal result. If punishing Chester does, perchance, promote the
good, then acquitting him is a suboptimal result, but nullum crimen is not
properly applied to his punishment in that instance.
I should also mention two opinions of mine that bear on the nullum
crimen issue, although my larger argument does not rely upon them. First,
I do not believe that modern Western states criminalize too little.40 There-
fore, I believe that acquittal is rarely a suboptimal result when the substan-
tive criminal law requires it. Acquittal might be suboptimal, however,
when procedurally required because of a statute of limitations, police or
prosecutorial misconduct, et cetera. This brings me to my second opinion:
rejecting 1R. In my view the fact that someone deserves to suffer, morally
speaking, is rarely (if ever) a good reason for the state to convict or punish
him. This is not the dominant opinion in contemporary philosophy, but it
is mine. If 1R is false, then the state is permitted to punish only for non-
retributive reasons, as 1C holds. Nevertheless, there are gap cases in which
it promotes the good to convict a defendant who must be acquitted under
law, as when prosecutorial misconduct requires acquitting a dangerous
defendant who should be incarcerated for the sake of the general welfare.
The procedural rules that generate such cases may, of course, be good rules
to promulgate, but I do not believe that courts should treat nullum crimen
as an independent moral principle that extends to such cases. I do not
believe that judges are morally obligated to acquit in all such cases,
although my argument in part II entails that they are so obligated in most
such cases.
Again, my larger argument does not depend upon my idiosyncratic
rejection of 1R. I have argued that reaching optimal results does not
violate nullum crimen, properly understood, whether 1R is true or false.

10.9 RECUSAL OR RESIGNATION

Part I is nearing an end. Its lesson is that judges retain pro tanto moral
reasons not to reach suboptimal results even when they are legally required

40. See, e.g., Douglas Husak, Overcriminalization: The Limits of the Criminal Law
(New York: Oxford University Press, 2008).
Legal Duty and Political Obligation 175

to do so. The law does not undermine these reasons to deviate. But are
these really reasons to deviate, specifically? Judges always have another
option: recusal.
Let us revisit a Catholic perspective on this issue from §6.11. As do
most commentators who have addressed the topic, Judge William Pryor
simply assumes that recusal is morally permissible but will rarely be
necessary for a Catholic judge.41 Pryor just wants to avoid material coop-
eration with specific actions designated as intrinsically evil by his church.
Avoiding such cooperation rarely requires recusal—only in cases involving
abortion, divorce, capital punishment, unjust warfare, et cetera. In this
book, by contrast, I consider a much larger set of suboptimal-result cases.
I assume that all uses and threats of force against other human beings
demand moral justification. Even threatening to use nonlethal force to
defend one’s life against an imminent threat from a culpable attacker
requires moral and legal justification, although adequate justifications
exist in such cases.42
As I explained in chapter 2, judges threaten and/or use force with vir-
tually every ruling. The judge’s cooperation with these actions is proxi-
mate, not remote. The judge orders his subordinates to threaten or use
force. The use of force or threats thereof is difficult to justify in
suboptimal-result cases. On many theories of justice, plenty of cases are
suboptimal-result cases, even in reasonably just legal systems. Therefore,
whether judges are morally permitted to recuse themselves in every sub-
optimal-result case is a more urgent question for me than for Catholic
scholars.
I begin by extending the scope of one tenet of Pryor’s position. Just as
he claims that judges are morally permitted to recuse themselves in order
to avoid “impermissible cooperation with evil,” I shall assume that judges
are morally permitted to recuse themselves in order to avoid legally man-
dated suboptimal results, whether or not these meet the Catholic defini-
tion of evil. Again, I am not sure that judges are morally permitted to
recuse themselves under these conditions, but I shall assume that they are
so as to reach more important issues.
The important question is whether judges are morally required to
recuse themselves in suboptimal-result cases. Pryor implies that they are
in the cases that concern him. Church doctrine forbids adhering to the
law when it conflicts too directly with Catholic teachings. Pryor does not
entertain the possibility of deviating in such cases. If adhering would

41. William H. Pryor, Jr., “The Religious Faith and Judicial Duty of an American Catholic
Judge,” Yale Law and Policy Review 24 (2006): 347–62, pp. 359, 361.
42. Kate brandishes a knife and demands Jodie’s purse. The remarkably even-tempered
Jodie threatens to punch Kate in the stomach. Under the law, both Kate and Jodie could be
prosecuted: Kate for assault with a deadly weapon, and Jodie for misdemeanor assault. Jodie,
however, has a complete legal defense. See Model Penal Code § 3.04(1) (1985) (use of force
in self-defense is justified when used to protect self from present threat).
176 Limits of Legality: Part I

conflict too directly with Catholic doctrine, then a Catholic judge has a
duty to recuse himself, according to Pryor.43
Interestingly, some individuals who have faced this conundrum iden-
tify resignation, rather than recusal, as the only permissible alternative to
adherence. Judge Thomas Gee of the Fifth Circuit expressed such a view
in the course of explaining his decision to defer to the Supreme Court
after it reversed the judgment of his court: “Subordinate magistrates such
as I must either obey the orders of higher authority or yield up their posts
to those who will. I obey, since in my view the action required of me by
the Court’s mandate is only to follow a mistaken course and not an evil
one.”44
Judge Lois Forer, by contrast, resigned:
I was faced with a legal and moral dilemma. As a judge I had sworn to uphold the
law, and I could find no legal grounds for violating an order of the supreme court.
Yet five years’ imprisonment was grossly disproportionate to the offense. . . .
Given the choice between defying a court order or my conscience, I decided to
leave the bench where I had sat for 16 years.45

Mahatma Gandhi seems to agree with Judges Gee and Forer that judges
must either enforce the law or resign. On March 18, 1922, Gandhi was
indicted and convicted at the Circuit House at Shahi Bag. Upon his
indictment, he addressed the judge as follows:
The only course open to you, the Judge, is either to resign your post and thus
dissociate yourself from evil, if you feel that the law you are called upon to
administer is an evil and that in reality I am innocent; or to inflict on me the
severest penalty if you believe that the system and the law you are assisting to
administer are good for the people of this country and that my activity is there-
fore injurious to the public weal.46

Apparently even recusal would not satisfy Gandhi. He wants his

judge to dissociate himself entirely from the colonial system, which
requires resignation from the bench, not just recusal. Notice also that
Gandhi insists that the judge has only two options: resign or inflict the
“severest penalty.” This is not literally true: the judge could have inflicted
a light sentence. He could also have deviated, overtly or covertly. He
could have announced that the law required him to convict and sen-
tence Gandhi, but that the law in this case dictated such an unjust result
that he would refuse to apply it. This outcome might not have served
Gandhi’s political strategy of embarrassing the British Empire, but if the
argument of part I is sound, then it would have been ethically permissible

43. Pryor, “The Religious Faith and Judicial Duty of an American Catholic Judge,”
p. 361.
44. Weber v. Kaiser Aluminum & Chemical Corp., 611 F.2d 132, 133 (5th Cir. 1980).
45. Lois Forer, “Justice by Numbers,” Washington Monthly, April 1992, p. 12.
46. M. K. Gandhi, Selected Writings of Mahatma Gandhi (Boston: Beacon Press, 1951),
p. 145. Thanks to Michèle Friend.
Legal Duty and Political Obligation 177

for the judge to make such an announcement, Gandhi’s claim

notwithstanding.
Recusal also raises its own moral issues. Judges have pro tanto moral
reasons against recusal—reasons to hear and decide all assigned cases. This
is obvious regarding permissible-result cases. Recusal delays proceedings
and imposes administrative costs on the court system. The presence of
such reasons explains why judges must not recuse themselves for private
reasons such as laziness. There are also fair play considerations. However,
I have argued that judges have pro tanto moral reasons not to adhere to
the law in suboptimal-result cases. One might argue that in a suboptimal-
result case, the judge’s reasons not to adhere to the law outweigh her
reasons to hear and decide the case, yielding a moral reason to recuse
herself.
Deviation is another option. By deviating, a judge respects both com-
peting reasons: his reason not to adhere and his reason to decide the case.
Ceteris paribus, deviating seems like the ideal choice, but Gandhi, Pryor,
Gee, and Forer do not so much as mention the possibility. Perhaps they
believe it to be morally impermissible as well, leaving recusal or resigna-
tion as the only permissible options. They could reply that the judge has
moral reasons not to deviate. These putative reasons would then combine
with his reasons not to adhere, outweighing his reasons to decide the case
and yielding a reason for recusal.
The problem with this argument is that we have not yet discovered
reasons against deviation that extend to suboptimal-result cases. I have
examined many reasons in part I and found that they are neutralized in
suboptimal-result cases. In fact, I see no arguments, even from Catholic
principles, against the conclusion that judges have pro tanto moral reasons
to deviate in suboptimal-result cases, although I am no expert in Catholic
ethics.47 The Catholic arguments for the conclusion that judges have
strong moral reasons to recuse themselves seem also to support the con-
clusion that judges have pro tanto moral reasons to deviate. This is so even
if Catholics also believe that these pro tanto reasons to deviate are usually
outweighed by reasons to adhere or to recuse. Church teachings appear to
entail that Catholic judges must not obey adherence rules, such as restric-
tive rule, that exclude as a reason against applying the law the fact that
doing so constitutes impermissible material cooperation with evil as the
Church understands it.
Moreover, we have reason to believe that a judge’s reason to deviate
in a suboptimal-result case is usually stronger than her reason to recuse
herself. By recusing herself in a suboptimal-result case, the judge simply
passes the buck. Not only does she delay proceedings and impose
administrative costs on the judicial system, but, more important, she

47. There are, however, also pro tanto reasons to adhere in suboptimal-result cases,
which I discuss in part II. Catholic ethicists could surely state their own pro tanto reasons to
adhere.
178 Limits of Legality: Part I

substantially and unnecessarily increases the chance that the case will
eventually be disposed of in a suboptimal way. The reason is simple.
After she recuses herself, the next judge who is assigned to the case will
either adhere, or deviate, or recuse himself, in turn. If he adheres to the
law, then a suboptimal result occurs, just as if the original judge had
adhered. If he deviates, then the suboptimal result is avoided, anyway.48
However, if the original judge decides the case and deviates, then she
reduces the chance that a suboptimal result will ultimately be reached
in the case after any appeals are exhausted. I conclude that the judge
has a pro tanto moral reason to deviate, not just a reason to self-recuse,
in a suboptimal-result case. This is not to say that she has an all-things-
considered reason to deviate—we have not yet reached that point. As I
shall discuss in part II, reasons to deviate face competition from reasons
to adhere. Therefore, the judge might or might not have an all-things-
considered reason to adhere to the law in a given suboptimal-result
case, depending on the relative strengths of the competing reasons for
and against deviation. Such a position contradicts the claim that re-
strictive rule has practical authority for judges.
If judges have no all-things-considered moral reason to obey restrictive
rule, then they are sometimes morally permitted to deviate in suboptimal-
result cases. Permissive rules and moderate rules permit this. In part II,
I shall defend moderate rules.

48. Notwithstanding appeals, et cetera.

Part II

Let us pause to notice where we stand in the dialectic. I have concluded

that judges have pro tanto moral reasons to deviate in suboptimal-result
cases. Therefore, suboptimal-result cases present apparent conflicts
between moral reasons and the demands of the law. In these cases judges
have strong, although not necessarily conclusive, moral reasons to commit
or omit actions that they must omit or commit, respectively, if they are to
adhere to the law. They have strong moral reasons to use force when the
law forbids them to do so and strong moral reasons to refrain from using
force when the law requires them to do so. Schauer observes:
The outcome of a legal decision may make a litigant a prince or a pauper,
famous or infamous, a success or a failure. And when the litigants for whom
such consequences attach are standing before the legal decision-maker, the
pressure to reach the correct result, rather than the substantively incorrect
result generated by faithful application of the rules, is likely to be enormous.1

In part I, I sought objective all-things-considered moral reasons for judg-

es to obey restrictive rule. My search was unsuccessful. If that were the
end of the story, then we would be left with the conclusion that judges
are objectively morally permitted to deviate in all suboptimal-result
cases.
Maybe that is the end of the story—maybe judges are morally per-
mitted to deviate in all suboptimal-result cases. If so, then we may be
forced to revise our conception of the rule of law, and the contribution
of this book consists exclusively in part I, which established the need
for revision. But most observers of the courts seem entirely confident
that judges act impermissibly if they deviate in every suboptimal-result
case. In fact, some writers believe that judges are never morally permit-
ted to deviate. Many others believe that judges are permitted to deviate
only when necessary to avoid extreme injustices. In part II, I look for
arguments to support these positions that are consistent with my find-
ings in part I. As you will see, I find some of these arguments fairly
persuasive. I conclude that deviating in every suboptimal-result case

1. Frederick Schauer, Playing by the Rules: A Philosophical Examination of Rule-Based

Decision-Making in Law and in Life (Oxford: Oxford University Press, 1991), p. 202.

179
180 Limits of Legality: Part II

constitutes an impermissible practice in realistic legal systems. That

should surprise no one who believes in the rule of law. But I also con-
clude that judges are morally permitted to deviate in some suboptimal-
result cases. Even in reasonably just legal systems there could, in
principle, be cases in which judges were morally permitted to deviate
in order to avoid results that were only moderately suboptimal. This
conclusion is unorthodox. So is my argument. I contend that the per-
missibility of deviation in a given case turns on some systemic factors
that are ordinarily seen as irrelevant to adjudication.
11

Systemic Effects

This chapter introduces some new arguments for the thesis that judges
have all-things-considered reasons to obey nonpermissive rules. These
arguments appeal to the systemic effects of deviating from the law:
effects on individuals other than parties to the case. Systemic-effects
arguments belong to nonideal theory—the study of how to act when
other agents do not act as they should.1 In this chapter, and in the rest
of part II, I shall explore the powers and limitations of systemic-effects
arguments.
First, a few words about systemic effects. We can distinguish three
aspects of any court decision: its content, its actual legality, and its per-
ceived legality. A decision has different effects in virtue of different aspects.
The content of a decision has direct effects on the parties, of course. Civil
damage awards redistribute money from defendants to plaintiffs, sentenc-
ing decisions send convicts to prison, et cetera. Losing money and being
incarcerated are disadvantageous whether or not the court’s decision was
legally correct and whether or not anyone perceives the court’s decision
as legally correct.2
Equally important for my purposes is whether observers classify a
decision as legally correct or incorrect. This classification has effects—
systemic effects—that extend beyond the parties to the case. I shall argue
that the systemic effects of deviation constitute important pro tanto moral
reasons to adhere to the law—reasons that apply in both optimal-result
and suboptimal-result cases.

1. The term comes from John Rawls, A Theory of Justice (Cambridge, Mass.: Harvard
University Press, 1971), pp. 8–9, 245–46, 351. It is also known as partial-compliance theory.
See Tamar Schapiro, “Compliance, Complicity, and the Nature of Nonideal Conditions,”
Journal of Philosophy 100 (2003): 329–55; Liam B. Murphy, Moral Demands in Nonideal
Theory (Oxford: Oxford University Press, 2000); George Sher, Approximate Justice: Studies
in Non-Ideal Theory (Lanham, Md.: Rowman & Littlefield, 1997).
2. The content of a decision also has effects on individuals who are affiliated with the
parties: friends, family members, business associates, et cetera. Again, what matters here
is content, not legality or perceived legality. Losing one’s father or business partner to
prison is disadvantageous whether or not the decision is, or is perceived to be, legally
correct. See Jeffrey Brand-Ballard, “Innocents Lost: Proportional Sentencing and the
Paradox of Collateral Damage,” Legal Theory 15 (2009): 65–105.

181
182 Limits of Legality: Part II

For your sake and mine, I wish my arguments were simple, decisive, and
proceeded from widely accepted premises. Sadly, that is not the case. The
arguments are complicated, unfamiliar, and far from conclusive. They
employ philosophical resources and arguments that remain controversial
in the literature. Persuading you to embrace these resources and argu-
ments would require wandering into these debates and take me too far
afield. I cite defenses of these positions in the literature, but these defenses
are not dispositive.
In addition to resting upon controversial philosophical premises,
systemic-effects arguments rely upon empirical hypotheses about
complex social facts, such as the extent to which judges actually imitate
the deviational tendencies of their colleagues. Reliable data supporting
or challenging these hypotheses would be difficult to collect. I cite the
studies I have found, but these are few, speculative, and not directly on
point.3
I shall not try to prove the philosophical and empirical premises of
systemic-effects arguments. Rather, I shall show how these premises could
be used to support a nonpermissive adherence rule. I anticipate that some
readers are so committed to such a rule that, in the absence of other
sound arguments for it, they will come to look more favorably upon pre-
mises that can be used to support it—even premises that seem dubious at
first.
The arguments needed to support a nonpermissive rule are quite dif-
ferent from those criticized in part I. They do not support the undermin-
ing principle. Rather, they support pro tanto reasons to adhere in
suboptimal-result cases—reasons that compete with judges’ natural rea-
sons to avoid suboptimal results. These arguments are prospective4 and
agent-neutral, not retrospective and agent-relative. In subsequent chap-
ters, I shall explain why these distinctions matter.

11.1 GOLDMAN ON SYSTEMIC EFFECTS

Alan Goldman presents a systemic-effects argument that, if sound, pro-

vides a reason for a judge to adhere to the law even if she correctly believes
it to require a suboptimal result. It is somewhat ironic that Goldman
should be the author to provide such a promising argument for obeying a

3. See §14.1.
4. Prospective arguments are often called consequentialist arguments, but this usage
causes confusion, because it connotes consequentialist moral theories that hold that pro-
spective reasons are the only ultimate moral reasons. Even nonconsequentialists accept pro-
spective moral reasons: “All ethical doctrines worth our attention take consequences into
account in judging rightness. One which did not would simply be irrational, crazy.” Rawls, A
Theory of Justice, p. 30.
Systemic Effects 183

nonpermissive rule. He is not especially enthusiastic about rules.5 As a

legal philosopher, he opposes formalism: the view that judges must obey
first-order rules of law.6 As a moral philosopher, he defends weak particu-
larism.7 Nevertheless, he faults strong particularists for ignoring “situa-
tions that do call for following rules even when ordinary judgment
opposes them. . . .” 8 He thinks there are special contexts in which agents
should, indeed, adopt strong rules. These include cases involving “individ-
ually harmless acts that have cumulatively harmful results.” 9 His para-
digm is, conveniently, judges deviating in suboptimal-result cases. He
refers to “the fundamental but unwritten rule of the legal system: that
judges must defer to legal requirements even when they disagree morally
with their implications in particular cases.” 10 He states that we want
judges “to accept a rule as morally binding that they will follow settled
law in their decisions, even when they disagree with the outcomes.”11 I
read this as claiming that judges must obey restrictive rule. In fact, as do
most writers on the topic,12 Goldman takes for granted that judges must
obey restrictive rule.13 He sets himself the task of explaining why they
must, but he never entertains the possibility that they might not have this
obligation at all.
First, let us consider Goldman’s defense of restrictive rule, beginning
with his description of a suboptimal-result case (this was my model for
Yasmin’s case in §5.3):
[C]onsider the specific case of a bank seeking to foreclose on a house and evict
a poor, elderly widow. The additional assets to the bank will have a negligible
effect on its overall financial position, while the widow will suffer greatly if
evicted. Similarly, a single court’s decision on moral merits instead of law will
have little effect on the stability or predictability of the legal system.

Using this case, Goldman defends restrictive rule, describing it as a

“second-best” strategy:
The problem is that the cumulative effects of many judges reasoning only
on these grounds could be disastrous to the legal and financial institutions
(also, in our example, to the ability of widows to obtain loans). This special
fallibility, the inability in the absence of a rule to take account of overall
effects in the single case, together with the fact that morally minded judges

5. Alan H. Goldman, Practical Rules: When We Need Them and When We Don’t
(Cambridge: Cambridge University Press, 2002), p. 36 (“[T]he adoption of rules in nor-
mal contexts represents a suspect and defeasible strategy”).
6. Ibid., ch. 3.
7. “[T]he justification for and coherence of particular judgments do not require their
support by universal rules.” Ibid., p. 2; also ch. 4.
8. Ibid., p. 3.
9. Ibid., p. 5.
10. Ibid., p. 8.
11. Ibid., p. 42.
12. See those quoted in §8.4.
13. But see §14.3.
184 Limits of Legality: Part II

will be tempted to bypass law on moral grounds in such cases, justifies the
imposition of a rule requiring a legal decision according to law and not
unfettered moral perception.14

More generally:
[J]udges may correctly perceive that they could do better morally by fol-
lowing their own moral perceptions in particular cases in which legal norms
require different decisions. But the cumulative effect of allowing such judg-
ments would be to destroy the legal authority of legislators and of the law
itself. Hence the rule that removes the authority of judges to decide cases
on direct moral grounds is both itself a moral rule and a cornerstone of the
legal system.15

I think Goldman outlines an argument with the right structure, but it

moves too quickly. In the next several sections I shall unpack his argu-
ment. I introduce some intermediate steps that he omits and rebut some
objections that he neglects.
First, Goldman’s argument requires some philosophically controversial
premises that he never states, much less defends. I shall draw out and
examine these premises. I am prepared to grant them, but doing so has
implications that we must eventually recognize.16
Second, Goldman’s systemic-effects argument as presented in his book,
Practical Rules, does not actually support the conclusion that judges must
obey an adherence rule. At most it supports the conclusion that lawmakers
should promulgate one. I present a supplemental argument that fills this
gap, supporting the conclusion that judges must obey an adherence rule.17
Adopting this argument, however, makes restrictive rule more difficult to
defend over moderate rules.
Third, Goldman’s systemic-effects argument by itself supports a non-
permissive rule but not necessarily restrictive rule. It leaves various mod-
erate rules in the running. Goldman notices this. In fact, he acknowledges
that a moderate rule initially appears to be optimal.18 Then he argues that
all moderate rules are flawed, leaving restrictive rule as the only remaining
option.19 I shall challenge his arguments. I think there is a moderate rule
to which his objections do not apply. Ultimately, I shall conclude that

14. Goldman, Practical Rules, p. 43.

15. Ibid.
16. See chapter 14.
17. In a subsequent paper, Goldman presents his own argument for a similar conclu-
sion, not pertaining to judges. Alan H. Goldman, “The Rationality of Complying with Rules:
Paradox Resolved,” Ethics 116 (2006): 453–70. Goldman, however, supports a counterpart
to restrictive rule that I reject. I discuss his paper in chapter 13.
18. “The optimal pattern would consist in some percentage of widows spared just below
the threshold at which cumulative damage to the legal system begins to outweigh the fur-
ther good to individual widows.” Goldman, Practical Rules, p. 44.
19. Ibid., pp. 44–55.
Systemic Effects 185

there may be a sound systemic-effects argument for moderate rules, but

not for restrictive rule.20
In discussing systemic effects it is important to remember that we are
seeking objective reasons for a judge to adhere when she correctly deter-
mines that the law requires a suboptimal result. We are imagining a judge
who deviates for moral reasons and advances a moral purpose as she sees
it. We must argue that deviation has systemic effects that compromise
other moral purposes to a greater degree and that these moral purposes
also provide the judge with reasons. There is no point in trying to persuade
a judge who is an avid rifle hunter to adhere to a gun-control law in the
case at bar by pointing out that deviating will encourage the sport of hunt-
ing. Instead, one must argue that deviating will compromise other values
to which the judge herself is committed. Ideally, the proponent of adher-
ence would argue that deviation is somehow self-defeating for the judge
given her own values. That is one of the ground rules of this dialectic.
In the rest of this chapter I shall examine systemic effects more
closely. They fall into two categories: adaptation effects and mimetic
effects. Goldman does not distinguish between these, but I shall argue
that they are not interchangeable. Mimetic effects, I shall suggest, are
much more useful than adaptation effects for defending the conclusion
that judges must obey nonpermissive rules.

11.2 ADAPTATION

First, consider the phenomenon of adaptation. When a judge deviates,

her decision can lead nonjudicial actors to adapt their behavior. Bank
sues Borrower for defaulting on her small-business loan. Judge rules for
Borrower. Practicing lawyers and scholars criticize this decision as legally
incorrect. They believe that the controlling statute entails a judgment
for Bank on the facts presented and that no higher source of law dictates
otherwise. Likewise, journalists report that Judge has disregarded the
law.
Bank officers at other banks also notice the decision. What effect will it
have on their behavior? This will depend in part on what they believe to
be Judge’s reasons for ruling in Borrower’s favor. If they conclude that
Judge simply made a random, careless mistake in applying the law, then
the effect on bank behavior will be minimal. Banks already knew that
judges make mistakes. Some mistakes favor lenders, others favor bor-
rowers. Banks have no incentive to alter their behavior provided they con-
clude that the mistake was random and does not portend an increase in
rulings unfavorable to lenders.
Suppose, instead, that bank officers reach the following conclusions:
first, that Judge’s decision evidences a trend of increasing judicial deviation

20. See chapter 13.

186 Limits of Legality: Part II

in loan-default cases in which the law requires results that judges consider
to be suboptimal; and second, that judges in their jurisdiction will often
conclude that results favoring lenders, rather than small-business owners,
are suboptimal. Rational bank officers who accept these premises will con-
clude that judges are becoming more inclined to deviate in order to reach
verdicts that favor small-business owners over lenders. Accordingly, banks
may begin to offer fewer small-business loans, and/or to offer them on less
favorable terms.
This is a simple argument from economic incentives. It may represent
a sound argument for lawmakers to promulgate a rule that requires a
judgment for Bank. The argument requires some unstated normative
and empirical premises, but they are fairly plausible ones. Later, how-
ever, I shall argue that adaptation effects cannot provide a general reason
for a judge to obey a nonpermissive rule.21

11.3 MIMESIS

In this section I introduce another kind of systemic effect, distinct from

adaptation. Whereas nonjudicial actors adapt to judicial decisions, other
judges may attempt to imitate prior decisions. The decisions of one or
more judges (anterior judges) can lead another judge (a posterior judge) to
make a decision that resembles the anterior decision(s) in some respect.
Consider judges who follow legal precedent. A judge applies a certain
legal doctrine to a certain fact pattern. A posterior judge, at the same level
or below, hears a case with a similar fact pattern and treats the anterior
judge’s decision as a reason to apply the same legal doctrine. Or a judge
interprets a legal standard in a certain way and a posterior judge treats the
anterior decision as a reason to interpret a similar standard similarly.22
Judges who follow precedent imitate one another at the first-order
level. I am interested in second-order imitation, which I call mimesis.23
Mimesis occurs when a judge imitates the way in which other judges treat
not a particular legal standard, but a whole category of legal standards. For
example, a judge engages in mimesis if she follows precedents more often,
the higher she believes to be the rate at which other judges follow prece-
dents.24 Likewise, she engages in mimesis if she adheres to statutes more
often, the higher she believes to be the rate at which other judges adhere
to statutes.
All legal standards—not just case law—can be the object of mimesis,
including constitutions, statutes, and administrative regulations. Also,

21. See §13.7.

22. I discuss precedent in chapter 16.
23. I apologize for the Greek, but if I called it imitation, then readers would assume I
meant “following precedent.”
24. The belief need not be conscious.
Systemic Effects 187

mimesis can refer to any way of treating a legal standard, not just adhering
to it. Mimesis occurs when a judge adheres to or deviates from a certain
legal standard in part because he believes that anterior judges have
adhered to or deviated from other legal standards. Furthermore, on my
definition of mimesis, the posterior judge need not consciously treat the
anterior decisions as a reason to decide as he does. All that I require for
mimesis is some kind of contributory causal link between anterior and
posterior decisions.
Mimesis is socially beneficial when anterior decisions causally con-
tribute to a posterior judge reaching an optimal result. For example, an
anterior decision to adhere in an optimal-result case might encourage a
posterior judge to adhere in another optimal-result case, one in which
he would otherwise have deviated. This is a success story for mimesis.
Mimetic failure occurs when optimal decisions by anterior judges make a
causal contribution to an outcome in which a posterior judge reaches a
suboptimal result. This happens because judges do not always agree
about which results are suboptimal. Imagine a group of judges who
deviate in all and only cases that Dave the Democrat considers to be
suboptimal-result cases. From Dave’s perspective, these are optimal
judges—Group O. There are, however, other judges in the system who
disagree with Dave about which results are suboptimal. From Dave’s
perspective, they are subpar judges. He might think them less enlight-
ened, less morally perceptive, less intelligent, less careful, or less scrupu-
lous than Group O. These subpar judges begin to notice when the
members of Group O deviate. Trying to imitate Group O, they begin
deviating in what they believe to be suboptimal-result cases, thereby
reaching results that Dave considers suboptimal. Thus, although Group
O always reaches results that Dave considers optimal, it causally con-
tributes to results that he considers suboptimal.
Of course, reasonable people disagree about which results are subop-
timal, so they disagree about who actually belongs to Group O. The
structure of the scenario just described remains the same if we substi-
tute Ruth the Republican for Dave the Democrat. Because I am not
trying to convince you that any particular result is suboptimal, I shall
simply refer to optimal and suboptimal results in the abstract. For the
sake of concreteness, I shall continue using Yasmin’s eviction as my proxy
for a suboptimal result, without presuming to have established that it is
in fact suboptimal.
If my story about Group O is realistic, then deviation, even in suboptimal-
result cases, can encourage deviation in optimal-result cases. This is a sig-
nificant claim. By deviating in an optimal-result case a judge performs the
same type of pro tanto wrongful action, and causes the same type of bad
effects, as when he adheres in a suboptimal-result case. A suboptimal
result, such as evicting Yasmin, is just as suboptimal when the law forbids
it as when the law requires it. It is no less wrongful for the judge and no
less harmful to Yasmin.
188 Limits of Legality: Part II

I shall argue that because deviating can be just as bad as adhering, the
need to prevent posterior judges from reaching suboptimal results can
constitute a pro tanto moral reason for a judge to reach a suboptimal result
in the case at bar. Mimetic failure can provide pro tanto reasons for judges
to adhere. These are, moreover, reasons of the right kind to compete with
the strong reasons that judges have to deviate in suboptimal-result cases.
In some suboptimal-result cases the former reasons win, producing an all-
things-considered reason to adhere. If there is any truth to the common
assumption that deviation, as such, “damages the rule of law,” then I think
the destructive mechanism at work must involve mimetic failure. Mimetic
failure also explains why even optimal judges inevitably confront nega-
tive-gap cases 25 if they inhabit jurisdictions in which more than a certain
fraction of the other judges are less accurate identifiers of suboptimal
results.
In the rest of this chapter and the next, I argue that the prospect of
mimetic failure gives the judges in Group O reasons to obey a nonpermis-
sive rule.

11.4 SYSTEMIC EFFECTS AS REASONS TO ADHERE

My ultimate position will be that mimetic failure provides an all-things-

considered reason to adhere in some suboptimal-result cases. However,
spelling out the argument will take much of part II. I begin by arguing
that judges in Group O have reasons to obey a nonpermissive rule. The
first obstacle to overcome is that for any given judge, the systemic effects
that she actually causes by deviating do not provide her with even a pro
tanto reason to adhere:

Individual Adherence Argument

1. If Agent B has an objective pro tanto moral reason to Φ, then Agent A has
an objective pro tanto moral reason not to act in ways that discourage B
from Φ-ing.
2. Judges have objective pro tanto moral reasons to adhere in optimal-result
cases.
3. If Judge J deviates in a suboptimal-result case, then she will discourage
some other judge from adhering in some optimal-result case.
4. Therefore, Judge J has an objective pro tanto moral reason to adhere in the
suboptimal-result case.

Here is a crude example of what the individual adherence argument

envisions. Picture a judge who initially believes that her reasons to adhere
are stronger than any reasons to deviate. Then she learns that an anterior

25. See §5.4 on gap cases.

Systemic Effects 189

judge has deviated, which leads her to revise downward her belief about
the strength of her own reasons to adhere. Perhaps she was concerned that
deviating would injure her professional reputation, but she is reassured by
learning that the anterior judge has not suffered reputational damage. Or
perhaps she previously believed that deviation was always wrong, but she
trusts the moral judgment of the anterior judge enough that his decision
changes her mind. Or perhaps the anterior decision influences her sublim-
inally without her paying it any conscious attention. Whatever her mental
process, the result is that she deviates in what she mistakenly believes to
be a suboptimal-result case. The fact that deviating in the anterior case has
this result gives the anterior judge a pro tanto reason to adhere, according
to the individual adherence argument.
The individual adherence argument is valid, but the truth of the first
and third premises depends on what it means to discourage someone from
acting. Here is a broad understanding of discourage. Agent B believes that
she has pro tanto reasons not to Φ.26 She intends to Φ, nevertheless, because
she believes that she has stronger reasons to Φ. If the actions of A cause B
to stop believing that she has an all-things-considered reason to Φ, and
this change in belief motivates B not to Φ, then A discourages B from
Φ-ing. On this understanding of discourage, A need not have the goal of
changing the beliefs or conduct of B, nor need A even be aware that his
actions will affect B. In the rest of this chapter, I shall examine further this
broad understanding of discourage.

11.5 MORAL-MORAL PRISONER’S DILEMMAS

The third premise of the individual adherence argument is empirically

false as to most suboptimal-result cases. Judge Jack’s choice to deviate in
a given case does not lead any identifiable judge, who would otherwise
have adhered, to deviate in a subsequent optimal-result case. Remember,
our concern is not that Jack’s deviation will cause other judges to decide
similar cases similarly, as when they follow precedent. Our concern is pos-
terior judges deviating in optimal-result cases as a result of Jack’s devia-
tion in a suboptimal-result case. One can agree that a pattern of widespread
deviation in suboptimal-result cases encourages deviation in optimal-
result cases and still deny that Jack’s individual choice to deviate in a single
case encourages any other identifiable judge to deviate in an optimal-result
case. So the individual adherence argument does not establish that any
individual judge always has objective pro tanto reasons to adhere.
Nevertheless, I concede the conceptual possibility of a particular judge
who, by choosing to deviate in a suboptimal-result case, provokes some
posterior judge to deviate in an optimal-result case. This would be the
case if, for example, the posterior judge so revered the anterior judge in

26. These reasons could be prudential, moral, or what have you.

190 Limits of Legality: Part II

particular that he would attempt to imitate her, but would do so incor-

rectly. The anterior judge in that case would have a pro tanto reason to
obey restrictive rule.
There is even the conceptual possibility of a “hypersensitive” legal
system in which any judge who deviates provokes deviation in an optimal-
result case. Imagine a posterior judge who is extremely aware of how all
other judges decide, extremely inclined to engage in mimesis, and an
extremely poor identifier of suboptimal results. A system with many such
judges is hypersensitive. Judges in such systems would have pro tanto rea-
sons to obey restrictive rule.
In realistic legal systems, however, few deviant decisions provoke devia-
tion by other judges in optimal-result cases. This entails that, as Goldman
observes, a judge assigned to decide a suboptimal-result case finds herself
in a special type of collective action problem: a multiplayer moral-moral
prisoner’s dilemma.27 Players in a standard prisoner’s dilemma have reasons
of self-interest to defect from the cooperative scheme, although collective
defection is collectively and individually self-defeating. Similarly, each
player in a moral-moral prisoner’s dilemma has pro tanto moral reasons to
defect, although collective defection produces a morally inferior outcome.
Anyone who believes that systemic effects give judges reasons to obey
a nonpermissive rule must identify reasons for parties to a moral-moral
prisoner’s dilemma to cooperate. Some readers will find intuitive the idea
that such reasons exist, just as some find intuitive the idea that parties to
a standard prisoner’s dilemma have reasons to cooperate.28 Those who do
not find the idea intuitive will want arguments. Several are possible. First,
one could argue that an agent who defects in a moral-moral prisoner’s
dilemma is morally responsible for more than the direct consequences of
his defection. Alternatively, one could “introduce a different notion of
rationality and of moral reasons.” 29 Goldman and Christopher Kutz take
the first path, Christopher McMahon the second.30 Other approaches are
surely possible. In the rest of this chapter I shall offer my own version of
the argument, drawing on the insights of these authors and others. My
argument is designed to describe reasons to cooperate in more detail and
to bolster the intuitions of readers who already believe in such reasons. My

27. Goldman, Practical Rules, pp. 13, 49–55. See related discussion in Garrett Cullity,
“Moral Free Riding,” Philosophy and Public Affairs 24 (1995): 3–34; Michael Otsuka, “The
Paradox of Group Beneficence,” Philosophy and Public Affairs 20 (1991): 132–49; George
Klosko, “Parfit’s Moral Arithmetic and the Obligation to Obey the Law,” Canadian Journal
of Philosophy 20 (1990): 191–214; Jean Hampton, “Free-Rider Problems in the Production
of Collective Goods,” Economics and Philosophy 3 (1987): 245–73.
28. “Intuitively, rational persons ought to be able to cooperate [in a prisoner’s dilemma].”
David Copp, “Introduction: Metaethics and Normative Ethics,” in Oxford Handbook of Ethi-
cal Theory, ed. David Copp (New York: Oxford University Press, 2006), p. 17.
29. Goldman, “The Rationality of Complying with Rules: Paradox Resolved,” p. 456.
30. Ibid.; Christopher Kutz, Complicity: Ethics and Law for a Collective Age (Cambridge:
Cambridge University Press, 2000); Christopher McMahon, Collective Rationality and
Collective Reasoning (Cambridge: Cambridge University Press, 2001).
Systemic Effects 191

argument might also persuade readers who do not share these intuitions,
but I feel little pressure to persuade them. If there are no reasons to coop-
erate in moral-moral prisoner’s dilemmas, and the arguments of part I are
sound, then judges have no all-things-considered reason to obey a nonper-
missive rule. I think even readers who find reasons to cooperate counter-
intuitive will find that conclusion still more counterintuitive. I think we
can ultimately avoid it, but only with some effort.

11.6 COLLECTIVE REASONS

As far as I know, Goldman was the first to try to use a systemic-effects

argument to solve a moral-moral prisoner’s dilemma. He uses the exam-
ple of citizens who contemplate rerouting their tax payments to charita-
ble causes. He adopts the premise that a rule requiring citizens to pay
taxes is justified by the value of the public goods provided by tax revenue.
He also assumes that “[g]iven the negligible effect of the contribution of
each on the provision of the public goods, and the amount of tax money
that must pay for government bureaucracy, it is easy to claim that each
could produce a morally better result by giving directly to a charity or
needy person.”31 Nevertheless, Goldman claims, the group of all citizens
has a “group obligation” to produce the public goods.32 This means more
than that the members of the group have individual obligations. It means
that the group as such has an obligation. I am obligated to feed my dogs
and my neighbor to feed hers, but we do not form a group with group
obligations. By contrast, my wife, my daughter, and I could have a group
obligation to feed our dogs. At least we could have such an obligation if
groups as such can have obligations.
A moral obligation is a kind of moral reason, so a group obligation is a
kind of group reason or (synonymously) collective reason. The question
whether groups can have reasons is prior to the question whether groups
can have obligations. Hereafter, when I refer to group reasons or group
obligations I shall mean reasons or obligations of the group as such.
If Group O can have reasons, then we can modify the individual adher-
ence argument to apply to Group O:

Group Adherence Argument

1. If Agent B has an objective pro tanto moral reason to Φ, then Agent A has
an objective pro tanto moral reason not to act in ways that discourage B
from Φ-ing.
2. Judges have objective pro tanto moral reasons to adhere in optimal-result
cases.

31. Goldman, “The Rationality of Complying with Rules: Paradox Resolved,” p. 454.
32. Ibid., p. 463.
192 Limits of Legality: Part II

3. By deviating in all suboptimal-result cases, Group O will discourage

some other judges from adhering in some optimal-result cases.
4. Therefore, Group O has an objective pro tanto moral reason not to
follow a policy of deviating in all suboptimal-result cases
(i.e., a reason to obey a nonpermissive rule).
If Goldman is correct, then something like the group adherence argu-
ment has potential. However, he says virtually nothing about the condi-
tions under which groups have reasons. In the next sections I shall describe
these conditions in greater detail. Then we can see how Group O satisfies
them.

11.7 COLLECTIVE REASONS AND SHARED INTENTIONS

A theory of collective reasons specifies conditions under which a group of

human beings, as such, has reasons. There is a substantial literature on
collective reasons, most of which concentrates on group obligations. For
the group adherence argument to succeed, some theory of collective rea-
sons must be correct. I expect that any of several theories could support
the argument. I shall not defend the superiority of any one theory, but
shall present an influential one that serves my purposes. This theory pro-
ceeds from the assumption that a group as such can have intentional
states, namely intentions. Some readers will question this assumption, rea-
soning as follows. Two human beings can share a single intentional state
only if they share a single mind. It is metaphysically impossible for two
human beings to share a single mind. Therefore, two human beings can-
not share a single intentional state.
The major premise of the preceding argument has been refuted by
Michael Bratman, Margaret Gilbert, David Velleman, John Searle, Carol
Rovane, Raimo Tuomela, and others. They advance conceptions of shared
intentional states that do not presuppose the existence of any kind of
metaphysically problematic group mind.33 Bratman’s theory of shared
intention illustrates. Bratman takes a functional approach, asking what
roles shared intentions play in our lives. He argues that shared intentions
serve three main functions: they help us to coordinate our intentional
activities, facilitate the coordination of further planning, and provide a

33. See, e.g., Carol Rovane, The Bounds of Agency (Princeton, N.J.: Princeton University
Press, 1998); David Velleman, “How to Share an Intention,” Philosophy and Phenomenological
Research 57 (1997): 29–50; Margaret Gilbert, Living Together: Rationality, Sociality, and Ob-
ligation (Lanham, Md.: Rowman & Littlefield, 1996); Raimo Tuomela, The Importance of Us
(Stanford, Calif.: Stanford University Press, 1995); Michael E. Bratman, “Shared Intention,”
Ethics 104 (1993): 97–113; John R. Searle, “Collective Intentions and Intentional Actions,”
in Intentions in Communication, eds. Philip R. Cohen, Jerry Morgan, and Martha E. Pollack
(Cambridge, Mass.: MIT Press, 1990); Margaret Gilbert, On Social Facts (Princeton, N.J.:
Princeton University Press, 1989).
Systemic Effects 193

background framework that structures bargaining. On Bratman’s view,

two individuals share an intention if and only if each has certain attitudes
that are specially interrelated with those of the other.34 A shared inten-
tion to paint the house might form in the following way. I say, “I’ll paint
the house with you if you’ll do the same.” You reply, “Then I’ll do so.” At
this point each of us has the intention that we paint the house together.
Furthermore, each of us intends that we paint the house in accordance
with and because of his own intentions and those of the other. Each of us
intends that the intentions of the other, along with his own, be jointly
efficacious. All of this is common knowledge between us. I know that
you know, you know that I know that you know, and so on.35 Here is
Bratman’s full analysis of shared intention:

We intend to J if and only if

1. (a) I intend that we J and (b) you intend that we J.
2. I intend that we J together in accordance with and because of 1a, 1b, and
meshing subplans of 1a and 1b; you intend that we J in accordance with and
because of 1a, 1b, and meshing subplans of 1a and 1b.
3. 1 and 2 are common knowledge between us.36

Bratman’s theory shows how the attitudes of individuals can constitute

a shared intention, notwithstanding the fact that an individual seems only
to be able to intend that she herself perform an action, not that anyone
else do so.37 Bratman appeals to the fact that we have the capacity not
only to intend to act but also to intend that some state of affairs obtain.
He proposes that we can intend a joint action just in the sense that we can
intend that we perform the joint action. Bratman acknowledges, however,
that such intentions are subject to an “influence condition.” 38 In other
words, I can intend that we jointly act only if I see myself as affecting both
the fact that you play your role in our joint action and the manner in
which you do so. I must believe that you will play your role in our activity
in part because of and in accordance with my intention that we act.
Velleman has argued that we can understand our two intentions as com-
bining to form a single joint intention—an intentional state that motivates
each of us precisely by virtue of representing itself as having this

34. This is not to suggest that shared intentions die with the brains of the human beings
who share them, any more than an individual intention dies with the brain of the human
being who has it.
35. For this sense of common knowledge see David Lewis, Convention (Cambridge, Mass.:
Harvard University Press, 1969).
36. Bratman, “Shared Intention,” p. 106.
37. For a different solution see Raimo Tuomela and Kaarlo Miller, “We-Intentions,” Phil-
osophical Studies 53 (1988): 115–37.
38. Michael E. Bratman, “I Intend that We J,” in Contemporary Action Theory, eds. Ghita
Holmström-Hintikka and Raimo Tuomela (Netherlands: Kluwer, 1997).
194 Limits of Legality: Part II

motivational efficacy.39 In this way we can understand the idea of two

human beings sharing a single token intention.
I find it both convenient and accurate to refer to a group as a collective
agent 40 if its members share an intention. A collective agent is a single
agent composed of two or more different human beings yet irreducible to
its constituents.41 Many have argued, and I agree, that a collective agent
can have its own reasons and obligations.42

11.8 OPTIMIZING COLLECTIVE AGENTS

For my purposes, two judicial intentions are especially important: to

adhere to the law and to avoid reaching results that one considers subop-
timal. Most judges in realistic legal systems have both. Only the most
incompetent or corrupt judge is indifferent either to the law or to the
optimality of results. A judge could, of course, find herself with desires
that conflict with these intentions. She could reconsider and/or revise
either or both of these intentions. But a typical judge maintains both of
these intentions throughout her career.
Some will claim that every judge is morally obligated to form these
intentions before hearing her first case and to maintain and fulfill them
while she remains on the bench. I think this is probably correct. Such
obligations can probably be derived from natural duties, combined with
the judge’s decision to remain a judge and the situation in which she finds
herself. However, my argument requires no defense of this position. I can
limit myself to judges who actually have these intentions, whether or not
they are morally obligated to form them. My argument depends only on
the premise that these intentions generate reasons for action.

39. Velleman, “How to Share an Intention.”

40. Also known as a collective subject, plural subject, group person, group agent, et
cetera.
41. See, e.g., Margaret Gilbert, Sociality and Responsibility: New Essays in Plural Subject
Theory (Lanham, Md.: Rowman & Littlefield, 2000); Rovane, The Bounds of Agency; Tuomela,
The Importance of Us; Philip Pettit, The Common Mind: An Essay on Psychology, Society, and
Politics (Oxford: Oxford University Press, 1993). The terminology of agency, although use-
ful, can mislead. The most familiar agents are individual human beings. They possess the
highest degree of moral status. They develop extremely complex psychologies as they
mature. Most collective agents, by contrast, are psychologically rudimentary and possess
only a low degree of moral status, if any. Consider a group that shares only a house-painting
intention. It constitutes an agent only in the most minimal sense. It does not enjoy any moral
rights beyond those of its individual members. When I refer to a group as an agent I do not
mean to imply that it possesses anything approaching the psychological complexity or moral
status of a typical adult.
42. See, e.g., Philip Pettit, “Responsibility Incorporated,” Ethics 117 (2007): 171–201;
Tracy Isaacs, “Collective Moral Responsibility and Collective Intention,” Midwest Studies
in Philosophy 30 (2006): 59–73; Abraham Sesshu Roth, “Shared Agency and Contralat-
eral Commitments,” Philosophical Review 113 (2004): 359–410; Gilbert, Sociality and
Responsibility.
Systemic Effects 195

In addition to forming these individual intentions, the members of a

group of judges could form shared intentions with similar content. Con-
sider a shared intention to avoid suboptimal results. On Bratman’s theory,
Judges Jack and Jill share this intention if and only if:

(1)(a)(i) Jack intends that Jack and Jill avoid suboptimal results.
(1)(a)(ii) Jack intends that Jack and Jill avoid suboptimal results in accordance
with and because of meshing subplans of (1)(a)(i) and (1)(b)(i).
(1)(b)(i) Jill intends that Jack and Jill avoid suboptimal results.
(1)(b)(ii) Jill intends that Jack and Jill avoid suboptimal results in accordance
with and because of meshing subplans of (1)(a)(i) and (1)(b)(i).
(2)It is common knowledge between Jack and Jill that (1).

If Jack and Jill share this intention, then they constitute an optimizing
collective agent. Notice that they cannot share it unless they intend to
mesh subplans.43 How well their subplans mesh will depend upon the
extent to which they share judgments about which results are optimal.
The more they disagree about which results are optimal, the more their
subplans conflict. So they cannot form an optimizing collective agent un-
less they share many, perhaps most, optimality judgments. If they share
enough optimality judgments, then they can form interlocking intentions
and meshing subplans.
If Jack and Jill share both the intention to avoid suboptimal results
and enough optimality judgments, then Jill encourages and helps Jack to
deviate in suboptimal-result cases and to adhere in optimal-result cases,
at least when she can do so without compromising more important
goals. One way in which Jill encourages Jack is by deviating herself in
suboptimal-result cases and adhering in optimal-result cases. She con-
siders her effects on Jack to be, at least, welcome side effects of her
decisions. Likewise for Jack, mutatis mutandis. Each sees deviation by
the other in suboptimal-result cases as partial fulfillment of the shared
intention of the optimizing collective agent to which they belong. Nei-
ther Jack nor Jill condemns, discourages, or disavows such deviation by
the other.
In the remainder of this book I shall present arguments requiring at
least one of the following premises: (1) some judges in realistic legal
systems actually form optimizing collective agents; or (2) judges in realis-
tic legal systems are morally obligated to form optimizing collective agents.
I think both premises are true. Some readers will disagree. Some will reject
the very idea of shared intentions. I shall not directly defend either pre-
mise or offer much more in defense of the underlying theory of collective
agency. Rather, I shall advance indirect arguments. I claim that despite the
arguments of part I, accepting one of these premises allows one to defend

43. A subplan is undertaken in order to advance another intention: a subplan to buy

eggs is undertaken in order to advance the intention to make an omelet.
196 Limits of Legality: Part II

the widely held conviction that judges must obey a nonpermissive rule.
Perhaps these premises are even necessary to the task, although I shall not
argue for this stronger claim. Part I found no individual moral reasons for
judges to adhere in any suboptimal-result case. Part II, by contrast, argues
that an optimizing collective agent could have collective reasons to adhere
in certain suboptimal-result cases. Readers who are moved by part I, but
who still believe that judges have all-things-considered reasons to obey a
nonpermissive rule, should be motivated to conclude either that judges
actually form optimizing collective agents or that they are morally obli-
gated to do so. That is my inspiration for part II.
In the rest of this section I shall clarify some points about optimizing
collective agents. A reader who accepts the Bratman-Velleman theory of
shared intention might object that judges never explicitly communicate
pledges to one another that would form the constitutive, shared inten-
tions of optimizing collective agents. But the Bratman-Velleman theory
does not require explicit exchanges. Explicit promises just illustrate how
shared intentions might form. In many real-life scenarios shared inten-
tions arise without any explicit exchange of promises.44 Judges need not
make explicit pledges to form optimizing collective agents.45
Any realistic judiciary will contain multiple optimizing collective
agents. One might wonder if a single judge could belong to more than
one. After all, it is usually possible for one person to belong to multiple

44. When the chair of a philosophy department offers you a position, part of the mean-
ing of her offer is, in effect, “We’ll treat you as a colleague, educate students, and advance the
discipline, so long as you proceed to do the same.” When you accept the offer, part of what
you implicitly communicate is that you will do the same, on the same conditions, mutatis
mutandis. These messages constitute part of the unspoken background of the social context
of academic job offers. This background is so taken for granted that none of the participants
may ever consciously reflect on it. But counterfactuals immediately reveal its presence. If a
member of the department begins giving trombone lessons in the middle of what is sup-
posed to be an introductory logic class, or begins publishing only detective novels, everyone
will understand him to have contravened the department’s constitutive shared intentions to
educate undergraduates, advance the discipline, and so forth. I suggest that this is how it
happens with Group O, as well. See Lewis, Convention; Gilbert, On Social Facts.
45. For convenience, I shall refer to the members of an optimizing collective agent as
judges, but this usage is not strictly accurate. The members of an optimizing collective agent
share optimality judgments, but a judge can change her mind over time. She exits an opti-
mizing collective agent when she ceases to share enough of its optimality judgments. Then
she can join another optimizing collective agent whose judgments she comes to share. Con-
sider a particular holding on a legal issue: that §271 of the Telecommunications Act of 1996
was not an unconstitutional bill of attainder, as applied to the Bellsouth Corporation in
1998. Judge Rhonda might believe in 2001 that this was a suboptimal result, but change her
mind in 2002 without ever changing her mind about the legally correct result. (Perhaps she
studies more economics in the interim.) The judge-segment, Rhonda-in-2001, is a member of
a group who believes that the holding was suboptimal, whereas Rhonda-in-2002 is not a
member thereof. So an optimizing collective agent is really composed of judge segments,
each of which comprises a judge during a certain finite time period. A judge is simply a judge
segment that happens to endure for an entire judicial career. Hereafter, for simplicity, I shall
refer to the members of groups as judges, but I really mean judge segments.
Systemic Effects 197

collective agents: one woman can be an Episcopal deacon, an FBI agent,

and an actor portraying Lady Macbeth. However, this is possible only
because these collective agents—the Episcopal Church, the FBI, the cast
of Macbeth—have substantially compatible, although largely nonoverlap-
ping, subplans. By contrast, because the members of one optimizing col-
lective agent share optimality judgments that differ from those of other
optimizing collective agents, the subplans of different optimizing collec-
tive agents always clash. Therefore, it is best to think of each judge as
belonging to one at a time.
Several legal philosophers have used theories of shared intention to
bring greater precision to the social fact thesis at the heart of positivism,
the thesis that “the possibility of legal authority is to be explained not in
terms of substantive morality, but, rather, in terms of certain social
facts.” 46 Scott Shapiro, Jules Coleman, and Christopher Kutz have all
adapted Bratman’s theory for this purpose.47 Matthew Noah Smith has
argued, against these three, that most actual public officials do not satisfy
Bratman’s requirements for joint intentional activity.48 Although I agree
with Shapiro and company that public officials can share intentions,
I take no position in this book on whether shared intentions are at the
foundations of law. I need not claim that all judges participate in a single
joint intentional activity. I plan to use shared intentions for a more limited
purpose—to explain how a group of judges, as such, can have practical
reasons.

11.9 GROUP O’S REASONS

We can define Group O as the optimizing collective agent whose mem-

bers share what we are assuming arguendo to be correct optimality judg-
ments. Different people will disagree about which group of judges is, in
fact, Group O, but that is not important here. The intention to avoid
suboptimal results gives Group O a pro tanto reason to deviate in all sub-
optimal-result cases. But Group O may also have reasons to adhere in
suboptimal-result cases. If Group O is large enough and deviates in enough
suboptimal-result cases (the anterior set), then it causes other judges to
deviate in certain optimal-result cases (the posterior set). This causal rela-
tionship could, in theory, be expressed as a mathematical function. Begin
by defining a judicial entity as an individual judge, a set of judges, a single

46. Jules Coleman, The Practice of Principle (Oxford: Oxford University Press, 2001),
p. 75. Coleman notes that “no claim is more central to legal positivism.”
47. Ibid., pp. 96–99; Christopher Kutz, “The Judicial Community,” Philosophical Issues
11 (2001): 442–69; Scott J. Shapiro, “Law, Plans, and Practical Reason,” Legal Theory 8
(2002): 387–441.
48. Matthew Noah Smith, “The Law as a Social Practice: Are Shared Activities at the
Foundations of Law?” Legal Theory 12 (2006): 265–92.
198 Limits of Legality: Part II

court,49 a court system,50 a geographic region,51 or something similar.

Next, define the deviation rate of a judicial entity over a specified period
of time as the fraction of cases in which deviation occurs relative to the
total number of cases decided by the entity during that period. In realistic
legal systems the function is probably monotonic: higher deviation rates
in anterior suboptimal-result cases correlate with higher deviation rates in
posterior optimal-result cases. If Group O deviates in all suboptimal-
result cases, then posterior judges try to imitate Group O. Not being
members of Group O, however, the posterior judges mistakenly deviate in
some optimal-result cases.
The causal relationship just described is expressed in the third premise
of the group adherence argument.52 I think the third premise is true in
realistic legal systems. However, as an empirical hypothesis it is not true
in all conceivable legal systems. There is a conceivable system in which
Group O deviates in all suboptimal-result cases yet no systemic damage
results. Imagine systems in which a very small fraction of the cases are
suboptimal-result cases, or in which judges outside Group O are few, or
rarely attempt to imitate judges in Group O. In such systems no mimetic
failure would occur, so mimetic failure would never give Group O even a
pro tanto reason to adhere in suboptimal-result cases. In such “hyposensi-
tive” systems, Group O would not be morally obligated to obey a nonper-
missive rule. It could permissibly obey an adherence rule as weak as
permissive rule.
If actual legal systems fit the foregoing description, then the third pre-
mise of the group adherence argument is false and my arguments in part
I support the objective moral permissibility of obeying permissive rule:
deviating in all suboptimal-result cases. However, I do not believe that
actual legal systems have any of these attractive features. In modern legal
systems there are many suboptimal-result cases. There are also many
judges who are outside Group O and who are vulnerable to mimetic
failure.53

11.10 DEVIATION DENSITY

We should also keep in mind the likelihood that the deviation rate in
posterior optimal-result cases is a function of more than the anterior de-
viation rate. The deviation density of a judicial entity is defined as the

49. The Supreme Court of the United States, the Federal District Court for the Southern
District of New York, the United States Court of Appeals for the Seventh Circuit, et cetera.
50. For example, the state courts of Ohio.
51. For example, every court in New England.
52. See §11.6.
53. Again, you and I need not agree about which results are suboptimal or which judges
belong to Group O.
Systemic Effects 199

modal deviation rate of all its subsidiaries.54 I think the density as well as
the rate of anterior deviation matters. The reason is variable salience. Sup-
pose one mixes a milliliter of red dye uniformly into a swimming pool.
The naked eye cannot detect the dye. But that same milliliter of dye is
easily visible if sealed in a tiny plastic bag in the middle of the pool. The
mimetic effects of a set of deviant decisions depends on how evenly dis-
tributed they are on various dimensions that affect salience, such as time,
jurisdiction, legal issues involved, et cetera. A set of deviant decisions by
the same court, or issued within a short period of time, will usually receive
more attention than the same set of decisions issued by different courts,
or over longer periods of time. A set of deviant decisions on the same or
related legal issues will usually receive more attention than a set contain-
ing the same number of equally deviant decisions on a diversity of unre-
lated legal issues. Therefore, I shall treat deviation density, rather than
mere deviation rate, as the relevant variable in the rest of this book.

11.11 INTRANSITIVITY AND SLIPPERY SLOPES

The mimetic-failure argument will remind some readers of slippery-slope

arguments. The parallel merits attention because such arguments are
familiar and often used fallaciously.55 In the words of coauthors Mario J.
Rizzo and Douglas Glen Whitman, a slippery-slope argument has three
components:
1. An initial, seemingly acceptable argument and decision
2. A “danger case”—a later argument and decision that are clearly
unacceptable
3. A “process” or “mechanism” by which accepting the initial argu-
ment and making the initial decision raise the likelihood of accept-
ing the later argument and making the later decision 56
In the mimetic-failure argument, the “initial, seemingly acceptable argu-
ment” is an argument addressed to Group O that runs as follows: (a) you
have all-things-considered reasons to avoid reaching suboptimal results;
(b) the law requires you to reach a suboptimal result in these cases;
(c) therefore, you have all-things-considered reasons to deviate in these
cases.

54. David Lyons uses the term density similarly in Forms and Limits of Utilitarianism
(Oxford: Clarendon Press, 1965), p. 72.
55. For discussion of valid and invalid uses see Eugene Volokh, “The Mechanisms of the
Slippery Slope,” Harvard Law Review 116 (2003): 1026–137; Eric Lode, “Slippery Slope
Arguments and Legal Reasoning,” California Law Review 87 (1999): 1469–543; Douglas
Walton, Slippery Slope Arguments (Oxford: Clarendon Press, 1992); Frederick Schauer, “Slip-
pery Slopes,” Harvard Law Review 99 (1985): 361–83.
56. Mario J. Rizzo and Douglas Glen Whitman, “The Camel’s Nose Is under the Tent:
Rules, Theories, and Slippery Slopes,” UCLA Law Review 51 (2003): 539–92, p. 544.
200 Limits of Legality: Part II

The “seemingly acceptable decisions” are the decisions by the members

of Group O to deviate in these cases. These decisions are acceptable
because ex hypothesi the cases are suboptimal-result cases.
The mimetic-failure argument also identifies a “danger case,” but it
does not necessarily fit Rizzo and Whitman’s description. In the mimetic-
failure argument the danger case involves posterior judges outside Group
O deciding to deviate in optimal-result cases, but the mimetic-failure ar-
gument does not require that the judges “accept an argument” that per-
suades them to do so. The posterior judges need not even be consciously
aware of the influence of Group O. However, as do Rizzo and Whitman,
the mimetic-failure argument posits a “process” or “mechanism” by which
Group O’s acceptance of the initial argument raises the likelihood that
posterior judges will deviate in optimal-result cases. The members of
Group O deviate in all suboptimal-result cases. Their deviation is recog-
nized by some observers (e.g., journalists, colleagues, scholars, lawyers,
members of the general public), but they receive, at most, only mild
public criticism. Some posterior judges who notice these developments
come to believe that they could deviate under similar conditions without
receiving any greater degree of criticism than Group O received. As a
result they proceed to deviate in cases that they believe to be suboptimal-
result cases and that they believe to be legally and factually similar to those
in which Group O deviated. For the sake of the story, imagine that the
first generation of posterior judges chooses to deviate in cases that happen
to be genuine suboptimal-result cases. Then the pattern recurs. A second
generation of posterior judges notices that the first generation deviated.
The second generation concludes that they could do the same without
suffering more criticism. So they proceed to deviate in cases that they
believe to be suboptimal-result cases, and that they believe to be legally
and factually similar to those in which the first generation deviated, and
so on.
The process just described could culminate in judges deviating in opti-
mal-result cases, even if every judge in the system is sincerely attempting
to imitate either Group O or judges who have themselves attempted to
imitate Group O. This can happen because the relation of similarity is
intransitive: Case A resembles B in the relevant respects, which resembles
C in the relevant respects, . . ., which resembles Y, which resembles Z, but
Z does not resemble A in the relevant respects.57 To be more concrete,
imagine a series of eviction cases, beginning with Yasmin’s. In each succes-
sive case the legally required result is less suboptimal than in the previous
case. As we move through the series, the defendants become increasingly
responsible, morally speaking, for their circumstances, whereas the plain-
tiffs become increasingly entitled, morally speaking, to have the corre-
sponding defendant evicted. Any two adjacent cases are legally and
factually similar to one another, but the final case in the series is legally

57. See Walton, Slippery Slope Arguments, p. 131.

Systemic Effects 201

and/or factually dissimilar to Yasmin’s case. In the final case in the series
we might find that the defendant is a lazy, middle-aged woman who could
easily afford rent whereas the plaintiff is a hardworking landlord who
needs the rent in order to support his own modest lifestyle. This is not a
suboptimal-result case at all. But a judge could end up deviating in this
case because he thinks it resembles previous suboptimal-result cases in
which judges have properly deviated.
Mimetic failure occurs because each successive generation of cases
chosen for deviation bears less and less resemblance to the original cases
with respect to the legal and factual features that justify deviation. At
some point, the danger case or “slippery-slope event” occurs: a judge devi-
ates in a case that resembles a case that resembles a case . . . that resembles
the original suboptimal-result cases, but the new case is actually an opti-
mal-result case. Because the similarity relation between cases is intransi-
tive, Group O’s deviation in suboptimal-result cases provokes mimetic
failure even if all posterior judges sincerely attempt to imitate Group
O—by deviating only in cases that they consider relevantly similar to
those in which Group O would deviate.
12

Agent-Relative Principles

12.1 DOES MIMETIC FAILURE GIVE GROUP O

REASONS TO ADHERE?

Recall the group adherence argument:

1. If Agent B has an objective pro tanto moral reason to Φ, then Agent
A has an objective pro tanto moral reason not to act in ways that
discourage B from Φ-ing.
2. Judges have objective pro tanto moral reasons to adhere in optimal-
result cases.
3. By deviating in all suboptimal-result cases, Group O will
discourage some other judges from adhering in some optimal-
result cases.
4. Therefore, Group O has an objective pro tanto moral reason not
to follow a policy of deviating in all suboptimal-result cases (i.e., a
reason to obey a nonpermissive rule).
One defining feature of the group adherence argument is that it assigns
to Group O some degree of moral responsibility for the decisions of other
judges. Applied to Group O, the first premise imputes to the group a col-
lective intention to discourage judges outside the group from reaching
suboptimal results. If Group O has this intention, then it has a pro tanto
reason to obey a nonpermissive rule.
Given Group O’s intention to avoid suboptimal results, one might
readily assume that it also has the intention to discourage other judges
from reaching suboptimal results. But the former intention does not
logically entail the latter. Group O could take a much more limited
view of its own moral responsibilities. It could regard itself as responsi-
ble for avoiding suboptimal results in its own decisions, while bearing
no responsibility whatsoever for discouraging other judges from reach-
ing suboptimal results. But the important question is not whether
Group O believes itself to be responsible for decisions it provokes other
judges to make. The question is whether Group O really is responsible
for such decisions. Does the fact that by deviating in suboptimal-result
cases, Group O encourages other judges to deviate in optimal-result

202
Agent-Relative Principles 203

cases give Group O pro tanto reasons to adhere?1 We must answer this
question in order to determine which adherence rule Group O has
an all-things-considered reason to obey (i.e., how often Group O may
deviate).
The answer depends on three factors: (1) the suboptimal results that
Group O avoids if it deviates in all suboptimal-result cases; (2) the sub-
optimal results that posterior judges reach if Group O deviates in all
suboptimal-result cases; and (3) the normative principle used to compare
these two sets of results.
First, consider an additive principle for comparing the results. To
apply it, assign to each result in the first set of cases a number express-
ing its degree of suboptimality. Total these numbers. Then do the same
for the second set. The additive principle entails that if the first total
exceeds the second, then Group O has a stronger reason to deviate than
to adhere in all suboptimal-result cases, so it has an all-things-consid-
ered reason to deviate in such cases. However, if the second total exceeds
the first, then Group O has a stronger reason to adhere than to deviate
in all suboptimal-result cases and thus an all-things-considered reason
to adhere.
The additive principle is agent-neutral. It reflects the controversial
idea that, ceteris paribus, an agent is no less responsible for leading
someone else to cause harm or to act wrongfully than for himself caus-
ing harm or acting wrongfully. The additive principle contrasts with
agent-relative principles, four of which I shall discuss in this chapter.
Any of these four principles in its strongest form entails that Group O’s
reason to deviate in suboptimal-result cases is always stronger than its
reason to adhere. This is because of the following contrast. When judges
adhere in suboptimal-result cases they cause negative effects in a direct
way. However, when they deviate, resulting in systemic effects, they
cause these effects indirectly. According to the agent-relative principles
that I shall discuss, only the effects of adherence—the relatively direct
effects—provide judges with reasons. Systemic effects do not. It is there-
fore necessary to determine whether any of these agent-relative princi-
ples is true.
Consider how systemic effects such as mimetic failure are related to
anterior deviant decisions. These relations have four noteworthy fea-
tures:

1. By deviating in suboptimal-result cases, Group O does not intend

to encourage other judges to deviate in optimal-result cases. At
most, it foresees that deviating will have this result.
2. Group O does not use the parties who suffer, in cases of mimetic
failure, as means to its end. The harm inflicted by posterior judges

1. For the purposes of this chapter, I am assuming as sound the mimetic-failure argu-
ment of the previous chapter.
204 Limits of Legality: Part II

who deviate in optimal-result cases is merely a side effect or aspect

of anterior deviation by Group O.
3. Mimetic-failure effects are spatiotemporally remote from the
original deviant decisions. Judges who deviate in optimal-result
cases do so at later times and in different places than the decisions
of Group O.
4. Mimetic-failure effects are mediated by voluntary interventions
subsequent to the anterior deviant decisions. Group O does not
coerce, bribe, command, or threaten posterior judges. Posterior
judges decide autonomously to deviate.

For each of these features there is an agent-relative principle with

serious defenders that assigns moral significance to the feature in question.
Most contemporary deontologists accept one or more of the following
principles, or variants thereof:

1. Unintended effects provide weaker reasons than intended effects.

2. Harm caused as a side effect or aspect of action provides weaker
reasons than harm caused as a means to an end.
3. Remote effects provide weaker reasons than local effects.
4. Effects mediated by subsequent voluntary interventions provide
weaker reasons than unmediated effects.

Many deontologists today accept only threshold versions of these

principles,2 not absolute versions, but I have formulated them as absolute
principles for now. As absolute principles they have the following impli-
cations. One is not permitted to cause harm intentionally in order to pre-
vent foreseeable harm, however great the quantity of harm that one could
prevent. One is not permitted to cause harm in order to avoid causing
harm, however great the quantity, as a side effect or aspect of one’s action.
One is not permitted to cause local harm in order to prevent remote
harm, however great the quantity. Finally, one is not permitted to cause
unmediated harm in order to prevent mediated harm, however great the
quantity.
None of the aforementioned relations holds between adherent
decisions in suboptimal-result cases and the effects of those decisions on
the legally disfavored parties. Compare the effects of adherence in a

2. See Leo Katz, “Incommensurable Choices and the Problem of Moral Ignorance,”
University of Pennsylvania Law Review 146 (1998): 1465–85, p. 1483 (“[M]ost people are
threshold deontologists”). Threshold deontologists include Michael S. Moore, Placing Blame
(Oxford: Clarendon Press, 1997), p. 723; Thomas Nagel, “War and Massacre,” in Mortal
Questions (Cambridge: Cambridge University Press, 1979), p. 53; Robert Nozick, Anarchy,
State, and Utopia (New York: Basic Books, 1974), p. 30. For critical discussion see Larry
Alexander, “Deontology at the Threshold,” San Diego Law Review 37 (2000): 893–912;
Russell L. Christopher, “Deterring Retributivism: The Injustice of ‘Just’ Punishment,” North-
western University Law Review 96 (2002): 843–976, pp. 877 n. 185, 878 n. 188, 879–80.
Agent-Relative Principles 205

suboptimal-result case to the effects of deviation. When a judge deviates,

the negative effects on the losing party are not mediated by subsequent
voluntary interventions. The effects are spatiotemporally local, not
remote. In some cases these effects are also intended, not just foreseen.
In some cases they are means or ends, not just effects or aspects.3
Therefore, if any of the absolute agent-relative principles is true,
then Group O’s reason to adhere is never stronger than its reason to
deviate in suboptimal-result cases. Accordingly, the systemic effects of
deviation do not provide Group O with reasons to adhere that are
strong enough to override countervailing reasons to deviate. If any of
these absolute agent-relative principles is true, then deviating in subop-
timal-result cases is morally preferable to adhering for Group O, which
means that permissive rule is vindicated. I conclude that anyone who
supports obeying a nonpermissive rule must reject or qualify these
absolute agent-relative principles. I shall now examine them in more
detail.

12.2 INTENTION

If Group O obeys a nonpermissive rule, then it intentionally reaches sub-

optimal results in some suboptimal-result cases, whereas if Group O
deviates in all suboptimal-result cases, then it does not intentionally
reach any suboptimal result. It merely provokes other judges to deviate
in optimal-result cases. These judges intentionally reach results that are
actually suboptimal, but Group O does not. Group O provokes these
bad decisions, but it does not intend them, nor does it intend to provoke
other judges to make them.
Recall our earlier discussion of the moral significance of intention.4
Again, I shall not claim to represent the doctrine of double effect, but
shall entertain the simpler precept that the intended effects of an action
supply stronger reasons than do the merely foreseen effects. If this inten-
tion principle is true, then the suboptimal results reached by other judges
in optimal-result cases provide relatively weak reasons for Group O to
adhere in suboptimal-result cases. This is so despite the fact that Group
O, by deviating in suboptimal-result cases, provokes the other judges to
reach suboptimal results. However, the suboptimal results reached when
Group O adheres in suboptimal-result cases provide Group O with rela-
tively strong reasons to deviate in these cases. If the intention principle is
true, then Group O has a stronger reason to deviate than to adhere in
suboptimal-result cases.

3. Remember also that adhering in a suboptimal-result case inflicts greater undeserved

disadvantages on the legally disfavored party than deviating would inflict on the legally
favored party.
4. See §6.9.
206 Limits of Legality: Part II

12.3 MEANS

If Group O obeys a nonpermissive rule in order to support the rule of law,

then it reaches suboptimal results in some suboptimal-result cases as a
means to that end. In those cases it uses unjustified force and inflicts
undeserved harm upon legally disfavored parties in order to avoid encour-
aging other judges to deviate in optimal-result cases. Contrast the scenario
in which Group O deviates in all suboptimal-result cases and provokes
other judges to deviate in optimal-result cases. These others use unjusti-
fied force and inflict undeserved harm, but Group O does not use this
force or harm as a means to its end. Its end is simply to avoid itself using
unjustified force and inflicting undeserved harm upon the legally disfa-
vored parties in suboptimal-result cases. Group O provokes this unfortu-
nate deviation by other judges, but merely as a side effect or aspect of its
own deviation.
F. M. Kamm and others argue that the effects of actions taken as means
or ends supply stronger reasons than do effects that are mere effects or
aspects of the action. This view is expressed in Kamm’s principle of per-
missible harm.5 If this principle is true, then Group O has a stronger pro
tanto reason to deviate in suboptimal-result cases than to adhere, no mat-
ter how much suboptimal deviation by other judges it encourages.

12.4 PROXIMITY

The proximity principle holds that the effects of an action supply stron-
ger reasons, the greater their spatiotemporal proximity to the action.6
Systemic effects are spatiotemporally remote from the original deviant
decision. Therefore, if the proximity principle is true, then Group O does
not have an especially strong reason to avoid provoking other judges to
deviate in optimal-result cases.
More important, the spatiotemporal distance between an act of devi-
ation in a suboptimal-result case and its mimetic effects is greater than

5. See §6.10; F. M. Kamm, Morality, Mortality: Rights, Duties, and Status, vol. 2
(New York: Oxford University Press, 1996), p. 172.
6. See, e.g., F. M. Kamm, Intricate Ethics: Rights, Responsibilities and Permissible Harm
(New York: Oxford University Press, 2007), chs. 11–12. Kamm believes that proximity may
affect the duty to aid, but not the duty not to harm. Ibid., p. 386. Most of the literature on
the moral significance of proximity concerns duties to aid distant, needy strangers. See, e.g.,
Francesco Orsi, “Obligations of Nearness,” Journal of Value Inquiry 42 (2008): 1–21; Richard
W. Miller, “Beneficence, Duty and Distance,” Philosophy and Public Affairs 32 (2004): 357–
83; Jeremy Waldron, “Who Is My Neighbor? Humanity and Proximity,” Monist 86 (2003):
333–54; Violetta Igneski, “Distance, Determinacy and the Duty to Aid: A Reply to Kamm,”
Law and Philosophy 20 (2001): 605–16; David Schmidtz, “Islands in a Sea of Obligation:
Limits of the Duty to Rescue,” Law and Philosophy 19 (2000): 683–705; Peter Singer,
“Famine, Affluence, and Morality,” Philosophy and Public Affairs 1 (1972): 229–43.
Agent-Relative Principles 207

the distance between an act of adherence in a suboptimal-result case and

its negative effects on parties to the case. Therefore, if the proximity
principle is true, then ceteris paribus the negative effects of adherence
provide stronger reasons to deviate in suboptimal-result cases than
mimetic failure provides to adhere.

12.5 MEDIATED CAUSATION

The fourth and final agent-relative principle holds that subsequent

voluntary interventions by other agents attenuate or eliminate an agent’s
moral responsibility for the foreseeable effects of his actions. At law the
principle is known as novus actus interveniens.7 I shall use this phrase to
refer also to the moral version of the principle.
There are exceptions to novus actus. If A causes B to Φ, but B’s
Φ-ing is less than fully voluntary, then A retains at least partial respon-
sibility for B’s Φ-ing. This includes cases in which B acts from mental
incapacity or reflex and those in which A coerces, commands, or
instructs B to Φ. In these cases, A retains full responsibility for the
Φ-ing. However, none of these exceptions applies to mimetic failure.
Group O does not coerce, command, or instruct other judges to devi-
ate. Nor does a judge who is provoked by the deviation of Group O
deviate due to mental incapacity or reflex. He deviates because of poor
judgment or misinformation (factual, legal, or moral) that leads him to
conclude erroneously that the legally required result in the case at bar
is suboptimal. Mimetic failure occurs when deviation by Group O pro-
vokes another judge to deviate in an optimal-result case. But for the
latter judge’s reaction, the group’s deviation would not have had this
negative effect. The judge’s decision is a subsequent voluntary inter-
vention.
Therefore, if novus actus is true, then the negative effects of mimetic
failure provide Group O with only weak reasons to adhere, if they provide
any reasons at all. By contrast, the negative effects on the disadvantaged
parties are not mediated by subsequent voluntary interventions. Accord-
ing to novus actus, the negative effects of adherence in a suboptimal-result
case give Group O a reason to deviate that is stronger than the reason to
adhere generated by the mimetic effects of deviation.

7. The locus classicus is H. L. A. Hart and Tony Honoré, Causation in the Law (Oxford:
Clarendon Press, 1959). See also Michael E. Bratman, “What Is the Accordion Effect?” Jour-
nal of Ethics 10 (2006): 5–19; Michael S. Moore, “The Metaphysics of Causal Intervention,”
California Law Review 88 (2000): 827–77; Michael J. Zimmerman, “Intervening Agents and
Moral Responsibility,” Philosophical Quarterly 35 (1985): 347–58; Joel Feinberg, “Causing
Voluntary Actions,” in Metaphysics and Explanation, eds. W. H. Capitan and D. D. Merrill
(Pittsburgh: University of Pittsburgh Press, 1966).
208 Limits of Legality: Part II

12.6 AGENT-RELATIVE PRINCIPLES AND THE MORAL

SIGNIFICANCE OF SYSTEMIC EFFECTS

If we wish to maintain that systemic effects provide Group O with all-

things-considered reasons to obey a nonpermissive rule, then we must
hold that an agent can have reasons to inflict local, intended, unmedi-
ated harm as a means to prevent comparable harms that are remote,
unintended, mediated, and merely effects or aspects. In the previous
sections I described four popular agent-relative principles: the inten-
tion principle, the principle of permissible harm, the proximity princi-
ple, and novus actus interveniens. If any of these constitutes a true,
absolute moral principle, then systemic effects never provide Group O
with an all-things-considered reason to adhere in suboptimal-result
cases.
I conclude that if we wish to preserve the possibility that systemic
effects provide strong reasons to obey a nonpermissive rule, then we
must reject absolute versions of the four agent-relative principles. How-
ever, if my arguments from part I are sound, then Group O has no other
reasons to obey a nonpermissive rule. Therefore, the price of absolute,
agent-relative principles is the conclusion that no nonpermissive rule has
moral authority for Group O.
If we wish to maintain that systemic effects generate all-things-considered
reasons to obey a nonpermissive rule, then we must take one of two paths.
First, we could reject the four agent-relative principles altogether, at least for
the purpose of adjudication theory, and embrace agent-neutral principles
such as the additive principle.8 Alternatively, we could qualify agent-relative
principles with thresholds. A threshold principle forbids a certain kind of
harm unless it is necessary to prevent consequences that are not just worse,
but much worse. A threshold version of the intention principle, for example,
forbids intending harm9 unless it is necessary to prevent much greater quan-
tities of foreseeable harm. Threshold versions of the other agent-relative
principles are similar, mutatis mutandis.
I shall consider both options in §12.8. Before doing so I shall examine
another way in which comparison principles differ from one another.

12.7 COMPARATIVE HARM FUNCTIONS

Different comparison principles incorporate different comparative harm

functions. Here are some possible comparison principles. I formulate
them initially in their harm-minimizing versions:

8. See §12.1.
9. More precisely: harm to innocent, nonconsenting, nonthreatening parties.
Agent-Relative Principles 209

Pareto: adhering in the anterior set of cases is justified if and only if deviat-
ing in the posterior set produces at least one result that is more suboptimal
than any result produced by adhering in the anterior set.
Aggregative: adhering in the anterior set of cases is justified if and only if
the aggregate results of deviating in the posterior set are worse than the
aggregate results produced by adhering in the anterior set.10

Suppose we have empirical evidence that by deviating in all subopti-

mal-result cases, Group O causes posterior judges to reach at least one
result that is more suboptimal than any result that it thereby avoids reach-
ing. The Pareto principle entails that under this condition deviating in all
suboptimal-result cases is not justifiable. The aggregative principle does
not entail this.
Alternatively, suppose we have empirical evidence that by deviating in
all suboptimal-result cases, Group O causes posterior judges to reach
results that are more suboptimal in the aggregate than the results that
Group O thereby avoids reaching. The aggregative principle entails that
under this condition deviating in all suboptimal-result cases is not justifi-
able. The Pareto principle does not entail this.
The comparison principles can also be reformulated as threshold
principles:
Pareto/threshold: adhering in the anterior set is justified if and only if devi-
ating in the posterior set produces at least one result that is much more
suboptimal than any result produced by adhering in the anterior set.
Aggregative/threshold: adhering in the anterior set is justified if and only if
the aggregate results of deviating in the posterior set are much worse than
the aggregate results produced by adhering in the anterior set.

Suppose we have empirical evidence that by deviating in all suboptimal-

result cases, Group O causes via mimetic failure at least one result that is
much more suboptimal than any result that it thereby avoids reaching.
The Pareto/threshold principle entails that under this condition deviating
in all suboptimal-result cases is not justifiable. The aggregative/threshold
principle does not entail this.
Suppose we have empirical evidence that by deviating in all suboptimal-
result cases, Group O causes posterior judges to reach results that are much
worse in the aggregate than the results that Group O thereby avoids reach-
ing. The aggregative/threshold principle entails that under this condition
deviating in all suboptimal-result cases is not justifiable. The Pareto/thresh-
old principle does not entail this.
Any of these four principles, combined with the right empirical evi-
dence, supports the conclusion that Group O has an all-things-considered
reason to obey a nonpermissive rule. Notice that the case is easier to
make under aggregative principles than under Pareto principles. It is also
easier to make under harm-minimizing principles than under threshold

10. The additive principle from §12.1 is a version of this.

210 Limits of Legality: Part II

principles. The easiest case for obeying a nonpermissive rule is made

under an aggregative/harm-minimizing principle. The hardest case is
made under a Pareto/threshold principle.

12.8 THRESHOLD PRINCIPLES

Threshold principles also raise their own problems.11 I confess that I am

skeptical about treating them as ultimate moral principles. I see no reason
to set the threshold at any particular point. Perhaps others have more
confidently held intuitions than I do about the number of lives that would
have to be at risk in order to justify incinerating an innocent patient
infected with a contagious disease or torturing a single innocent person to
death (e.g., at the behest of a terrorist with a ticking bomb). “One hun-
dred,” people say. Or one thousand. Or 104, 105, 106. . . . Intuitions vary
widely. It is not simply that we cannot agree on a precise figure (25,444
yes, 25,443 no). It is not, as some think, a vagueness problem. The prob-
lem is that we cannot even achieve rough consensus on an order of mag-
nitude. We are reduced to mumbling that infringing agent-relative
restrictions is permissible in “extreme” cases without specifying the exten-
sion of “extreme,” even vaguely.
Instead of trying to intuit our way to thresholds, we could base our
theory of adjudication on agent-neutral principles. These, in contrast to
agent-relative principles, assign no ultimate moral significance to the rela-
tions between an action and its effects. This shift does not entail abandon-
ing agent-relative principles altogether. We could simply demote them
from foundational to derivative status. Agent-relative principles would
survive, but only as derivative principles, presumably with thresholds
specified by the foundational, agent-neutral principles. Agent-relative
principles make very useful heuristics.12 No one should object to them in
that capacity.
Debates about agent-relative principles are among the most central
and contentious in normative ethics. Different philosophers try to ground
agent-relative principles in different moral theories, including contractu-
alism, consequentialism, Kantianism, virtue theories, et cetera.13 Some
argue that agent-relative principles have enough intuitive support that
they need no theoretical defense. Meanwhile, disagreement persists about

11. See Alexander, “Deontology at the Threshold”; Christopher, “Deterring Retributiv-

ism: The Injustice of ‘Just’ Punishment”; Anthony Ellis, “Deontology, Incommensurability
and the Arbitrary,” Philosophy and Phenomenological Research 52 (1992): 855–75.
12. See, e.g., Cass R. Sunstein, “Moral Heuristics,” Behavioral and Brain Sciences 28
(2005): 531–73.
13. See, e.g., Paul Hurley, “Agent-Centered Restrictions: Clearing the Air of Paradox,”
Ethics 108 (1997): 120–46; Philip Pettit, “The Consequentialist Can Recognise Rights,” Phil-
osophical Quarterly 38 (1988): 42–55; Stephen L. Darwall, “Agent-Centered Restrictions
from the Inside Out,” Philosophical Studies 50 (1986): 291–319.
Agent-Relative Principles 211

where the thresholds lie.14 I believe that agent-relative principles are true
only insofar as they can be derived from agent-neutral first principles and
that the only principles that can be so derived have relatively low thresh-
olds.15 However, this book is the wrong place for me to defend these
positions. You can accept the rest of my arguments even if you disagree
with me about agent-relative principles, perhaps preferring to treat them
as foundational although qualified with thresholds. I shall not offer any
direct, global critique of principles that are both foundational and agent-
relative. But my arguments call into question the applicability of such
principles to adjudication theory, however applicable they may be to
interpersonal ethics or anything else. At least, I question the applicability
of these principles to adjudication theory for anyone who believes that
judges are ever morally obligated to adhere to the law in a suboptimal-
result case. If my arguments are sound, then anyone who would support
foundational, agent-relative principles in interpersonal ethics, for exam-
ple, must explain why these ostensibly foundational principles cease
applying to judges when they enter the courtroom and being applying to
them again when they leave, at the close of business, to resume their lives
as private individuals.
Notice, also, that stronger adherence rules require lower thresholds.
The lower we set the threshold, the closer our overall theory becomes to
one with no agent-relative restrictions whatsoever—one that simply per-
mits agents to minimize harm. As we approach that limit, treating agent-
relative principles as foundational appears increasingly ad hoc and
implausible. In these respects my arguments converge with those of
writers outside adjudication theory who conclude that agent-relative
principles do not bind states, however important they may be in interper-
sonal ethics.16

14. As discussed in Samantha Brennan, “Thresholds for Rights,” Southern Journal of

Philosophy 33 (1995): 143–68.
15. See Jeffrey Brand-Ballard, “Contractualism and Deontic Restrictions,” Ethics 114
(2004): 269–300.
16. See David Enoch, “Intending, Foreseeing, and the State,” Legal Theory 13 (2007):
69–99; Cass R. Sunstein and Adrian Vermeule, “Deterring Murder: A Reply,” Stanford
Law Review 58 (2006): 847–57, pp. 849–52; Louis Kaplow and Steven Shavell, Fairness
versus Welfare (Cambridge, Mass.: Harvard University Press, 2001); Robert E. Goodin,
Utilitarianism as a Public Philosophy (New York: Cambridge University Press, 1995),
pp. 51–57.
13

Optimal Adherence Rules

So far in part II, I have argued that Group O has an all-things-considered

reason to obey a rule that sometimes requires adherence in suboptimal-
result cases—a nonpermissive rule. My argument has many controversial
premises: certain empirical assumptions about mimetic failure, a theory
of collective practical reason, and a normative ethics for adjudication that
includes either no agent-relative principles whatsoever, or else principles
with relatively low thresholds.
After all this, we still do not know which adherence rule Group O has
reason to obey: restrictive rule or a moderate rule. Each adherence rule
maps onto a set of decision patterns that are consistent with Group O
obeying that rule. The optimal adherence rules are the ones associated
with the optimal decision patterns. The more often Group O adheres, the
greater is the quantity of pro tanto unjustified harm that it inflicts upon
legally disfavored parties in suboptimal-result cases. The more often
Group O deviates, the more pro tanto unjustified harm it encourages other
judges to inflict upon legally favored parties in optimal-result cases, via
mimetic failure. So Group O has pro tanto reasons to adhere and pro tanto
reasons to deviate. The relative strength of those two sets of reasons dic-
tates what Group O has an all-things-considered reason to do. The opti-
mal decision patterns are those with the highest possible deviation rate in
suboptimal-result cases that conform to whichever harm-comparison
principle we accept. Group O has an all-things-considered reason to obey
a certain adherence rule if and only if doing so brings the system as close
as possible to the optimal decision pattern.
Goldman recognizes the idea of an optimal decision pattern when dis-
cussing his taxpayer hypothetical. He appears to assume what I called, in
chapter 12, an aggregative/minimization principle: “[G]iven that the effect
of each individual’s deviation on the overall desired outcome . . . is mini-
mal or negligible, there will be an optimal level of deviation just below
the threshold at which collective harm begins to set in and outweigh fur-
ther individual benefits.”1 For any given harm-comparison principle, there

1. Alan H. Goldman, “The Rationality of Complying with Rules: Paradox Resolved,”

Ethics 116 (2006): 453–70, p. 455.

212
Optimal Adherence Rules 213

is a level of deviation density that produces a rate of deviation in optimal-

result cases that the principle classifies as “too high,” given the associated
rate of deviation in suboptimal-result cases.2 This is the deviation density
threshold, or threshold. An optimal decision pattern is one that approaches
threshold without exceeding it. Group O achieves the optimal pattern if
it successfully conforms to group restriction: do not deviate so often as to
cross the deviation density threshold.
The higher the thresholds specified for the agent-relative principles
mentioned in chapter 12, the more deviation group restriction permits.
The lower these thresholds, the less deviation group restriction permits.
The poles are absolute deontology and harm minimization. Absolute de-
ontology permits deviation in all suboptimal-result cases. Harm minimi-
zation permits only as much deviation as minimizes suboptimal results.
Threshold deontology permits less than absolute deontology but more
than harm minimization.
Obeying group restriction is, of course, impossible without some idea
where threshold falls in realistic legal systems. This question has a large
empirical component. The answer partly depends on how mimetic failure
varies with deviation density. Consider two judges who disagree with one
another about which results are suboptimal. The first judge makes a
decision that the second judge believes to be deviant. To what extent, if
any, does this belief increase the probability that the second judge will
deviate in cases in which the first judge would adhere (i.e., provoke mi-
metic failure)? This is an empirical question to which I have no good
empirical answer. The closest thing we have is evidence that judges engage
in first-order imitation: they sometimes announce that they are following
precedents with which they disagree. We can infer from such an an-
nouncement that the judge engages in first-order imitation.
We can also infer that judges engage in second-order imitation (mime-
sis), although the evidence is less conspicuous. When judges follow prec-
edent they do not ordinarily assert that they are following precedent
“because others have done so.” However, courts regularly cite authority
for the doctrine of stare decisis, which warrants the same inference.
The same is true of adhering to statutes, constitutions, and other legal
standards.
If adhering encourages other judges to adhere, then perhaps deviating
encourages others to deviate. This is harder to establish because judges
rarely admit when they deviate. In fact, they usually pretend to adhere in
such cases.3 I cannot find any empirical research on the tendency of judges
to imitate the deviational tendencies of others. This is not surprising.
Nevertheless, we can fall back on our general understanding of how social
norms function. Common sense tells me that judges adhere in perceived

2. Assuming we are not discussing hyposensitive systems. See §11.9.

3. See §16.1 on “surreptitious deviation.”
214 Limits of Legality: Part II

suboptimal-result cases in part because of a social norm to do so. Consider

a judge with the following attributes: he wants to deviate in perceived
suboptimal-result cases, and he adheres in such cases in part because of a
social norm to do so. If other judges deviate at an increasing rate, then at
some point he will notice and become more likely to deviate himself.
So common sense tells me that judges engage in mimesis, but it tells
me little about the shape or endpoints of the curve representing the rela-
tionship between deviation density and mimetic failure. Common sense
does not tell me how far along the curve my legal system falls or how
close it is to its threshold. It would be helpful for social scientists to de-
velop mathematical models of the relationship between deviation density
and mimetic failure and to test them in the field. I believe that progress in
the prescriptive theory of adjudication requires such models. While we
await these models, however, it is not unreasonable to believe that judges
could sometimes deviate without pushing a realistic legal system past its
threshold. If so, then group restriction permits more deviation than re-
strictive rule in realistic legal systems. I shall grant arguendo that it permits
less than permissive rule: consistent deviation in suboptimal-result cases
would eventually push a realistic system past its threshold. In other words,
I think group restriction supports a moderate rule.
Group O has an all-things-considered reason to obey group restriction.
Remember that this group reason is an objective appraisal reason. Recall
the distinctions among the following:
1. The best rule to which an agent’s decisions could conform
2. The best rule that a rule maker could promulgate for the agent
3. The best rule that a rule enforcer could enforce against the agent
4. The best rule for an agent to obey as a subjective, conscious
matter—to use as a guide for conduct4
Group restriction fits the first description. It answers the question, what is
the rule to which the decisions of Group O should conform, given the
tendencies of judges outside of Group O? This rule is not necessarily the
same as the rule that Group O should try to obey as a subjective, con-
scious matter. Conscious efforts to obey group restriction could be coun-
terproductive for Group O.
Group O obeys group restriction if and only if its members collectively
deviate at a rate that keeps their system below threshold. One way to
maintain acceptable deviation density is for every member to deviate at
the same rate—the optimal collective rate. But there are thousands of
other patterns that also maintain optimal density. For example, half the
members could deviate at 10 percent above the optimal collective rate
and half at 10 percent below. Only the average rate matters for the pur-
pose of maintaining optimal deviation density.

4. See §8.1.
Optimal Adherence Rules 215

13.1 GENERAL COMPLIANCE AND FREE RIDING

The next question is, which adherence rules do the individual members
of Group O have objective reasons to obey? This is now a question about
the individual judge rather than the group, but it is still a question about
the best rule to which the agent’s decisions could conform. It is not yet a
question about which rule a judge should subjectively try to obey.
Goldman claims that if a group has an obligation to perform a task,
then each individual member has an obligation to contribute her fair
share to the enterprise.5 The most familiar forms of free riding involve
tangible benefits and disadvantages such as collective goods and unpleas-
ant labor, respectively. But the structure of free riding applies to any ben-
efit or disadvantage whatsoever. Someone who desires to act morally
benefits when she satisfies this rational desire. Conversely, she suffers
when she performs a pro tanto impermissible action. Performing an action
that is pro tanto impermissible is worse for her than performing an action
that is not pro tanto impermissible, even if the former action is all-things-
considered permissible. When pro tanto impermissible actions must be
performed, there is a pro tanto reason to distribute them evenly across
agents who are disadvantaged by performing them, just as with any other
unpleasant task that must be done.
Recall the citizen who reroutes her tax payment to charity. She treats
unfairly those who pay. Goldman states that “other complying members
of the group justifiably would resent defection.”6 He concludes that each
citizen has an obligation to pay taxes rather than rerouting the funds to
charity, even if rerouting would do more good. Note that the consumers
of public goods are not the people who are wronged when a citizen
reroutes his taxes to charity. Those consumers have no greater moral claim
on the citizen’s resources than do the beneficiaries of his chosen charity.
Rather, the citizen who reroutes his taxes wrongs those who pay their
taxes. They, too, could have rerouted their funds to charities, thereby do-
ing more good. But they did not. They bore their fair share of a collective
burden. The citizen who reroutes his taxes makes an unjustified excep-
tion of himself. That is why his free riding is wrong. He violates the duty
of fair play. This reasoning applies to judges, too:
Fair Share Argument
1. If, by Φ-ing, a group acts on its group reasons, then each member has
an individual pro tanto reason to contribute his fair share to the group
effort to Φ.
2. Group O has an all-things-considered group reason to obey group
restriction (by the group adherence argument).
3. Group O obeys group restriction.

5. Goldman, “The Rationality of Complying with Rules: Paradox Resolved,” pp. 462–63.
6. Ibid., p. 465.
216 Limits of Legality: Part II

4. Therefore, each member of Group O has an individual pro tanto reason

to contribute his fair share to the group effort to obey group restriction.
5. A member of a group that obeys group restriction contributes his fair
share to the group effort if and only if he adheres to the law at the optimal
average rate or higher.
6. Therefore, every member of Group O has an individual pro tanto
reason to adhere to the law at the optimal average rate or higher.

The conclusion of the fair share argument is that every member of

Group O has an individual pro tanto reason to obey the following:
First Individual Restriction
If Group O deviates as often as group restriction permits, but no more
often, then deviate at no higher than the optimal average deviation rate.

If most judges adhere in suboptimal-result cases, then they inflict un-

justified disadvantages. If a judge deviates while others adhere, then she
rides free on them. By refusing to reach a suboptimal result, she wipes her
hands clean, but she wipes them on the hands of her fellow judges. “What
is so special about her?” her peers might ask. “Why must we adhere while
she deviates?” We could make this same point by reintroducing Scanlo-
nian contractualism, which now has some real work to do. Rather than
focusing on potential litigants,7 we now ask what adherence rules judges
could reasonably reject. That is another way of highlighting the partiality
of deviating when others obey.
Christopher McMahon offers yet another path to this conclusion. His
principle of collective rationality states in part, “One has sufficient reason
to contribute as provided to a cooperative scheme that produces some-
thing that one regards as good if the value to one of the outcome of the
scheme, when one’s contribution is added to the others that will actually
be made, exceeds the value to one of the noncooperative outcome.”8
Let us stipulate that in the judicial case the cooperative outcome is all
members of Group O adhering in all suboptimal-result cases, whereas the
noncooperative outcome is all members deviating. Given that stipulation,
each member prefers the cooperative to the noncooperative outcome,
so the principle of collective rationality entails that each has a pro tanto
reason to adhere.

13.2 AVERAGE DEVIATION RATE IS OPTIMAL OR BELOW

If Group O’s average deviation rate is optimal, according to group restric-

tion, then Judge Jack is permitted to deviate at the optimal average rate as

7. As in §8.4.5.
8. Christopher McMahon, Collective Rationality and Collective Reasoning (Cambridge:
Cambridge University Press, 2001), pp. 21–22.
Optimal Adherence Rules 217

well. This inflicts pro tanto unjustified harm on legally disfavored parties
in suboptimal-result cases, but it is justifiable, all things considered,
because necessary to avoid free riding. Jack must not deviate above the
optimal average rate. Otherwise he rides free on his fellow judges who
properly limit their deviation.
In realistic legal systems, however, Group O does not conform to group
restriction. What reasons does a member have under these nonideal con-
ditions? It depends where the system stands in relation to threshold. First,
consider a scenario in which the system is below the optimum. The anal-
ogous scenario for Goldman’s taxpayers is one in which so many citizens
pay their taxes that more good would be done if some of them were to
reroute. Goldman’s position appears to be that each citizen still has an
all-things-considered reason to pay his taxes under these conditions.
Goldman is correct if his position is that the rerouters ride free on the
taxpayers. But this just gives each rerouter a pro tanto reason to pay. He
also has a pro tanto reason to reroute based on impartial considerations: he
can do more good by rerouting. A rerouter rides free on every taxpayer, so
he has a pro tanto reason to pay. But a taxpayer fails to benefit all those
who would have benefited from his diversion, so he has a pro tanto reason
to reroute. Therefore, we cannot conclude that he has an all-things-
considered reason to pay taxes unless we can show that the former reason
is stronger than the latter. Goldman does not show this, so he offers
an inconclusive argument for the conclusion that citizens are obligated
to cooperate when they could do more good by rerouting their funds.
Goldman reaches a similar conclusion about judges: they must adhere
in suboptimal-result cases, with no exception for scenarios in which all
other judges adhere.9 But Goldman’s fairness argument is even less per-
suasive as applied to judges. Someone who pays his taxes rather than
rerouting his payment to charity does not infringe anyone’s rights. He
does not use force against the beneficiaries of the charity. He withholds
from them a benefit, but it is one to which they had no moral right as
against him.
By contrast, if Judge Juan adheres to the law in a suboptimal-result
case, then he infringes the moral rights of the losing party. He uses or
threatens pro tanto unjustified force against someone. Although he is unfair
to judges who adhere when he deviates, he neither uses nor threatens any
force against them. So his reason to deviate may be stronger than his fair-
ness reasons to adhere. Therefore, Juan may be permitted to deviate at the
optimal average rate even if the other members of Group O deviate less
often.
Juan may even be permitted to deviate above the optimal average rate
under these conditions. The system has “room” for additional deviation by
Juan. Consider a thought experiment. Ned lives on an island inhabited by

9. Alan H. Goldman, Practical Rules: When We Need Them and When We Don’t
(Cambridge: Cambridge University Press, 2002), pp. 104, 125–26, 137–39, 145–48.
218 Limits of Legality: Part II

four other individuals, A, B, C, and D. Ned cannot communicate with the

others, but he knows that there are four and that they all, himself in-
cluded, have plenty to eat. In addition to the ample food supply, delicious
wild berries grow on the island. Shortly after they ripen they fall from the
bush and spoil. Each inhabitant has a natural claim to an equal share of
the berries. Every day Ned picks some berries. He would gladly use all the
berries himself, but he knows that he is only entitled to one fifth of the
crop, so he limits his daily harvest so that by summer’s end he will have
taken only his share. However, he notices that only one of the other
inhabitants—D—is taking anything. The other three take nothing at all.
Finally, after watching the unclaimed shares rot for several weeks in a row,
Ned decides to take twice his share. This continues for several weeks. Ned
thinks maybe D will likewise begin to take more than his share, but D
never does. So Ned proceeds to take three times his share. Again he waits,
but the final share continues to go unclaimed. By summer’s end Ned’s
daily harvest amounts to four times his share.
Of course, a better solution would be for Ned to consult with the other
four at some point, to confirm that none wants the shares to which A, B,
and C are entitled. But I have ruled out communication. Under these
conditions, Ned may not be blameworthy for exceeding his entitlement. I
do not think that Ned is morally obligated to limit himself to his fair share
no matter how clear it becomes that berries are going to waste. Whether
his actions are objectively wrong depends on whether the others actually
wanted the berries that Ned takes beyond his entitlement. We could also
argue about how confident Ned must be that the berries are unwanted
before he begins taking them. But there must be a level of confidence that
allows Ned blamelessly to take the extra berries.
Let us now modify the facts of the hypothetical. Suppose Ned has
been taking only his fair share and observes that the others are doing like-
wise for several weeks. One day Ned greedily takes more than his share.
He is blameworthy for doing so because his maxim is something like “I
will take as many berries as I want.” However, his actions may not be ob-
jectively wrong. If, by coincidence, the others had no plans to harvest the
extra berries that Ned takes, then his greedy behavior is objectively per-
missible (although still as blameworthy as ever).
Apply these lessons to the situation of the judge. First, consider the
objective permissibility of a judge deviating in every suboptimal-result
case he hears. Suppose Judge Juan’s system is below its threshold and few
other judges have any interest in deviating. They would not deviate even
if Juan himself adhered consistently. Nor will Juan push the system over
threshold if he deviates in every suboptimal-result case. Under such con-
ditions, Juan is objectively permitted to deviate in every suboptimal-result
case. His situation parallels Ned’s when the other islanders have no inter-
est in the berries.
The question of whether Juan is blameworthy is more difficult. It
depends on his maxim of action, not just his behavior. Juan is blameworthy
Optimal Adherence Rules 219

if his maxim is permissive rule (“I will deviate in every suboptimal-result

case”), just as Ned was blameworthy for using the maxim, “I will take as
many berries as I want.” But suppose Juan’s maxim is “I will deviate in
every suboptimal-result case unless my system is close to its deviation
density threshold.” This maxim is a moderate rule. Using this maxim, Juan
is not blameworthy even if he deviates in every suboptimal-result case.
He is not blameworthy because his maxim responds appropriately to the
behavior of his fellow judges. He is like Ned in the first hypothetical,
when he has adequate reason to conclude that his fellow islanders have no
plans to harvest their shares. We might say that Juan is “consuming oppor-
tunities to deviate” that would otherwise be “wasted.” He takes more than
his fair share, but doing so is neither blameworthy nor objectively imper-
missible under the conditions described.
If Juan wants to stay on the safe side, of course, he can deviate less fre-
quently than the aforementioned moderate rule permits, obeying a more
restrictive moderate rule, or even restrictive rule. Any judge who main-
tains a deviation rate higher than the optimal average is playing a risky
game. His fellow judges will always be able to claim that they wanted to
deviate more often than they did but refrained because of judges such as
him. To this Juan could reply that the others should have made their
desires clearer to him, but he may find it difficult to persuade others that
his maxim is, indeed, sufficiently responsive to facts about the proximity
of his system to its threshold. A more restrictive rule is safer for the judge
who wants to avoid blame.
Deviating above the optimal average rate constitutes free riding and is
pro tanto wrong, but it may be all-things-considered permissible if the
system is below threshold, for the following reason: if Juan deviates less
often, then he inflicts pro tanto unjustified harm on legally disfavored
parties in suboptimal-result cases. Infringing moral rights is usually worse
than free riding. However, the further Juan goes above the optimal aver-
age rate, the more of his fellow judges he “passes” with his deviation rate.
His fair play reasons to adhere strengthen commensurately the further he
goes. So there is still a limit to how often he is permitted to deviate,
although he is not limited to the optimal average rate. I suggest the fol-
lowing rule:
Second Individual Restriction
Where n is the maximum number of cases in which group restriction per-
mits deviating during a certain time period and m is the number of cases in
which Group O actually deviates during that period, if m < n during that
period, then do not deviate in more than n – m cases during that period.

Remember, I am not claiming that the best guidance rule for judges directs
them to estimate how close their system is to threshold. That might prove
impractical. I am still addressing the level of appraisal rules.
220 Limits of Legality: Part II

13.3 GENERAL DEFECTION

Now consider a legal system that is far above its threshold. In this system,
every member of Group O deviates in more suboptimal-result cases than
first individual restriction permits. Every member is above the optimal
average rate. Other judges attempt to imitate Group O, but they misiden-
tify suboptimal-result cases and end up deviating unknowingly in too
many optimal-result cases, based on the stipulated threshold. In other
words, the deviation of Group O causes too much mimetic failure.
Group O still has an all-things-considered reason to obey group
restriction—a stronger rule than its members actually obey. But does any
individual member have a reason to obey first individual restriction when
no one else obeys it? Goldman’s taxpayer argument does not support
the unconditional conclusion that taxpayers have a reason to pay their
taxes. It supports, at most, the conditional conclusion that each citizen has
a reason to pay her taxes if enough other citizens pay theirs. Consider the
citizen who reroutes her tax payment to charity while all other citizens
do the same. She rides free on no one, so she has no reason of fairness to
pay. Moreover, the public goods are not provided whatever she does. Her
act makes no perceptible difference. Does she still have a reason to pay?
Goldman does not distinguish between general cooperation and general
defection or entertain the possibility that citizens’ obligations differ across
these states of affairs. But his argument does not support the conclusion
that all citizens have unconditional reasons to pay their taxes.
Similarly, the fair share argument establishes only that judges have rea-
sons to obey first individual restriction if the other members of Group O
obey it. In order to cover the neglected case of general defection, we must
supplement the argument. We must hold each deviating judge responsi-
ble for the cumulative effects of patterns of general deviation in which he
participates, even if his own decisions have no perceptible effect and he
rides free on no one.
Of course, a pattern in which every member of Group O deviates more
often than first individual restriction permits is not realistic. In a more
realistic pattern, some members deviate more often than first individual
restriction permits, others conform to first individual restriction, and the
overall rate is too high. The greater the number of conformists, the stron-
ger the judge’s fairness reasons to adhere. But these reasons still may not
be strong enough to outweigh his reasons to deviate, even if the system is
above threshold. Indeed, as the system rises beyond threshold, the number
of conformists falls, commensurately weakening fairness reasons to
adhere.
There are, however, additional arguments for adherence that are based
not on fairness but more directly on systemic effects. I turn to these argu-
ments in the next sections.
Optimal Adherence Rules 221

13.4 IMPERCEPTIBLE EFFECTS

How might systemic effects provide a member of Group O with a reason

to adhere in a suboptimal-result case, given that a single choice to deviate
makes no perceptible difference? Such a reason might exist if what I shall
call the causal limitation is false. This limitation holds that only the per-
ceptible effects of an action constitute objective reasons for or against
performing it. Act consequentialists accept the causal limitation, holding
that an act is wrong, all things considered, if and only if it causes less
agent-neutral value to be produced than would an alternative act. Act
consequentialism combines the causal limitation with agent neutrality. It
is important to keep these elements separate as we proceed. One could,
for example, hold that an agent acts wrongly if and only if her actions
amuse someone who maximizes agent-neutral value. This preposterous
principle is agent-neutral because it ultimately defines the right in agent-
neutral terms. But it is noncausal because it does not define right action in
terms of how much value it actually promotes. In this chapter and the
next, I shall argue that anyone who believes that judges must obey a non-
permissive rule should embrace foundational agent-neutral principles but
reject the causal limitation. This seems like a coherent position to me,
although maybe not the only coherent position on the subject.
The fact that a deviant decision typically has no perceptible systemic
effects is especially significant because the immediate adverse effects of
adherence are typically very perceptible. A judge who adheres to the law
in a suboptimal-result case disadvantages particular, identified individuals,
notably the losing party. Of course, adherence also benefits the winner,
but in suboptimal-result cases the winner does not receive enough
deserved benefits to outweigh the loser’s undeserved losses. So the objec-
tion from the imperceptibility of systemic effects has great urgency for
anyone who believes that judges should obey a nonpermissive rule. We
still have not identified a pro tanto reason to adhere that is strong enough to
override the pro tanto moral reason to deviate in suboptimal-result cases.
If we wish to advance a systemic-effects argument for the conclusion
that judges have pro tanto reasons to adhere in suboptimal-result cases,
then we must maintain that a judge who deviates acts pro tanto wrongly,
despite the fact that her deviation makes no perceptible causal contribu-
tion to harming the legal system. Many philosophers believe that an ac-
tion that causes no perceptible harm to anyone can still be pro tanto
wrong. Some consider it intrinsically pro tanto wrong for a soldier to tell
his fallen comrade’s grieving mother, falsely, that her son died painlessly. I
myself do not. But deviating as such does not require deception, so I can
drop the issue.10

10. A judge who deviates might decide, in addition, to misrepresent his decision as
conforming to the law. He might have a pro tanto moral reason not to do so under conditions
that will foreseeably mislead others.
222 Limits of Legality: Part II

More analogous is one’s decision to break a promise, made to a dying

woman, to deliver a pointlessly hurtful message to someone after her
death. Some consider it intrinsically pro tanto wrong to break this prom-
ise.11 I myself do not. But I concluded in chapter 9 that even if deviating
breaks a promise, the judge’s reason to keep it is undermined in subopti-
mal-result cases.
In the next three subsections I shall consider three different theories,
any one of which would allow us to hold an agent morally accountable for
a state of affairs to which she makes no perceptible contribution: imper-
ceptible harms, triggering risks, and complicity.

13.4.1 Imperceptible Harms

First, one could embrace the existence and moral significance of impercep-
tible harms. The classic hypothetical that motivates belief in the moral
significance of imperceptible harms involves 100 bandits, who steal 100
beans from hungry villagers. Each bandit steals 1 bean from each villager.
No villager can detect the difference any given bandit makes to his lunch,
yet we still think the bandits act immorally. One way to explain this is to
claim that each bandit causes imperceptible harm and that doing so is
wrong.12
Analogously, one could argue that a deviant decision inflicts impercep-
tible harm on the legal system. True, it is rare for a single deviant decision
to cause any posterior judge to weaken her subjective reasons to adhere,
causing her to deviate in an optimal-result case. This is to say that the
decision inflicts no perceptible harm. Nonetheless, one could insist that the
deviating judge inflicts harm by contributing, albeit imperceptibly, to
a “culture of judicial lawlessness” in which deviation is more likely in
optimal-result cases.
However, some have argued that assigning moral significance to imper-
ceptible harms entails various repugnant conclusions.13 I am not sure that
these objections to the moral significance of imperceptible harms are con-
clusive, but they worry me enough to motivate me to continue my search
for reasons for a judge to adhere even when deviating causes no percepti-
ble harm.

11. And to make it in the first place, if one has no intention of keeping it.
12. See Jonathan Glover, “It Makes No Difference Whether or Not I Do It,” in Applied
Ethics, ed. Peter Singer (Oxford: Oxford University Press, 1990).
13. Derek Parfit, Reasons and Persons (Oxford: Clarendon Press, 1984), pp. 381–90. See
also Michael J. Almeida, ed., Imperceptible Harms and Benefits (Dordrecht: Kluwer, 2000);
Larry S. Temkin, “Rethinking the Good, Moral Ideals and the Nature of Practical Reasons,”
in Reading Parfit, ed. Jonathan Dancy (Oxford: Blackwell, 1997).
Optimal Adherence Rules 223

13.4.2 Triggering Risks

Here is a second way of supporting the idea that an action that causes no
perceptible harm to anyone—nor any deception, nor any infidelity, et
cetera—can still be objectively wrong. One could treat each deviant
decision as imposing a small risk of causing perceptible harm and hold
agents objectively responsible for exposing others to such risks.14 Unlikely
as it may be in any given case, there could in principle be a deviant decision
in a suboptimal-result case that directly causes one or more deviant
decisions in posterior optimal-result cases. Judge Karl reads or otherwise
learns about a deviant decision by Judge Rita. Thereafter, Karl has to
decide a case that he erroneously believes to be a suboptimal-result case.
Remembering Rita’s deviant decision or subconsciously aware of it, Karl
is now encouraged to deviate and does so. Perhaps Rita’s decision reduces
Karl’s sense of how “unprofessional” deviation is, just to the point at which
he summons his resolve to deviate. Whatever the causal mechanism, the
fact is that if Rita had adhered, then so would have Karl. Rita’s deviation
is decisive for Karl: it is a trigger. She has “provoked” him to deviate. So she
has an objective pro tanto reason to adhere that competes with any pro
tanto reason she might have to deviate. If Karl’s decision is more subopti-
mal than an adherent decision by Rita would be, then she has an objective
all-things-considered reason to adhere. If she deviates, then she runs a risk
of acting against her objective, all-things-considered reasons. Some would
claim that she has an objective reason not to take such risks.
Any deviant decision could potentially have a triggering effect. But
how great is the risk? Of course, virtually every ruling, deviant or not,
adversely affects the losing party, his family, and others in virtue of its
content. But only in extraordinary cases does anyone exist whose interests
are perceptibly affected for the worse by the fact that a decision is deviant.
In realistic, stable legal systems very few deviant decisions directly cause
subsequent deviation in optimal-result cases. Any given deviant decision
runs only a miniscule risk of triggering.
The question, then, is whether people have a right not to be exposed
to miniscule risks of undeserved harm. If they do, then judges have objec-
tive, pro tanto moral reasons to adhere, despite the fact that deviation
rarely has harmful effects. In this section I ask how we should evaluate
actions that impose miniscule risks of harm.

14. See, e.g., Frank Jackson and Michael Smith, “Absolutist Ethical Theories and Uncer-
tainty,” Journal of Philosophy 103 (2006): 267–283; David McCarthy, “Rights, Explanation,
and Risks,” Ethics 107 (1997): 205–25; Dennis McKerlie, “Rights and Risk,” Canadian Jour-
nal of Philosophy 16 (1986): 239–51; Judith Jarvis Thomson, Rights, Restitution, and Risk
(Cambridge: Harvard University Press, 1986); Peter Railton, “Locke, Stock, and Peril: Natu-
ral Property Rights, Pollution, and Risk,” in To Breathe Freely, ed. Mary Gibson (Totowa, N.J.:
Rowman & Littlefield, 1985).
224 Limits of Legality: Part II

Dan pops a hole in a bag of pretzels on a supermarket shelf and inserts

a peanut, creating a small chance that someone with a severe peanut
allergy will eat it and die. If this happens, then Dan’s act is an unjustifiable
homicide. Objectively, it is extremely immoral. Nevertheless, Dan might
not be guilty of criminal homicide. Reckless homicide requires risks to be
substantial, whereas the risk Dan runs may be insubstantial.15
If no allergic person eats the peanut, then Dan’s act has no effects that
make it objectively very immoral.16 But it might have aspects other than
its effects that make it objectively very immoral. Dan exposes allergic
consumers to a risk of severe harm. It is, however, a miniscule risk. One
might suggest that agents should simply ignore risks below a certain prob-
ability, however severe the harm risked. One could claim that below a
certain level of risk, agents who knowingly take a risk of doing something
objectively wrong (even extremely wrong) are not pro tanto blameworthy.
If a fully informed agent is not pro tanto blameworthy for an action, then
he has no objective pro tanto moral reason not to perform it. So Dan has
no objective pro tanto moral reason to refrain from hiding the peanut on
this view. Again, the criminal law has something in common with this
position. Dan is probably not guilty of reckless endangerment, either,
given the low level of risk.
At the other extreme, one could maintain that individuals have an
absolute right not to have others expose them to even miniscule risks of
harm. This entails that Dan violates the rights of allergic individuals and
has an objective all-things-considered moral reason not to hide the pea-
nut. This view, however, is highly restrictive. It entails that pedestrians
have an absolute right to keep cars off the road.
There is a moderate view that is less restrictive than the former, but
more restrictive than the criminal law. This is the view that individuals
have a right not to have others expose them to unjustifiable risks of harm—
even miniscule risks.17 This is a nonabsolute right against risks because
some risks are justifiable. This view entails that Dan has an objective pro
tanto moral reason not to hide the peanut, which becomes an all-things-
considered reason because he has no justification for doing so. This is a
theory that we could use to analyze a judge’s decision to deviate. If devi-
ating creates a risk of provoking deviation in optimal-result cases, then the
judge has an objective pro tanto reason to adhere, even if that risk is min-
iscule (as it probably is). Unlike Dan, however, the judge also has pro
tanto reasons to deviate in suboptimal-result cases.

15. Model Penal Code § 2.02(2)(c) (reckless actors consciously disregard a substantial
and unjustifiable risk).
16. Popping holes in pretzel bags at the market and inserting peanuts is also wrong in
itself, as destruction and adulteration of groceries, but these are petty wrongs that I shall
disregard.
17. See, e.g., Larry Alexander and Kimberly Kessler Ferzan, Crime and Culpability: A
Theory of Criminal Law (Cambridge: Cambridge University Press, 2009).
Optimal Adherence Rules 225

13.4.3 Complicity
Now for a third way of supporting the idea that an action that causes no
perceptible harm to anyone can be objectively wrong. Some philosophers
have recently argued that a contributing agent is accountable for the cu-
mulative effects of the actions of the members of a group to which she
belongs. Christopher Kutz’s complicity principle holds an agent account-
able for what others do when she “intentionally participates” in the wrong
they do or the harm they cause, independently of the actual difference
she makes.18 The complicity principle holds individuals accountable for
such actions as driving a car that emits greenhouse gases in quantities too
small to register on the global scale.19 It holds accountable each of the
Allied pilots who participated in the firebombing of Dresden civilians
during World War II, whether or not any particular pilot caused any par-
ticular injury.20 It also condemns anyone who indifferently provides tools
to criminals, even if the tools are widely available elsewhere.21 These ac-
tions make no perceptible difference to outcomes but the complicity
principle holds the agents accountable nonetheless.
Assuming that the citizenry as a group acts wrongly by failing to pro-
vide public goods through taxpaying, the complicity principle could sup-
port the claim that anyone who reroutes her tax payment intentionally
participates in the group wrongdoing. Under the complicity principle she
acts wrongly herself, even though her act makes no perceptible difference.
The goods will not be provided whatever she does. Assume that all other
citizens are rerouting their payments to charity, so public goods are not
provided whatever the citizen does. This constitutes a great misfortune for
everyone. Any citizen who reroutes his payment under these conditions
intentionally participates in causing this misfortune, although he makes
no difference to the outcome. Therefore, under the complicity principle
each is morally accountable for the general pattern of nonpayment. Each
has a pro tanto reason to pay even though no one else is paying.
Returning to the judicial case, we have noted that the systemic effects
of a single deviant decision are usually imperceptible. However, the cu-
mulative systemic effects of a set of deviant decisions can still give a judge
a reason to adhere. The complicity principle links deviant decisions to the
systemic effects of deviation patterns. Each agent is objectively responsi-
ble for the effects (intended or not) of her actions and for the effects
(intended or not) of any joint intentional activity in which she partici-
pates.22 A pattern of deviant decisions in suboptimal-result cases by Group

18. Christopher Kutz, Complicity: Ethics and Law for a Collective Age (Cambridge:
Cambridge University Press, 2000), p. 122.
19. Ibid., pp. 171–77.
20. Ibid., pp. 117–20.
21. Ibid., pp. 168–70.
22. See §13.4.3.
226 Limits of Legality: Part II

O provokes posterior judges to deviate in a larger number of optimal-

result cases than the number in which they would have otherwise devi-
ated. Because of mimetic failure, a widespread practice of deviation, even
if limited to suboptimal-result cases, indirectly encourages deviation in
optimal-result cases. Judges who deviate in optimal-result cases often use
unjustified force. They also violate their judicial oaths and frustrate parties’
reasonable expectations. The complicity principle holds each judge in
Group O responsible for the collective actions of Group O that provoke
posterior judges (outside Group O) to engage in this behavior.

This section (13.4) has described three different ways to conceptualize

the pro tanto wrong of deviation in suboptimal-result cases when other
members of Group O deviate excessively. We can understand deviation
under these conditions as inflicting imperceptible systemic effects, or as
running a risk of triggering perceptible systemic effects, or as participating
in a wrongful enterprise that causes perceptible systemic effects. Each of
these theories is controversial, and I shall not attempt to defend any of
them. Any of the three supports the claim that systemic effects give judg-
es pro tanto reasons to adhere in suboptimal-result cases. Recall that my
argument in part II addresses those who believe that judges are morally
forbidden to make a practice of deviating in every suboptimal-result case.
Anyone who believes this must hope that one of these theories, or an-
other that does the same work, is correct.
If any of the principles associated with these theories has normative
authority, then agents have more natural reasons than I originally identi-
fied.23 They still have pro tanto natural reasons to use justified force and
not to use unjustified force.24 But they also have pro tanto natural reasons
to join enterprises with participants who use justified force and enter-
prises that cause nonparticipants to use justified force. They have pro
tanto natural reasons not to participate in enterprises with participants
who use unjustified force or fail to use justified force. And they have pro
tanto natural reasons not to participate in enterprises that cause nonpar-
ticipants to use unjustified force or to fail to use justified force.
We can now address the case of general deviation in suboptimal-result
cases with the following argument:
Clean Hands Argument
1. If a group has an all-things-considered group reason to Φ, but does not
Φ, then its members have individual pro tanto reasons to Φ so as not to
contribute to the failure of the group to Φ.
2. Group O has an all-things-considered group reason to obey group
restriction (by the group adherence argument).

23. See §2.5.

24. By force, as usual, I mean force, threat of force, or other coercion.
Optimal Adherence Rules 227

3. If Group O disobeys group restriction, then adhering at no less than

the optimal average rate is the only way for a judge in Group O to avoid
contributing to the group’s disobedience of group restriction.
4. Therefore, if Group O disobeys group restriction, then each member
of Group O has an individual pro tanto reason to adhere at no less than
the optimal average rate.

13.5 THE COMPARATIVE WEAKNESS OF ADHERENCE REASONS

I pause in this section to forestall a potential misunderstanding. Theories of

imperceptible effects, triggering risks, and complicity can be used to sup-
port strict principles of abstention: the bandit must not take the single
bean; Dan must not hide the peanut; the pilot must not firebomb Dresden.
One might infer that if one of these theories is correct, then judges in
Group O must never deviate if the deviation rate of Group O is excessive.
This conclusion does not follow because a judge is not similarly situ-
ated to the agents in the aforementioned scenarios. The agents in these
scenarios have no moral reasons whatsoever to act as they do. The bandit
has no moral reason to take the villager’s bean. Nor has Dan any moral
reason to hide the peanut in the pretzel bag. Nor has the pilot any (ade-
quate) moral reason to bomb Dresden. The puzzles arise because, even so,
theories that incorporate a causal limitation cannot easily explain the ev-
ident immorality of these actions. Embracing the moral significance of
imperceptible harm, or responsibility for imposing miniscule risks, or the
complicity principle at least allows one to condemn the bandits, Dan, and
the Dresden bombers, respectively. These agents lack even pro tanto moral
reasons for their actions.
By contrast, judges in suboptimal-result cases have strong pro tanto
moral reasons to deviate. Theories based on imperceptible harms, trigger-
ing risks, or complicity provide competing pro tanto reasons to adhere, but
these reasons do not necessarily prevail in the competition. Administering
a vaccination subjects the patient to a minute risk of death. This risk gives
the doctor a pro tanto reason not to administer the vaccine, but it is a weak
reason. Of course if the vaccine serves no medical purpose, then this weak
reason becomes an all-things-considered reason. But if the vaccine will
prevent disease, then the doctor has a pro tanto reason to administer it
that overrides the opposing pro tanto reason and generates an all-things-
considered reason to do so.
Similarly, imagine that each bandit is a “Robin Hood” who steals a bean
from one of a hundred wealthy villagers in order to contribute it to a food
bank that saves much hungrier people in the next town from starvation.
In that case the imperceptible harm that he inflicts may be outweighed by
benefits to others.25

25. Depending, of course, upon the correct theory of distributive justice.

228 Limits of Legality: Part II

The point is that small disadvantages and risks of harm may be justified
as unavoidable side effects of permissible means to permissible ends. The
negative effects of adherence, on the losing party and others, are often
direct and substantial in suboptimal-result cases. However, the systemic
effects of a single decision to deviate rarely constitute adherence reasons
that are strong enough to override a judge’s natural duties and sustain an
all-things-considered reason to adhere. Because deviation either imposes
a triggering risk of systemic effects, or inflicts imperceptible systemic ef-
fects, or constitutes complicity in a harmful enterprise, judges have pro
tanto reasons to adhere. However, these reasons may not be strong enough
to defeat the judge’s pro tanto reasons to deviate in all suboptimal-result
cases. Running the risk of provoking deviation by Judge Karl in an opti-
mal-result case is an unavoidable side effect of Judge Rita’s choice to de-
viate in a suboptimal-result case. So the risk gives Rita a pro tanto reason
to adhere, but may not provide an all-things-considered reason, depend-
ing on the deviation density of the system at the time.

13.6 DEVIATION RATE ABOVE THE OPTIMUM

We can now address the scenario in which Group O’s average deviation
rate is above the optimum. What is Judge Juan permitted to do under
those conditions? He may not deviate above the optimal average rate.
Otherwise he participates in the ongoing deviation enterprise that is caus-
ing excessive mimetic failure. If some members deviate at or below the
optimal average rate, then he has an additional reason not to deviate above
the optimal average rate. Otherwise he rides free on the members who
deviate less often than he.
So Juan is forbidden to deviate above the optimal average rate. Beyond
that there are two possibilities that make some intuitive sense. The first is
that Juan is permitted to deviate at the optimal average rate. In doing so
he does not ride free on anyone. He inflicts pro tanto unjustified harm on
legally disfavored parties in the suboptimal-result cases in which he
adheres, but some adherence is necessary lest he participate in a deviation
enterprise that causes mimetic failure. If some members of Group O
deviate at or below the optimal average rate, then Juan must not exceed
the optimal average rate either, in order to avoid free riding. If everyone
else deviates above the optimal average rate, then Juan could, without
riding free on anyone, deviate as often as the one who deviates least fre-
quently. But he would still participate in an enterprise of excessive devi-
ation by deviating that often, so he must deviate at or below the optimal
rate.
If Group O’s average deviation rate is above the optimum, then devi-
ating at the optimal average rate is sufficient to avoid free riding. How-
ever, one might wonder whether deviating at the optimal average rate is
sufficient to avoid participation in the enterprise of excessive deviation. It
Optimal Adherence Rules 229

might seem that Juan participates in this enterprise if he deviates at all.

This is the second of the two, intuitively sensible possibilities mentioned
in the previous paragraph.
I think this suggestion about Juan is false. The fact that other judges are
deviating above the optimal average rate does not obligate Juan to keep
his deviation rate below the optimal average rate. He is not obligated to
compensate for the excessive deviation of other members of Group O.26
More to the point, the legally disfavored parties whom Juan could protect
by deviating should not have to suffer on account of excessive deviation
by other members of Group O. Juan does not participate in an enterprise
of deviating above the optimal average rate unless he deviates above that
rate. I suggest the following rule:
Third Individual Restriction
If Group O deviates more often than group restriction permits, then do not
deviate at a rate higher than the optimal average deviation rate.

I have proposed three individual restrictions that I now incorporate

into the following:
Combined Restriction
1. If Group O deviates as often as, or more often than, group restriction
permits, then deviate at no higher than the optimal average deviation rate.
2. Where n is the maximum number of cases in which group restriction
permits Group O to deviate during a certain time period and m is the
number of cases in which Group O actually deviates during that period, if
m < n, then do not deviate in more than n – m cases during that period.

13.7 ADAPTATION EFFECTS REVISITED

Earlier I distinguished adaptation effects from mimetic effects.27 Adapta-

tion effects result when nonjudicial actors adapt to patterns of deviation
by Group O. Mimetic effects result when judges outside Group O unsuc-
cessfully attempt to imitate Group O’s deviation. Since drawing this dis-
tinction I have neglected adaptation effects, using only mimetic effects in
my arguments. I can now explain why we should not substitute adapta-
tion effects for mimetic effects in these arguments.
The problem is simple: depending on the law adaptation effects can be
bad, as in cases like Yasmin’s, but they can also be good, from the moral
standpoint of anterior judges. Consider judges who are persuaded by
economic arguments against rent-control statutes. Assume that a
widespread practice of judicial deviation from such statutes would have

26. I see parallels with arguments in Garrett Cullity, The Moral Demands of Affluence
(Oxford: Oxford University Press, 2004), but I cannot explore them here.
27. See §§11.2–11.3.
230 Limits of Legality: Part II

adaptation effects. Do these effects give prior judges a reason to adhere to

such statutes? From the moral standpoint of the anterior judges these ef-
fects are likely to be positive. At least, the effects are likely to be positive
if lawyers and their clients can discern a pattern of judicial deviation.
From the moral standpoint of judges who disapprove of rent-control stat-
utes, this pattern of otherwise optimal deviation will have negative adap-
tation effects only if no pattern becomes apparent to other actors.
We can, in theory, imagine legal actors reacting to a pattern of judicial
deviation from rent-control statutes by arriving at the hysterical conclu-
sion that judges have become less willing to enforce all rental contracts or
even contracts in general. In that case the prior pattern would discourage
beneficial economic activity. More likely, however, lawyers will figure out
that judges are deviating from rent-control statutes because the judges
think them economically inefficient. Lawyers will not, therefore, advise
their clients to refrain from other economically efficient transactions in-
volving real estate. Private parties who are broadly self-interested have
powerful incentives to obtain accurate predictions of judicial behavior.
Inaccurate predictions lead to lost opportunities and poor investments.
There is no economic reward for paranoia. Suppose judges are regularly
nullifying rent-control statutes. Lawyers respond by advising prospective
tenants not to assume that rent control will be enforced in the future. But
lawyers will not interpret the trend as evidence that courts will also devi-
ate from economically efficient laws that favor landlords. Lawyers will not,
for example, advise their clients not to sign leases as a general matter, re-
gardless of whether the building has rent control. The point is that adap-
tation effects provide reasons to adhere in some cases but not others
depending upon the content of the controlling law. In fact, adaptation
effects provide reasons to deviate in some cases. They provide no general
reason to obey a nonpermissive rule.
Now we can see why mimetic failure is a better candidate for a general
reason to obey a nonpermissive rule. Compared with the incentives for
legal subjects to predict judicial decisions accurately, the incentives are
weak for judges to imitate one another accurately when deviating.
Although a judge has various incentives to adhere to the law,28 on those
occasions when she chooses to deviate she has little incentive to do so in
accordance with anyone’s values but her own. When Judge Dave the
Democrat deviates from a statute providing mandatory minimum sen-
tences for felons, he may inspire Judge Ruth the Republican to deviate
from a statute extending civil rights to lesbians and gay men—a statute
that Dan favors. The symmetrical possibility exists if Ruth’s decision pre-
cedes Dan’s. Dave sees Ruth’s decision as suboptimal, as she sees his, no
matter who goes first.

28. See §§3.2, 4.6.

Optimal Adherence Rules 231

Adaptation effects can provide pro tanto reasons for a group of judges
to adhere in certain suboptimal-result cases, but these reasons vary with
the content of the laws in question. Adaptation effects do not provide
content-independent reasons to adhere in suboptimal-result cases unless
lawyers and their clients are very bad at predicting judicial behavior—
worse, I think, than they are in realistic modern legal systems. By contrast,
mimetic failure provides content-independent pro tanto reasons for the
group to adhere in suboptimal-result cases under a much weaker condi-
tion: that a substantial plurality of their fellow judges will respond to the
group’s deviation pattern by deviating in optimal-result cases. This is so if
other judges attempt to imitate the group by deviating, but either reject
or misunderstand the group’s classification of suboptimal-result cases.
Given the range of moral opinions and the variations in moral perceptive-
ness on the bench in modern legal systems, this condition plausibly holds.
This is why mimetic failure could, perhaps, provide content-independent
pro tanto reasons for judges to adhere in suboptimal-result cases.

13.8 COHERENCE AND COMMON SENSE

Many people believe that judges in reasonably just legal systems have an
all-things-considered moral obligation to adhere to the law in all cases—
to obey restrictive rule. Given current circumstances as I understand
them, my arguments thus far do not support this conclusion. My conclu-
sion thus far appears to conflict with a common belief about the ethical
obligations of modern judges. I have admitted that I cannot demonstrate
that judges have no obligation to obey restrictive rule. I have simply con-
sidered arguments for the opposite conclusion and found them wanting.
Therefore, we must choose between (1) denying that judges must obey
restrictive rule, because we have found no sound argument for that con-
clusion, and (2) accepting as a foundational principle, requiring no argu-
ment, that judges must obey restrictive rule.
One reason to prefer the first position is that the second seems to
impute to the law certain unexplained, primitive justificatory powers: the
law always overrides moral reasons to deviate in reasonably just legal
systems, we are told, but we are not told why. A defender of the second
position might respond that popular support for obeying restrictive rule
is so pervasive that we should simply accept it on faith. But this argument
ignores a crucial variable. The two positions also differ in their levels of
coherence with the rest of common morality and practical philosophy. I
think the first position dominates the second on this dimension. This is
not to say that the second position is less consistent than the first with
common morality and modern practical philosophy. The positions might
be comparably consistent in this respect. But consistency is only one
232 Limits of Legality: Part II

element of coherence.29 Another is the extent to which beliefs form mu-

tually supportive, justificatory connections with one another. The second
position has few justificatory connections to the rest of common morality
and practical philosophy. We are not told how any of our other beliefs
connects with the principle that judges must obey restrictive rule. By con-
trast, the first position connects adjudication to a range of plausible moral
judgments and principles. I have offered arguments for the conclusion
that judges in reasonably just legal systems have pro tanto reasons to
adhere in all cases and all-things-considered reasons to adhere in most.
The arguments of part II depend on premises that, although controversial,
have not been refuted. It is not unreasonable to accept them. The first
position contradicts common sense to some extent, but I think its revi-
sionism is more than compensated by its advantages in linking adjudica-
tion theory to the rest of common morality and practical philosophy.
I think the real choice we face is between (1) a theory that conflicts with
one popular belief about judicial duties, but makes more mutually sup-
portive connections to the rest of common morality and practical philos-
ophy, and (2) a theory that matches one popular belief, but makes fewer
such connections. I have tried to persuade you that the first position is at
least as reasonable as the second. Perhaps you will even come to prefer
the first, as I do.

29. See, e.g., Norman Daniels, Justice and Justification: Reflective Equilibrium in Theory
and Practice (Cambridge: Cambridge University Press, 1996); Michael R. DePaul, Balance
and Refinement: Beyond Coherence Methods of Moral Inquiry (New York: Routledge, 1993).
14

Guidance Rules

The previous chapter culminated with support for the following:

Combined Restriction
1. If Group O deviates as often as, or more often than, group restriction
permits, then deviate at no higher than the optimal average deviation rate.
2. Where n is the maximum number of cases in which group restriction
permits Group O to deviate during a certain time period and m is the
number of cases in which Group O actually deviates during that period, if
m < n, then do not deviate in more than n – m cases during that period.

Combined restriction simply imposes some constraints on deviation.

Many possible adherence rules are consistent with combined restriction.
We can now conjoin combined restriction with the imperative embodied
in the basic intention shared by Group O: to avoid suboptimal results.
The result is as follows:

Individual Policy
Deviating in suboptimal-result cases is permissible, subject to combined
restriction.

14.1 APPRAISAL AND GUIDANCE

Individual policy specifies criteria for the objective appraisal of decisions

made by a judge in Group O. It is designed for use by an outside observer
who wants to determine whether a member of Group O has produced a
decision sequence over the course of her career that is objectively permis-
sible, all things considered. If a judge makes decisions that conform to
individual policy, then her behavior is objectively permissible. This is not
to say anything about her conscious deliberation processes. Nor is it to
comment on her character: to say that she necessarily deserves praise.
Perhaps she made all her decisions randomly and they coincidentally
happen to conform to individual policy. Such a judge would deserve
blame for her irresponsible methods despite having performed objectively
permissible actions.

233
234 Limits of Legality: Part II

I think a good case can be made for individual policy as an objective

appraisal standard. It is a moderate rule, an intermediate position between
permissive rule and restrictive rule, and superior to both. But major ques-
tions remain. Could judges in realistic legal systems actually implement
individual policy? What guidance rules might they use?
Alan Goldman rejects the whole idea of trying to maintain the “opti-
mal level of deviation.” He concludes that every taxpayer must simply pay
his taxes and every judge must simply adhere to the law. I shall pause at
intervals to consider Goldman’s objections to optimizing. His first is epi-
stemic: “[I]ndividuals will not have a way of knowing how many other
individuals are reasoning in the same way in justifying deviation, how
close to the threshold of excessive deviation they may be. They will not
be able to coordinate their decisions to achieve the optimal level of devi-
ation.”1 Analogously, he thinks, each judge suffers “an inability . . . to iden-
tify the optimal pattern or to know how close to the threshold of damage
to the legal system he might be in not enforcing the law.”2
This objection might apply to taxpayers, who do not communicate to
one another their plans to reroute tax payments and cannot do so without
risking criminal penalties. But the decisions of judges are matters of public
record, extensively indexed and available for research. Judges can, in the-
ory, learn how many other judges are deviating. To some extent their job
even requires them to stay informed about these facts. They could
develop a sense of how close to threshold their system is. Goldman’s epi-
stemic argument succeeds only if one of the following two assumptions
holds:
1. Judges decide under conditions of uncertainty. They do not know
even the probability that their system is at its threshold; or
2. Judges decide under conditions of risk: they do not know whether
their system is at its threshold, but they know that the probability
is relatively high.
However, Goldman’s epistemic argument fails if
3. Judges decide under conditions of risk: they do not know whether
their system is at its threshold, but they know that the probability
is relatively low.
I submit that the third assumption describes realistic legal systems
more accurately than the first two. Goldman briefly considers this possi-
bility. His response consists of two assertions. The first is that “the negative
consequences of crossing [the threshold] are enormous.”3 I shall consider
this assertion before turning to the second. The consequences of crossing

1. Alan H. Goldman, Practical Rules: When We Need Them and When We Don’t
(Cambridge: Cambridge University Press, 2002), p. 44.
2. Ibid.
3. Ibid., pp. 44–45.
Guidance Rules 235

threshold are, by definition, bad. But attempting to optimize is rational,

nevertheless, if they are not too bad and if the risk of crossing threshold is
low enough. The negative consequences of crossing threshold are “enor-
mous” only if two conditions hold: first, crossing threshold dramatically
increases the number and/or severity of suboptimal results; second, hav-
ing crossed threshold, the system takes too long to return to subthreshold
equilibrium, or maybe never returns. If these two conditions obtain, then
crossing threshold is like falling off a cliff: we should stay far from the
edge. Whether these conditions obtain depends upon the shape of the
curve relating deviation density to mimetic failure. We can assume that it
slopes monotonically upward, but what else can we say about it?
If it is fairly straight and shallow, then there is no dramatic difference
between the system at one level of deviation density and at a slightly
higher level, even when the incremental increase pushes the system over
threshold. If Group O accidentally pushes the system over threshold,
then it can return to subthreshold equilibrium by slightly reducing its
own deviation rate.
Imagine, instead, that the curve is roughly shaped like a backward L: its
slope is shallow for awhile and then suddenly becomes steep. Imagine,
also, that threshold is located very close to the sharp upward turn. If
these conditions hold, then crossing threshold triggers a cascade of devia-
tion, including much deviation in optimal-result cases. Once the fact that
judges are deviating so often becomes common knowledge, it is difficult
to persuade judges to return to a more adherent pattern. Even if Group O
reduces its own deviation rate to previous levels, it takes too long to
restore subthreshold equilibrium. The system has fallen off the cliff.4
If the curve has a backward-L shape, then the argument for restrictive
rule is strengthened. I have no empirical data about the shape of the curve,
but there is potentially relevant social science: a body of research docu-
menting “informational cascades” and other forms of “herd behavior” and
“bandwagon effects” in various settings.5 Under certain conditions even
otherwise rational actors who “know better” will adopt the beliefs of
others and imitate their behavior. General convergence in the population

4. See the literature on the provision of collective step goods. Ramzi Suleiman, “Provi-
sion of Step-Level Public Goods under Uncertainty,” Rationality and Society 9 (1997):
163–87; Jean Hampton, “Free-Rider Problems in the Production of Collective Goods,” Eco-
nomics and Philosophy 3 (1987): 245–73.
5. See, e.g., Timur Kuran and Cass R. Sunstein, “Availability Cascades and Risk Regula-
tion,” Stanford Law Review 51 (1999): 683–768; Sushil Bikhchandani, David Hirshleifer, and
Ivo Welch, “Learning from the Behavior of Others: Conformity, Fads, and Informational
Cascades,” Journal of Economic Perspectives 12 (1998): 151–70; Lisa R. Anderson and Charles
A. Holt, “Information Cascades in the Laboratory,” American Economic Review 87 (1997):
847–62; Abhijit Banerjee, “A Simple Model of Herd Behavior,” Quarterly Journal of Eco-
nomics 107 (1992): 797–818; Sushil Bikhchandani, David Hirshleifer and Ivo Welch, “A
Theory of Fads, Fashion, Custom and Cultural Change as Informational Cascades,” Journal
of Political Economy 100 (1992): 992–1026.
236 Limits of Legality: Part II

can occur rapidly, like a stampede. A few scholars have applied cascade
theory to judicial behavior, but they have concentrated on precedent, not
mimesis.6 The backward-L hypothesis remains speculative, but it could be
true. If so, then judges who attempt to optimize risk falling off the cliff. I
shall assume arguendo that this risk is real.
I note, however, that Goldman provides no reason to assume that
threshold falls close to the point at which the curve begins its sharp
upward slope. I shall assume arguendo that it does—that crossing thresh-
old runs an unacceptably high risk of destroying the rule of law indefi-
nitely. If this assumption proves false, then my argument for moderate
rules becomes more compelling.
This brings me to Goldman’s second objection to optimizing: “a judge
can assume that, in the absence of a rule, she or one of her colleagues
will [cross the threshold].”7 If the judges in Group O deviate in all sub-
optimal-result cases, with no regard for threshold, then it is probable
that the judiciary will eventually cross threshold. If that is what “absence
of a rule” means, then Goldman is correct. But he treats restrictive rule
as the only alternative to no rule at all. Judges can guard against falling
off the cliff by obeying moderate rules.8 If the members of Group O
conform to individual policy, then they deviate in some but not all sub-
optimal-result cases. No one crosses threshold. I conclude that Gold-
man’s objection to optimizing does not reach judges who conform to
individual policy. Even if crossing threshold destroys the rule of law, in-
definitely, judges who conform to individual policy do not push their
system over that cliff.
My argument in this section has assumed that judges internalize indi-
vidual policy—they use it as their guidance rule. Real judges could, in
theory, do this. In practice, however, it may be a terrible prescription, for
reasons that mirror the well-known shortcomings of act utilitarianism as
a decision procedure.9 Much as utility may not be maximized if every
individual attempts to maximize utility, so the goals of individual policy
may be achieved less effectively if judges try to apply it directly than if
they follow another guidance rule. Accurately applying individual policy
requires extensive gathering of information and deliberating. It requires
each judge to determine how often other judges are deviating relative to
the maximum rate permitted by group restriction. It requires judges to
calculate the proximity of current deviation density to threshold, to ascer-
tain what standards their colleagues have adopted to regulate their own
deviation, and to determine in which deviation enterprises they would

6. Cass R. Sunstein, Why Societies Need Dissent (Cambridge, Mass.: Harvard University
Press, 2003), pp. 59–60; Eric Talley, “Precedential Cascades: An Appraisal,” Southern Califor-
nia Law Review 73 (1999): 87–137.
7. Goldman, Practical Rules, pp. 44–45.
8. Goldman rejects moderate rules for moral reasons that I shall discuss in chapter 15.
9. See, e.g., J. L. Mackie, Ethics: Inventing Right and Wrong (Harmondsworth: Penguin
Books, 1977), p. 129.
Guidance Rules 237

participate were they to deviate in a given case. Judges who attempt to

follow individual policy as a guidance rule will rarely succeed in approxi-
mating the optimal decision pattern. This observation does not assume
that the judges to whom guidance rules are addressed are imperfect iden-
tifiers of suboptimal-result cases. I am still discussing the judges in Group
O. My reservations about individual policy merely assume that directly
determining the location of threshold and the current level of deviation
density are extremely difficult and time-consuming tasks. The research
takes time, effort, and resources that judges should not be expected to
expend. If we asked judges to use individual policy as a guidance rule,
then we could not blame them for failing to approximate the ideal decision
pattern because reasonably competent agents with their beliefs and infor-
mation would perform no better. Thus, individual policy makes a poor
guidance rule and I would not urge judges to adopt it. I merely maintain
that individual policy is the objectively correct appraisal policy for pat-
terns of judicial decision. The best guidance rules are those that most
closely approximate the patterns approved by individual policy. These
rules I have yet to identify.

14.2 SIMPLER GUIDANCE RULES

We can formulate simpler guidance rules for implementing individual

policy. Consider two old friends: restrictive rule and permissive rule. Nei-
ther assigns any relevance to deviation density. Although I have argued
that neither succeeds as an appraisal rule, I have not yet considered their
potential as guidance rules.
Permissive rule fares no better as a guidance rule than it did as an ap-
praisal rule. In real legal systems, deviating in all suboptimal-result cases
probably exceeds threshold. Restrictive rule, by contrast, forbids all devi-
ation. It has some definite advantages over individual policy as a guidance
rule: it is simple, easy to learn, and relatively easy to follow.10 Obeying
restrictive rule guarantees that the two conditions of individual policy are
met. It keeps the system below threshold, so it is at least a candidate for a
viable guidance rule. However, restrictive rule also has substantial moral
disadvantages, discussed earlier,11 that carry over to its employment as a
guidance rule. Restrictive rule forbids judges to attain the optimal decision
pattern in any legal system in which the optimal average deviation rate is
greater than zero, which is all realistic legal systems. Restrictive rule
requires judges to use pro tanto unjustified force, or threats thereof, in an
unnecessarily large number of suboptimal-result cases. The higher the
optimal average deviation rate in the system, the greater becomes the

10. I am not sure whether restrictive rule is easier to follow than permissive rule or vice
versa, but I need not address that question.
11. See §§8.4–8.6.
238 Limits of Legality: Part II

divergence between the optimal pattern and the best patterns permitted
by restrictive rule.

14.3 BETTER GUIDANCE RULES

I shall now evaluate candidates for a guidance rule better than restrictive
rule—one that allows judges to deviate at closer to the optimal rate. Con-
sider a rule slightly more permissive than restrictive rule:
First Moderate Rule: Over the course of your judicial career, deviate in no more
than 1 percent of suboptimal-result cases.

First moderate rule is more difficult to apply than restrictive rule, but
still manageable. It is much easier to apply than individual policy. First
moderate rule relieves judges of the burden of monitoring the deviation
rate of other judges. Instead, it merely requires each judge to monitor her
own deviation rate over the course of his career. It makes the permissibil-
ity of deviation in a given case by a certain judge partly depend upon the
number of other cases in which she has already deviated, or anticipates
deviating. First moderate rule seems like an improvement on restrictive
rule, at least for legal systems that are well below threshold. Deviating in
1 percent of suboptimal-result cases will not push any such system above
threshold.
First moderate rule answers Goldman’s epistemic objection. Interest-
ingly, with respect to prudential reasoning by individuals in their private
lives, Goldman supports a position similar to first moderate rule. He
observes that
we can more easily solve the coordination problem over time for our own
actions than we can coordinate with the actions of others in the relevant moral
contexts. This ability to coordinate actions over time allows the adoption of
strategies to optimize instead of settling for the second-best strategy of strict
rules. We can have the occasional snack or martini.12
Goldman’s reference to the “occasional snack or martini” shows that he
accepts a prudential, intrapersonal counterpart to first moderate rule.
However, he sees moral and epistemic problems in the interpersonal case
that do not arise in the intrapersonal case. I shall argue that these prob-
lems are not fatal in the interpersonal case, either.
An obvious objection is that first moderate rule provides no basis for
selecting the 1 percent. It seems to invite judges to use arbitrary criteria.
This is a good objection. When allocating a scarce, indivisible privilege to
one of several deserving beneficiaries, the allocator is not necessarily per-
mitted to choose arbitrarily if the beneficiaries differ on morally relevant
criteria. For example, if every potential beneficiary deserves or needs the

12. Goldman, Practical Rules, p. 6.

Guidance Rules 239

privilege, but one is more deserving or needy than the others, then the
allocator may be morally obligated to favor the most deserving or needy,
as in emergencies calling for medical triage. First moderate rule allows
judges to violate this principle. Fortunately, we can revise first moderate
rule to correct this problem:
Rebuttable Presumption: Adhere in suboptimal-result cases, unless the law
requires an extremely suboptimal result.

In realistic legal systems, extreme suboptimal-result cases are relatively

rare, comprising less than 1 percent of suboptimal-result cases (my esti-
mate). In such systems, the members of Group O can obey rebuttable
presumption without pushing the system above threshold. As does first
moderate rule, rebuttable presumption allows for occasional deviation,
but unlike first moderate rule, it provides a nonarbitrary criterion for
selecting cases for deviation. Rebuttable presumption requires judges to
choose their “1 percent” on a morally appropriate basis.
Rebuttable presumption captures a familiar idea.13 I have mentioned
that even writers who deny that it is ethically permissible for a judge to
deviate from the law simpliciter often make exceptions for extremely
suboptimal results.14 Faced with cases involving slavery laws in the
American South, anti-Semitic laws in Nazi Germany, the laws of the
Taliban in Afghanistan, or the laws sustaining South African apartheid,
all but the most committed supporters of restrictive rule tend to admit
that they really favor only something like rebuttable presumption. A
judge is morally permitted to deviate from a Nazi law that forbids har-
boring Jews, even if his fellow judges uphold it faithfully. By deviating,
he rides free on his fellow judges, but he is justified in doing so because
the law requires an extremely unjust result. Goldman actually implies,
occasionally, that judges should obey rebuttable presumption, not re-
strictive rule.15
Rebuttable presumption will remind some readers of the Radbruch
formula, which denies legal validity to extremely unjust positive laws.16
Both the Radbruch formula and rebuttable presumption entail that judges
are not morally obligated to adhere to extremely unjust laws. However,
rebuttable presumption does not impugn the legal validity of extremely
unjust laws, as does the Radbruch formula. Also, although rebuttable

13. “[I]n a flourishing legal system the fact of law provides a case for coercion that must
stand unless some exceptional counterargument is available.” Ronald Dworkin, Law’s Empire
(Cambridge, Mass.: Harvard University Press, 1986), p. 110. See also Gerald J. Postema,
Bentham and the Common Law Tradition (Oxford: Oxford University Press, 1986), ch. 12.
14. See §8.4.
15. He states that judges “can ignore recognized legal requirements only in the direst of
circumstances” (p. 148) and that they should not treat like cases differently “except in the
direst of circumstances that would call the validity of the legal system itself into question”
(p. 46). Goldman, Practical Rules.
16. See §8.4.
240 Limits of Legality: Part II

presumption applies to both suboptimal-rule cases and negative-gap cases,

the Radbruch formula might not apply to the latter.
One might doubt whether rebuttable presumption even constitutes a
serious rule, because applying it seems to require moral judgment con-
cerning whether a certain result is extremely suboptimal. But it could
work as a serious rule if there were sufficient consensus regarding which
results are extremely suboptimal. Ex hypothesi such consensus prevails in
Group O. However, obeying rebuttable presumption as a serious rule is
equivalent to treating restrictive rule as something less than a serious rule.
It is equivalent to acting as a “rule-sensitive particularist” with respect to
restrictive rule. Another variation of rebuttable presumption would be
akin to taking a “presumptive positivist” approach to restrictive rule,17
directing judges to obey restrictive rule, unless they just happen to
perceive especially strong reasons to deviate.
As I mentioned, rebuttable presumption makes sense in reasonably just
systems: it is a permissible implementation of individual policy. However,
there are conceivable legal systems so unjust that rebuttable presumption
permits too much deviation in them. Imagine a horribly unjust system in
which the law requires so many extremely suboptimal results that if
Group O deviates in all of them, then it pushes the system past threshold.
In such systems, rebuttable presumption is the wrong adherence rule.18 In
more just legal systems, by contrast, rebuttable presumption is acceptable.
In such systems Group O can safely obey rebuttable presumption with-
out pushing the system past threshold.
Applying rebuttable presumption requires some sense of what makes
a result extremely suboptimal as opposed to just moderately suboptimal.
This assumes that some results are more suboptimal than others and that
people can distinguish between extremely and moderately suboptimal.
Returning a slave to his master, for example, is an extremely suboptimal
result. Someone who believes this might also believe that preventing a
university from giving preferential admission to African-Americans,
although a suboptimal result, is not extremely suboptimal. I have been
assuming that the extension of extremely suboptimal is fixed independently
of facts about threshold or the types of cases heard in any particular
system. A result is extremely suboptimal if and only if it involves a degree
of injustice, unfairness, or undeserved misfortune that is very high, in
absolute terms. Because such cases are very rare in reasonably just systems,
rebuttable presumption happens to be an acceptable implementation of
individual policy in such systems. But rebuttable presumption is not spe-
cifically calibrated to any particular system. Therefore, rebuttable pre-
sumption is not the best implementation of individual policy in all legal

17. Frederick Schauer, Playing by the Rules: A Philosophical Examination of Rule-Based

Decision-Making in Law and in Life (Oxford: Oxford University Press, 1991), pp. 196–206.
18. Unless, of course, the system is so unjust that the rule of law should, morally speak-
ing, be destroyed there.
Guidance Rules 241

systems. Some systems are sufficiently just that Group O can deviate in
more cases than rebuttable presumption permits, without exceeding
threshold. It remains an open question whether modern legal systems
meet this condition. If they do, then rebuttable presumption is needlessly
restrictive for those systems. It would be desirable to formulate a more
precisely tailored rule than rebuttable presumption—one that allows
judges to approach threshold more closely without going over. In the next
section, I present an analogy meant to illustrate the irrational restrictive-
ness of rebuttable presumption in reasonably just systems.

14.4 EMISSIONS TRADING ANALOGY

Think of a deviant decision as analogous to an act that releases into the

environment a very small quantity of a toxin that, in much larger quan-
tities, unjustifiably harms identifiable individuals to a perceptible extent.19
This is like mimetic failure. Suppose the toxin is released as the undesired
byproduct of a process that itself prevents societal harm (e.g., the manu-
facture of safety devices or beneficial pharmaceuticals). Now picture two
different environmental policies. Both are versions of cap-and-trade envi-
ronmental regulations. These policies efficiently limit the aggregate level
of pollution by allocating pollution credits to firms and allowing firms that
pollute less to trade their credits to firms that pollute more.
Standard cap-and-trade policies allow firms to purchase any quantity
of credits and spend them as they wish, either gradually or all at once. Call
this a no-minimum policy. It does not attempt to ensure that each firm
expels even approximately the same quantity of pollution. The state does
not monitor whether any given firm is profitable, either. Rather, the state
creates market incentives that induce firms to respond rationally to the
overall level of pollution. As the overall level of pollution increases, credits
are spent and the remaining credits command commensurately higher
prices in trade. It makes little difference if one firm buys most of the
credits and pollutes extensively or if the credits are spread across many

19. Stephen Burton draws a similar analogy:

The political common ground is like an economic commons that benefits the commu-
nity only as long as everyone acts with self-restraint. Any one person, however, can be
made better off by acting opportunistically in his or her individual or subgroup self-
interest if everyone does not do so and thereby bring down the cooperative scheme
altogether. So some people will try in any case quietly to take a piece of the common
ground for themselves while continuing to benefit from the general scheme of order and
justice that prevails nonetheless. When too many get away with it the common ground
may give way to a repellent kind of politics that will strike most as a degeneration of the
society.

Stephen J. Burton, Judging in Good Faith (Cambridge: Cambridge University Press,

1992), p. 258.
242 Limits of Legality: Part II

firms, each of which pollutes modestly. Overall pollution levels matter;

the identities of the polluting firms do not.
This is the way real cap-and-trade policies work.20 However, we can
imagine a bizarre alternative policy that requires firms to choose between
(1) buying no credits and (2) buying and spending a large minimum
number of credits all at once. Call this a minimum-purchase policy. A
minimum-purchase policy might be acceptable in a society given the fol-
lowing scenario. Most of the time, most of the many firms in this society
have no opportunity to prevent even small degrees of harm (thereby
earning profits) by engaging in harm-preventing but polluting processes.
Therefore, they have no reason to pollute. They buy no pollution credits.
On rare occasions, however, one of these firms or another has the oppor-
tunity to prevent great harm, thereby earning large profits, although doing
so will produce high levels of pollution all at once. When a firm finds itself
in this rare position, it purchases the prescribed minimum credits and
spends them all at once.
Although a minimum-purchase policy is acceptable in such a society,
it is excessively restrictive for one in which opportunities to prevent small
quantities of harm, by means of moderately polluting processes, are dis-
tributed widely across the firms. In such a society the no-minimum policy
is superior. It allows firms to purchase credits in a continuous range of
quantities. Each firm can tailor its purchases to its individual needs and
opportunities.
Now apply the analogy back to the legal system. A market in which a
single firm buys a large bundle of credits and spends them all at once is
analogous to a legal system in which a single judge deviates from a law
that requires an extremely suboptimal result, but she thereby causes sub-
stantial mimetic failure. For some reason her decision leads many other
judges to deviate in optimal-result cases. By contrast, a market in which
the same credits are spread more evenly across many firms is analogous to
a legal system in which many judges deviate at various times from laws
that require moderately suboptimal results, each contributing modestly to
an overall degree of mimetic failure.
Rebuttable presumption is analogous to the minimum-purchase pol-
icy. It permits a judge to deviate from the law only if she must in order to
avoid an extremely suboptimal result. This rule may be appropriate to
very unjust legal systems in which a single deviant decision can prevent an
extreme injustice. The prerogative is useless, however, to a judge in a rea-
sonably just legal system who may never hear an extreme suboptimal-
result case in her entire career. Rather, she will hear hundreds of cases in
which the law requires moderately suboptimal results. Preventing the

20. These policies, as implemented in the real world, invite objections from distributive
justice and expressive meaning that I shall not address. See Mark Sagoff, “Controlling Global
Climate: The Debate over Pollution Trading,” in Philosophical Dimensions of Public Policy,
ed. Verna V. Gehring (New Brunswick, N.J.: Transaction, 2002).
Guidance Rules 243

same “amount” of injustice as the judge in the highly unjust system pre-
vented with a single decision would require her to deviate in many of
these moderate cases. She could do so without contributing to any enter-
prise that will push her system past threshold. Unfortunately, rebuttable
presumption forbids her from doing so. As a result, she reaches many
moderately suboptimal results that she could have avoided without
undermining the rule of law had she obeyed a better adherence rule.

14.5 IMPROVING UPON THE REBUTTABLE PRESUMPTION

Is there an alternative to rebuttable presumption that allows judges in

reasonably just systems more closely to approximate individual policy?
The problem with rebuttable presumption is that it is not calibrated to
the system’s optimal deviation rate. We need a rule that is properly cali-
brated while still feasible to apply. Suppose we determine a value for n
that is the optimal average rate of deviation in suboptimal-result cases for
all judges in Group O. Then we can formulate the following:
Final Moderate Rule: Over the course of your judicial career, deviate in no more
than n percent of suboptimal-result cases.

Final moderate rule improves upon rebuttable presumption. I think a

good implementation of individual policy will incorporate final moder-
ate rule. By itself, however, it has the same flaw as first moderate rule. A
typical judge in an actual legal system might decide a much larger
number of suboptimal-result cases than the total number in which final
moderate rule permits deviation. Final moderate rule provides no crite-
ria for selecting which cases will comprise the n percent. So judges need
additional standards for deciding between two suboptimal-result cases
when final moderate rule permits deviation in only one of them. These
standards could take the form of priority rules. Our first priority rule
recalls rebuttable presumption, in that it assigns relevance to degrees of
suboptimality:

First Priority Rule: If final moderate rule permits you to deviate in only one of
two suboptimal-result cases, then deviate in the one in which the legally
required result is more suboptimal.

First priority rule ensures that the n percent will comprise the more
serious of the suboptimal-result cases that arise in the system. Notice that
this is not the same as limiting deviation to “extreme” suboptimal-result
cases as rebuttable presumption does. First priority rule specifies no abso-
lute level of suboptimality as a floor. The maximum percentage of subop-
timal-result cases in which moderate rule permits deviation is determined
by n. The value of n varies directly with the overall level of justice that
prevails in the legal system. In an extremely just system there are only a
few suboptimal-result cases and the legally required results are only
244 Limits of Legality: Part II

slightly suboptimal, yet first priority rule permits judges to deviate

in most of them. At the limit there are no suboptimal-result cases and
n = 100.21
First priority rule instructs judges to give higher priority to avoiding a
result, the more suboptimal it is. Ignoring for the moment the objection
that judges should not deviate in the first place, this is an intuitive idea.
The opposite instruction would give priority to avoiding less suboptimal
results. At the limit it would instruct judges to deviate in optimal-result
cases. This view makes no sense. Somewhat more defensible is the view
that judges should give no priority based on degrees of suboptimality. On
this view judges should not treat the fact that the law requires a more
suboptimal result in one case than in another as a reason to deviate in the
first case rather than the second. Perhaps some people accept this view,
but it is hard to defend. The argument would appear to rest upon the
assumption that the systemic effects of deviating, which vary from one
case to another, fluctuate directly with the suboptimality of the legally
required result. The idea would be that deviating in order to avoid a more
suboptimal result causes more systemic damage than does deviating in
order to avoid a less suboptimal result. If that were true, then the rela-
tively greater benefits of deviating in a more extreme suboptimal-result
case would be cancelled out by the relatively greater systemic damage.
However, there is no reason to believe in a direct correlation between
suboptimality and systemic effects. This is not to claim that the systemic
effects of deviation are constant across all cases. Some cases naturally
receive more attention than others and provoke more mimetic failure.
But there is no reason to think that these are necessarily cases in which
the avoided result was more suboptimal, or more suboptimal than the
judge believed it to be. If anything, one might predict that deviating causes
more mimetic failure the less suboptimal the avoided result is perceived
to be: the more subtle the argument for deviation, the more easily is it
misconstrued by posterior judges.

14.6 TYPE-SELECTIVE DEVIATION

I suggested that judges in Group O can move their system toward thresh-
old, without going over, by obeying first priority rule in addition to final
moderate rule. First priority rule provides some basis for choosing in
which suboptimal-result cases to deviate. But first priority rule does not
entirely solve the arbitrariness problem. First priority rule permits judges
to deviate in n percent of suboptimal-result cases and to give priority to
more serious suboptimal-result cases, but it does not require a judge who

21. Obviously, in that scenario there are no suboptimal-result cases in which to deviate.
One hundred percent of zero is zero.
Guidance Rules 245

deviates in one case to deviate in every other case at the same level of
suboptimality. The following story illustrates.
Your friend, Rita, is a federal district judge. Once a month you visit her
courtroom. In July she hears a free speech challenge to a statute making
it a misdemeanor to engage in certain kinds of bias-motivated speech. A
higher court has ruled that a similar statute violates the Free Speech
Clause.22 The defendant moves for dismissal for that reason, but Judge
Rita ignores the precedent and applies the statute. You ask her privately if
she believes that the precedent did not apply to the facts of the case. Was
there some legally relevant difference between the statute challenged and
the statute that was invalidated by the higher court? Rita confides in you
that she sees no such difference. The precedent clearly applies to the
recent case, but she fundamentally disagrees with the higher court’s
decision. She believes that hate speech should be subject to such regula-
tions. That is why she ignored the precedent.
You return to the courtroom in August to find Judge Rita hearing
another free speech challenge, this one to a statute that makes it a mis-
demeanor to engage in nude dancing. The higher court has ruled that a
similar statute violates the Free Speech Clause.23 Rita cites this precedent
as authority and rules the statute unconstitutional.
You privately question the judge. Did the second free speech prece-
dent actually control her reasoning, as the language in her opinion sug-
gests, or did she invalidate the nude dancing statute because she disapproves
of it? She replies that the precedent controlled her reasoning, although
she disagrees with it. She believes that nude dancing should be subject to
regulation. She approves of the challenged statute and believes it to be
consistent with a proper reading of the First Amendment.
So why did she follow precedent in the second case but not the first?
Does she find nude dancing intrinsically less eligible for regulation than
hate speech? No, she replies, she finds the two practices equally eligible
for regulation. So what was her reasoning?
Rita explains that, in a certain sense, she invalidated the nude dancing
statute because she upheld the hate speech statute. She followed prece-
dent in the second case because she disregarded it in the first. In a world
without systemic effects, she gladly would have upheld both statutes. But
every month she hears cases in which higher-court precedent requires her
to reach results that she considers suboptimal. If judges such as she were
to disregard too many precedents with which they disagree, then less
“wise” judges would begin disregarding precedents with which they dis-
agree and many of these would be good precedents. Therefore judges (wise
or not) must limit themselves to deviating in a certain percentage of cases

22. Compare R.A.V. v. City of St. Paul, 505 U.S. 377 (1992).
23. As if the opposite result had been reached in Barnes v. Glen Theatre, Inc., 501 U.S.
560 (1991).
246 Limits of Legality: Part II

and no more. Rita limits herself to deviating in no more than 15 percent

of the cases in which the law requires her to reach a result that is (in her
opinion) suboptimal.
First priority rule permits judges to practice type-selective deviation—
deviating in some, but not all, suboptimal-result cases at a certain level of
suboptimality. This is the basis for a major objection to first priority rule
and a defense of restrictive rule: the objection from comparative justice or
“treating like cases alike.” I think the comparative justice objection to
type-selective deviation can be answered. As I shall explain, individual
policy permits judges, in principle, to practice an even more surprising
kind of selectivity: token-selective deviation. The comparative justice argu-
ment has its greatest force against this kind of selectivity. I shall challenge
the argument even as an objection to token-selective deviation. If it fails
there, then it fails a fortiori with respect to type-selective deviation.24

14.7 SUBOPTIMAL-RULE CASES

Assume that n = 15, and picture Judge Rita hearing a series of 100 subop-
timal-result cases. Fifteen of these are cases in which the law requires an
extremely suboptimal result: high-level cases. In remaining 85, the law
requires moderately suboptimal results: these are low-level cases. Judge
Rita can deviate in all 15 of the high-level cases and adhere in the remain-
ing low-level cases. This result conforms to final moderate rule and first
priority rule. It does not require her to choose arbitrarily.
However, first priority rule tolerates a different kind of arbitrariness
that is almost as troubling. Other judges have not deviated and will not
deviate in every high-level case. In fact, they have adhered and will adhere
in case tokens very similar to those in which Rita deviates. Rita can point
to no features of the tokens in which she deviates that justify deviating in
them but not in other tokens of the same type, nor in suboptimal-result
cases of a different type at the same level of suboptimality.25
Furthermore, Rita got lucky. She heard exactly the right number of
high-level cases. What if she hears fewer? We can depict the case tokens
in a legal system as a layered pyramid. Each layer represents a set of cases
at the same level of suboptimality. The top layer represents the most
extreme suboptimal-result cases. The area of each layer represents the
number of cases at that level. Picture a system represented by a pyramid
with only two layers: a tiny top layer and a large bottom layer. In this
system, 5 of the suboptimal-result cases are high-level cases, while the
remaining 95 are low-level cases. None of the low-level cases is noticeably

24. See §15.2.

25. C. S. Peirce first used the terms type and token to mark such distinctions. Charles S.
Peirce, Collected Papers of Charles Sanders Peirce, eds. Charles Hartshorne and Paul Weiss
(Cambridge, Mass.: Harvard University Press, 1932), §4.537.
Guidance Rules 247

more or less suboptimal than the other low-level cases. That is why they
are represented by a single layer.
Rita can still deviate in the 5 high-level cases, but beyond that she
faces a conundrum. If she deviates in all 95 of the low-level cases, then she
violates first priority rule, because n = 15. First priority rule permits her to
adhere in all 10 of the low-level cases that she hears, but if she does so
then she reaches 10 unnecessary suboptimal results. Alternatively, first
priority rule permits her to deviate in 10 of the low-level cases, adhering
in the rest. But first priority rule provides no reason to choose any partic-
ular 10 from the 95. The arbitrariness problem recurs.
It would be desirable to find further criteria for choosing between sub-
optimal-result cases at the same level. I think there are such criteria. In
this section I entertain the idea that judges should give some deviation
priority to suboptimal-rule cases over gap cases, assuming equally subop-
timal results. The limits of this priority are decision patterns involving
deviation only in suboptimal-rule cases but never in gap cases. One might
interpret the claim of the natural law tradition that “unjust laws do not
bind in conscience” as requiring such a pattern.26
Consider two criminal statutes: one criminalizes fornication, whereas
the other imposes strict liability for possessing 500 grams of heroin.
Each carries a $500 fine. Assume that fornication is not so immoral as to
merit this fine and that the fornication law serves no other adequate
public purpose. Therefore, virtually all convictions under that law are
suboptimal results. Assume, also, that heroin possessors usually deserve
to be convicted, so that only a few convictions under the heroin law
are suboptimal, as in the rare case when a defendant receives a box
containing heroin, unbeknownst to him.27 However, a conviction in
that gap case is just as suboptimal as is the typical conviction under
the fornication law.
Do judges have a stronger reason to deviate in a suboptimal-rule case,
such as one brought under the fornication law, than in a gap case under
the heroin law? The reason to deviate in a suboptimal-rule case is not in-
trinsically stronger than the reason to deviate in a gap case if the law in
the latter requires a comparably suboptimal result. However, imagine a
judge who must choose between (1) deviating in m gap cases, each under
different suboptimal rules, and (2) deviating in m cases under one subop-
timal rule. I think she has reasons to choose to deviate in the cases falling
under the single rule. This idea is captured by the following:
Second Priority Rule: If final moderate rule and first priority rule permit devia-
tion in only one of two sets, each composed of the same number of cases,
where the members of the first set fall under more than one rule and the

26. Thomas Aquinas, Summa Theologiae, trans. Fathers of the English Dominican Prov-
ince (London: Washbourne, 1915), Question 96.
27. This defendant could still be convicted under a strict liability statute.
248 Limits of Legality: Part II

members of the second set fall under a single rule, and are otherwise as apt for
deviation as are the members of the first set (regarding suboptimality, etc.),
then deviate in the second set.

Second priority rule is consistent with some deviation priority for sub-
optimal-rule cases. It does not, however, entail that judges must never
deviate in negative-gap cases. Recall Rita’s conundrum. It is statistically
likely that fewer than 10 of the remaining high-level cases are subopti-
mal-rule cases under the same rule. If Rita does not hear as many subop-
timal-rule cases as the overall number in which second priority rule
permits her to deviate, then she is permitted to deviate in some negative-
gap cases until she reaches her permitted deviation allotment, whatever
that may be.
Second priority rule instructs judges to give deviation priority to sub-
optimal-rule cases over gap cases. This is an intuitive idea. By definition a
suboptimal rule can be improved. By deviating from a suboptimal rule, a
judge paves the way for improvements, judicial or legislative. This is obvi-
ous in cases in which the judge has legal authority to improve the rule,
but it is true even in cases in which the rule could be improved although
the judge lacks legal authority to improve it, as when a superior court has
already ruled out the improvement. A judge who improves the rule by
departing from vertical precedent deviates from the law, but she points
the way to a more coherent and optimal set of legal standards from which
deviation will be less often necessary.
By contrast, a judge who deviates in a gap case does not point the way
to a more coherent and optimal set of legal standards. In a gap case, by
definition, the controlling legal standards cannot be improved as a prac-
tical matter, yet they still dictate suboptimal results in some cases. A
judge who deviates in a gap case does not suggest any superior legal stan-
dard. He does not propose a new exception that should be appended to
existing standards. He simply deviates in order to avoid a suboptimal
result. This can be morally permissible if my argument is correct. It pro-
tects the legally disfavored party from unjustified disadvantages and it
might inspire posterior judges to deviate in other suboptimal-result cases.
But it does nothing to improve legal standards themselves. By contrast,
deviating properly in a suboptimal-rule case benefits the legally disfa-
vored party and also makes a potential contribution to an improvement
in legal standards, possibly benefiting future parties in this additional way.
Ceteris paribus, judges should give priority to suboptimal-rule cases over
gap cases.
Consider cases that arise under the following conditions. Under exist-
ing law, each of a class of persons is suffering or will suffer a suboptimal
result. Changing the law would protect these persons. However, if the law
were changed to protect everyone in the class, then this would have neg-
ative effects on balance. The law would not have these negative effects if
it were changed to protect some but not all members of the class, but
Guidance Rules 249

there are no politically feasible criteria for the law to use to select the
beneficiaries.
I classify such cases as gap cases even though, in theory, a better rule is
possible. A judge who hears such a case should not attempt to change the
law because such attempts are futile. Ex hypothesi there are no politically
feasible criteria for selecting the beneficiaries. If he announces a new legal
rule that selects the right number of beneficiaries, then his decision will
be reversed by a higher court, overridden by the political branches, or
simply ignored by other judges. He will fail to change the law. The most
he can realistically hope to accomplish is an optimal result in the case at
bar. So the difference between a suboptimal-rule case and a gap case is
that a judge who deviates and announces a new rule in a suboptimal-rule
case has a better chance of improving the law than a judge who does the
same in a gap case.

14.8 BLAME AND WARRANT

I have defended guidance rules that permit deviation in some suboptimal-

result cases. I have stipulated that the judges in Group O always accu-
rately identify suboptimal-result cases. By obeying my proposed rules,
they can be very confident that their decisions are objectively permissible,
given that they accurately identify suboptimal-result cases. However, they
probably want to avoid subjectively impermissible actions—blameworthy
actions—as well as impermissible ones. Even if a judge is objectively per-
mitted to deviate, she is blameworthy if she deviates without having ade-
quate evidence that the deviant result is at least not normatively inferior
to the legally required result. She is like a motorist who drives with
severely impaired vision, but gets lucky and hits no one.
Judges know that reasonable citizens in pluralistic societies disagree
about many moral issues. They know that even competent judges reach
some morally mistaken conclusions. Therefore, before she knowingly
deviates for the first time, any judge who wishes to avoid blameworthy
decisions will want to know how strong her evidence of suboptimality
must be if she is to remain blameless.28 The following argument
elaborates:

1. An agent is blameworthy if she knowingly performs an action that

she does not reasonably believe to have a probability below y of
being objectively wrong.
2. Judges cannot reasonably deny that reaching suboptimal results is
objectively wrong.

28. I am referring to the concern that blame would be justified, not that some actual
observer will blame her.
250 Limits of Legality: Part II

3. Therefore, a judge is blameworthy if she knowingly deviates in

order to avoid a result that she believes to be suboptimal and she
does not reasonably believe that the probability is below y that she
is mistaken.
What is the value of y? In other words, how subjectively confident in
her judgments of suboptimality must a judge be before she deviates?29
How confident should we require a judge to be before she can deviate
without risking our condemnation? These are difficult questions. Fortu-
nately, legal systems deal constantly with questions of probability and
confidence in judgment. A judge who concludes that a result is subopti-
mal can always ask himself whether his level of confidence satisfies a spec-
ified evidentiary standard. In Anglo-American law the most familiar
standards are preponderance of the evidence, clear and convincing evi-
dence, and reasonable doubt. Judges could use one of these standards, or
something similar, to determine whether they are sufficiently confident
that a result is suboptimal.
First consider a reasonable-doubt standard. Some might argue that
judges who used this standard in reasonably just legal systems would
never deviate in socially controversial cases, because they could not have
sufficient confidence in the suboptimality of any result in such cases. Con-
sider controversies about economic justice, environmental policy, affirma-
tive action, capital punishment, abortion, assisted suicide, prostitution,
recreational drugs, et cetera. If, as philosophers often suggest, “reasonable
people” disagree about such issues, then one might claim that no one can
ever believe “beyond a reasonable doubt” that his own position is correct.
If so, and if the proper evidentiary standard is reasonable doubt, then
judges who deviate in such cases are always blameworthy even when their
decisions are objectively correct as a moral matter. Therefore, they should
never deviate in such cases.
I dispute the major premise of this argument: that if reasonable people
disagree about an issue, then one cannot have confidence beyond a rea-
sonable doubt in one’s own position. Judge John T. Noonan, Jr., a federal
appellate judge, believes that abortion is as objectively immoral as infan-
ticide, ceteris paribus. He believes that criminalizing it is morally obliga-
tory and that the right to abort is not protected by the U.S. Constitution,
properly understood.30 He knows that reasonable people disagree with

29. See the discussion of the epistemic threshold in Douglas E. Edlin, Judges and Unjust
Laws: Common Law Constitutionalism and the Foundations of Judicial Review (Ann Arbor:
University of Michigan Press, 2008), pp. 139–49. See also Stephen R. Perry, “Second-Order
Reasons, Uncertainty, and Legal Theory,” Southern California Law Review 62 (1989): 913–
94, p. 967.
30. John T. Noonan, Jr., Persons and Masks of the Law: Cardozo, Holmes, Jefferson,
Holmes, and Wythe as Makers of the Masks (New York: Farrar, Straus & Giroux, 1976).
Guidance Rules 251

him, but their general reasonableness does not make their views on this
topic reasonable. I think Judge Noonan could consistently claim that he is
confident beyond a reasonable doubt that abortion should be criminal-
ized, knowing that many reasonable people disagree with him.31 This is
not, of course, to imply that the esteemed Judge Noonan would actually
deviate from the law. Nor is my point that pro-life judges have better
reasons than pro-choice judges to claim that they have no reasonable
doubts about the correctness of their respective positions. Justice Ruth
Bader Ginsburg may be just as confident that abortion should not be
criminalized. I happen to disagree with Judge Noonan about abortion. I
have simply used him as an example of a judge whose subjective level of
confidence in one of his moral judgments would seem to satisfy even a
reasonable-doubt standard.32 I conclude that even a standard as stringent
as reasonable doubt could allow judges to deviate in socially controversial
cases on the basis of their own moral convictions.
I have made no argument in favor of any particular evidentiary standard.
Different standards might be equally compatible with the guidance rules
proposed. However, the stringency of the chosen standard has a direct
impact on n: the maximum permissible deviation rate in final moderate rule.
Fewer cases satisfy more stringent standards, so judges who use a more strin-
gent standard can permissibly deviate in a greater percentage of the cases
that satisfy it. However, more cases satisfy less stringent standards, so judges
who use such standards must deviate in a smaller percentage of those cases.
Judges should not assign lexical priority to the cases in which they are
most confident that the legally required result is suboptimal. Picture a
judge who hears two cases on the same day. In both cases he is sufficiently
confident that the legally required result is suboptimal, so deviation is
permitted according to his chosen evidentiary standard. However, he is
more confident in the first case than in the second. Furthermore, in the
first case an adverse judgment will not greatly disadvantage the legally
disfavored party: the amount in controversy is small and the party is well-
off. The stakes are much higher in the second case: the law requires prison
time. Assuming that he has not exceeded his deviation limit and that he

31. Referring to the late Justice William Brennan, also Catholic, Noonan writes, “I have not
understood how a Catholic or any judge who was guided by the terms of the Constitution could
conscientiously [join the majority in Roe v. Wade]. But obviously Catholic consciences differ.
Brennan in Roe showed that they can differ on abortion. It is not, I think, the business of anyone
to judge the conscience of another.” John T. Noonan, Jr., “The Religion of the Justice: Does It
Affect Constitutional Decision Making?” Tulsa Law Review 42 (2007): 761–70, p. 763.
32. Concerning results that impose undeserved disadvantages or infringe rights, it is
often the case that the greater the magnitude of the disadvantage or infringement, the high-
er our level of confidence in our judgments of suboptimality. Forcing a suspect to take a
forensic blood test is a lesser infringement than forcing him to undergo abdominal surgery
for forensic purposes, so I am more confident that the latter is unjust. But we can also be
extremely confident that a trivial infringement, such as shoplifting a magazine without jus-
tification, is wrong.
252 Limits of Legality: Part II

must choose between these cases, he is morally permitted to deviate in

the second case rather than the first.
Judges probably have as many mistaken moral opinions as anyone else.
Judges may disagree about morality as often as anyone else. Moderate
rules do not presuppose judicial infallibility or judicial consensus. They
merely assume that judges can apply evidentiary standards to their judg-
ments of suboptimality, much as they apply these standards when they
arrive at legal judgments. Using a standard as stringent as reasonable doubt
may not be necessary in order to avoid engaging in blameworthy devia-
tion, but even a reasonable-doubt standard would seem to permit some
deviation in reasonably just legal systems. There is, of course, much more
to be said about this topic, but I must leave it there.
15

Treating Like Cases Alike

15.1 AN ILLUSTRATION

After obeying second priority rule, judges will still find that many
suboptimal-result cases are tied for deviation priority. Now we face what
may be the most counterintuitive implication of individual policy.
Individual policy does not, as a matter of principle, forbid judges to prac-
tice token-selective deviation—deviating in some, but not all, tokens of a
single case type. Such deviation invites the objection from comparative
justice in its strongest form. The time has come to consider this objection
in detail. The following scenarios illustrate token-selective deviation.
Once a month you visit the courtroom of Susan, a state trial judge. In
February, Judge Susan hears the case of Fred, a teenager who has shop-
lifted jewelry from Walmart, a large, corporate retail store. Fred has
returned the jewelry. The statute classifies shoplifting as a Class A misde-
meanor, minimum. Fred will lose his college scholarship if he is convicted
of a Class A misdemeanor, whereas he can keep it if he is convicted of a
lesser Class B. The Class A conviction will ruin Fred’s promising future.
Let us also stipulate that the Class A conviction will not benefit anyone,
including Walmart (more than negligibly).1
You return to the courtroom in March and watch Susan hear the case
of George, another teenager who has shoplifted jewelry from Walmart.
There are no morally or legally relevant distinctions between the cases:
George has returned the jewelry, the Class A conviction will cost him his
scholarship, et cetera.
Assuming all these facts, how should Susan decide the two cases? First,
she has a pro tanto moral reason not to convict either Fred or George of a
Class A misdemeanor. Many would argue that the law gives her a stronger
reason to convict both defendants of Class A misdemeanors. I rejected
some of these arguments in part I. However, I have also defended
final moderate rule, which entails that Susan could have an all-
things-considered reason not to deviate in both Fred’s and George’s cases.

1. The judge and I are ignoring the likelihood that losing his scholarship might teach
Fred a “valuable lesson” and thereby benefit him in some broader sense.

253
254 Limits of Legality: Part II

Final moderate rule forbids judges to deviate in more than n percent

of cases. It is heuristically useful to think of each judge as being allo-
cated a corresponding number of “deviation credits.” I put the term in
scare quotes because these credits are fictional—no one literally allo-
cates them to judges in actual legal systems. Nevertheless, judges
could consciously or subconsciously regulate their own decisional
conduct as if they had a finite supply of deviation credits to “spend” in
a given period of time. When Susan deviates in Fred’s case, she spends
some of her finite allocation. At some point she has no credits left.
Whether she rations her credits over the course of her career or spends
them all at the beginning, she will hear many suboptimal-result cases
in which final moderate rule forbids her to deviate. If she spends her
credits early, then these cases will all arise in the latter portion of her
career. If she rations her credits, then these cases will arise at the end
of each rationing period, however that period is defined. Therefore,
it is possible that Susan will deviate in a case, at some point, only to
be assigned thereafter a token of the same type with her credits
now exhausted. In such situations, final moderate rule forbids Susan to
deviate.
To your surprise, Susan convicts Fred of a Class B misdemeanor, but
proceeds to convict George of a Class A. You privately question her. Did
she forget Fred’s case? Was she privy to distinguishing facts of which you
were unaware? Susan assures you that there were no distinguishing facts
and that she remembers Fred’s case. In fact, she explains, in a sense she
was hard on George because she was easy on Fred. She would have liked
to have given George the Class B as well: the Class A conviction benefited
no one and greatly harmed George. But she “spent her last available devi-
ation credit” on Fred. It would be wrong for her to deviate in George’s
case, having deviated in Fred’s.
If my arguments so far are sound, and Susan has only one deviation
credit left when she hears Fred’s case, then she must not deviate in both
Fred’s and George’s cases. Her only remaining choices are adhering in
both or deviating selectively. Could she have all-things-considered reasons
to deviate in one case but not the other?

15.2 COMPARATIVE JUSTICE

Comparative justice now takes the stage, supporting the claim that Susan
has all-things-considered reasons to adhere in both cases. Goldman thinks
comparative justice provides a decisive objection to token-selective devi-
ation. He advocates what he calls the Kantian constraint: “We must not
judge two cases differently without being able to cite a relevant differ-
ence between them, where relevance is relative to domain of discourse
Treating Like Cases Alike 255

(e.g., moral or legal).”2 He also argues that judges “ought to distinguish cases
only on legally recognized grounds.”3 He applies the Kantian constraint to
courts as follows, referencing his hypothetical of the impoverished widow:

If some impoverished people were allowed to keep their homes while oth-
ers were evicted, these egregious violations of the fundamental principle
not to treat like cases differently without morally relevant differences
between them would be as damaging to the legal system as would crossing
the original threshold. The optimal pattern from a purely consequentialist
viewpoint would soon be upset as citizens reacted to these considerations
of (comparative) fairness.4

Goldman’s argument appears to be as follows:

1. Judges have a pro tanto reason not to follow a policy of deviating in

every token of case type, t.
2. Judges have a pro tanto reason to treat tokens of the same case type
alike.
3. Therefore, judges have a pro tanto reason to adhere in every token
of t.

This argument is sound. The first premise follows from a systemic-

effects argument, such as the one described in chapter 11. But the
argument supports only a pro tanto reason to adhere. In order to support
an all-things-considered reason to adhere consistently, Goldman must
argue that the reason mentioned in the second premise is always strong
enough to outweigh any reasons to deviate. In order to evaluate this
claim, we must examine the second premise more closely. The rest of
this section seeks a conception of comparative justice that makes this
premise plausible.
The comparative justice principle goes by many names: justice, fairness,
equality, formal justice, formal equality, et cetera. It is often stated in Aristo-
telian terms, as requiring adjudicators to “treat like cases alike.”5 It receives
great attention and deference, under many descriptions, in Anglo-American

2. Alan H. Goldman, Practical Rules: When We Need Them and When We Don’t
(Cambridge: Cambridge University Press, 2002), p. 2, also pp. 3, 9, 22, 41, 79, 138–40,
143–44, 159, 161, 167.
3. Ibid., p. 139.
4. Ibid., p. 45.
5. See, e.g., Michael Bayles, Procedural Justice: Allocating to Individuals (Dordrecht: Kluwer
Academic, 1989); Neil MacCormick, Legal Reasoning and Legal Theory (Oxford: Clarendon
Press, 1978), ch. 4; Joel Feinberg, “Noncomparative Justice,” Philosophical Review 83 (1974):
297–338; James I. MacAdam, “The Precepts of Justice,” Mind 77 (1968): 360–71; S. I. Benn
and R. S. Peters, The Principles of Political Thought (New York: Free Press, 1965), pp. 127–28;
William K. Frankena, “The Concept of Social Justice,” in Social Justice, ed. Richard Brandt
(Englewood Cliffs, N.J.: Prentice-Hall, 1962), pp. 9–13.
256 Limits of Legality: Part II

jurisprudence.6 To say that one case is “like” another is to say that the two
cases are tokens of one case type. Only a miniscule fraction of the features
of a case are features upon which judges should base their decisions. Most
are legally and morally irrelevant: the day of the week on which the case is
filed, the number of letters in the names of the parties, the mass of the
defendant’s kidneys, et cetera. The fact that Fred’s case was heard on a
Tuesday is a property of Fred’s case, as is the fact that Fred’s name begins
with a consonant and the fact that he enjoys opera. But none of these is a
case-type-specific feature of Fred’s case. None determines the legal or moral
type to which his case belongs. If such factors determined case types, then
every case type would be a singleton. Every case token differs from all
others in some respect. No two cases would be “alike” so it would be impos-
sible to infringe comparative justice.
If we define case types so inclusively, then the suggestion that a judge
has a good reason to deviate in one token of a type but not another becomes
logically incoherent. If the judge really has such a reason, then the tokens
must differ in some respect. And they always do. Two case tokens cannot
be identical in both intrinsic and extrinsic attributes. A case filed on Tuesday
could, in principle, be intrinsically identical to a case filed on Wednesday,
but the cases are extrinsically distinct, nevertheless: one was filed on
Tuesday, the other on Wednesday. The day of the week on which a case is
filed is an extrinsic attribute. It is generally irrelevant to the disposition of
the case (unless, of course, the law makes Tuesday a filing deadline or some-
thing). On pain of trivializing the concept of a case type, we must
understand each case type as supervening exclusively on what lawyers call
the fact pattern of the case. I call the elements of the fact pattern the intrinsic
features of the case type or its case-type-specific features.
The legal type to which a case belongs supervenes on intrinsic features
of the case. Many of these features are mentioned in legal standards. Legal
standards assign legal consequences to specified combinations of actions,
mental states, and circumstances. These standards are always relevant to

6. See, e.g., Timothy A. O. Endicott, “The Impossibility of the Rule of Law,” Oxford
Journal of Legal Studies 19 (1999): 1–18, p. 3 (“Government is arbitrary if it does not treat
like cases alike—if it does not treat people consistently”); H. L. A. Hart, The Concept of Law,
2nd ed. (Oxford: Oxford University Press, 1994), p. 157 (noting the “peculiarly intimate
connection” between justice and law); David Lyons, “On Formal Justice,” in Moral Aspects
of Legal Theory (Cambridge: Cambridge University Press, 1993); Peter Westen, Speaking
of Equality: An Analysis of the Rhetorical Force of “Equality” in Moral and Legal Discourse
(Princeton, N.J.: Princeton University Press, 1990); William Van Alstyne, “Notes on a Bicen-
tennial Constitution: Part II, Antinomial Choices and the Role of the Supreme Court,” Iowa
Law Review 72 (1987): 1281–99 (advocating a consistency criterion of judicial good faith);
Ronald Dworkin, Law’s Empire (Cambridge, Mass.: Harvard University Press, 1986),
p. 183; Peter Ingram, “Maintaining the Rule of Law,” Philosophical Quarterly 35 (1985):
359–81, p. 361 (“To follow a rule, or more universally, to be committed to a system of rules
as operative for guiding behavior, is to accept the principle that like cases should be treated
alike”).
Treating Like Cases Alike 257

judicial decisions.7 The central version of comparative justice in legal

discourse is legal comparative justice.8 Legal comparative justice requires
judges to give identical treatment to each token of a given legal case type.
Fred’s and George’s cases are tokens of a single legal case type. If Susan
deviates in Fred’s case but not in George’s, then she infringes legal
comparative justice.
Most legal writers assume that legal comparative justice has some
moral weight. There are several reasons for this. The first is that someone
may suffer emotional distress if he believes that a court has treated him
less favorably than someone in a legally indistinguishable situation. He
may feel resentment, frustration, envy, et cetera. He may lose respect for,
or question the legitimacy of, the legal system, the judiciary, the state, or
his entire society. At the least, the value of preventing these feelings and
their consequences gives courts a pro tanto reason to create the impression
that they follow legal comparative justice. Actually following legal
comparative justice is one way to create this impression. If courts have a
pro tanto duty of candor,9 then they have a pro tanto reason not to create
the false impression of following legal comparative justice—hence a
reason to actually follow it.
These reasons to follow legal comparative justice might be weak if the
aforementioned feelings usually resulted from irrationality or misinfor-
mation, but they do not. Violations of legal comparative justice correlate,
albeit imperfectly, with other wrongs and injustices. A judge violates legal
comparative justice when she adheres to the law in one case but deviates
in another with indistinguishable facts presented. If these are optimal-
result cases, then her deviation in the second case is morally unjustified.
Because many cases are optimal-result cases, decisions that violate legal
comparative justice are often morally unjustified. However, the second
decision is independently pro tanto unjustified, legal comparative justice
notwithstanding, because of this conjunction: it deviates from the law and

7. But a legal case type is not limited to facts mentioned in preexisting legal standards.
Judges often decide on the basis of other facts: sociological, economic, historical, et cetera.
When judges create new legal standards or revise old ones, especially in common law cases,
they often advert to new social conditions concerning the economy, technology, social mores,
demographics, crime, et cetera. We can regard the social conditions in which a case arises as
part of its legal case type. It is consistent for a judge in an industrial society to announce that
he is modifying a doctrine because it no longer makes sense in a post-agrarian economy.
Economic facts about his society are part of the legal case type to which the cases he hears
belong. The judge would consider revising any similarly obsolete doctrine. Social conditions
can also be relevant when a judge decides to deviate from the law without revising the
applicable legal standards. See, e.g., Morton J. Horwitz, The Transformation of American
Law: 1780–1860 (Cambridge, Mass.: Harvard University Press, 1979).
8. See, e.g., Andrei Marmor, “Should Like Cases Be Treated Alike?” Legal Theory 11
(2005): 27–38; Ken Winston, “On Treating Like Cases Alike,” California Law Review 62
(1974): 1–39; Lyons, “On Formal Justice.”
9. See §16.1 on judicial candor.
258 Limits of Legality: Part II

it generates a suboptimal result. It would be pro tanto unjustified even if

the judge had also deviated in the first case. The injustice of deviating in
the second case might be overdetermined, but we have yet to find a reason
to believe that violations of legal comparative justice are intrinsically
wrong.
There are other arguments for legal comparative justice. Judges who
follow legal comparative justice are more likely to apply general rules
when this is possible and general rules have some virtues. Also, judges who
anticipate that violations of legal comparative justice will receive extra
scrutiny will be less likely to allow bias, partiality, or prejudice to compro-
mise their decisions when that would lead them to violate legal compar-
ative justice. But these, too, are instrumental defenses of legal comparative
justice. I confess that I do not believe that violations of legal comparative
justice are intrinsically wrong, although they are extrinsically wrong, pro
tanto, in realistic legal systems. Despite this belief of mine, my next
arguments are compatible with the proposition that violations of legal
comparative justice are intrinsically wrong.
Recall the second premise of the argument attributed to Goldman:
that judges have a pro tanto reason to treat tokens of the same case type
alike. I shall now explain why this premise is implausible if we use legal
comparative justice as our conception of comparative justice. Consider
two tokens of the same legal case type that differ from one another in
morally relevant ways. If the law grants the judge who hears both cases
discretion to take these differences into account, then she is morally
permitted to treat the tokens differently, rather than reaching a suboptimal
result. Judges are morally permitted to infringe legal comparative justice
in order to avoid sufficiently suboptimal results. Alternatively, we might
say that two tokens do not actually belong to the same legal case type if
the law grants the judge discretion to treat them differently.
So Goldman’s objection to token-selective deviation cannot simply
appeal to legal comparative justice. It must appeal to another version
of comparative justice. One such version is conditional legal comparative
justice, which states that judges have a pro tanto moral reason to treat two
cases alike if the law requires it. The law requires Judge Susan to treat
Fred’s and George’s cases alike so conditional legal comparative justice
entails that she has a pro tanto reason to do so.
However, in suboptimal-result cases judges also have a pro tanto moral
reason to deviate. That reason will often outweigh the pro tanto reason
specified by conditional legal comparative justice. Conditional legal com-
parative justice does not preclude token-selective deviation in moderate
suboptimal-result cases unless deviating from the law is worse, morally
speaking, than reaching all but the most suboptimal results. I disputed
this claim in part I.
Therefore, Goldman’s second premise needs yet a different version of
comparative justice. Consider moral comparative justice: judges have a
pro tanto moral duty to treat alike cases that share all morally relevant
Treating Like Cases Alike 259

case-type-specific features. Moral comparative justice is somewhat

plausible, but the reasons it provides must be extremely weak. The fact
that the law requires a certain result in a case is not itself a case-
type-specific feature of the case. Courts infringe moral comparative
justice whenever laws change. Sue parks on the street at 5 P.M. on July
10 and again on July 17. She is ticketed on both occasions. On July 10,
however, parking at 5 P.M. was legal. The parking officer mistakenly
believed that the city council had banned street parking during rush
hour. The ban did not actually take effect until July 11. Sue contests
both tickets. The court nullifies the July 10 ticket but upholds the July
17 ticket. The morally relevant intrinsic features of the two cases are
identical. Perhaps the court has a pro tanto moral reason to treat these
cases alike, but it must be a very weak reason.
Any change in the law can make the same point. When Oregon passed
the Death with Dignity Act in 1997 certain acts of assisted suicide that
had been criminal became legal.10 Moral comparative justice entails that
Oregon courts have a pro tanto moral reason to continue convicting defen-
dants for actions that are no longer criminal. Perhaps they do, but it must
be a very weak reason. This is significant because the argument I attributed
to Goldman does not support the conclusion that judges have an
all-things-considered reason to adhere consistently unless moral compara-
tive justice always provides reasons that are strong enough to outweigh
their reasons to deviate. If this is ever so it is only in the most trivial
of suboptimal-result cases.
That leaves us with the narrower comparative justice principle of
formal equality: judges have a pro tanto moral duty to treat alike cases that
share all morally relevant features, local or nonlocal. Formal equality pro-
vides a very strong reason when it applies, but it applies only to certain
kinds of cases, such as those involving judicial discretion. Judge Jack hears
the case of State v. Adams. The law gives him discretion to impose upon
Adams a prison sentence between one and twelve months. Let us suppose
that no penal purposes, such as retribution, uniquely determine a sentence
within this range. As are many important concepts, desert is vague.
Although we can say that Adams deserves more than one week in prison
and less than three years, there is no fact of the matter as to whether he
deserves four, five, or six months. Jack arbitrarily chooses a five-month
sentence. We can agree that this is an optimal sentence based on what
Adams deserves (and all other penal purposes), but it is no more optimal
than a sentence of four months or six. Jack needs no reason to choose five
months over four or six.
Jack then hears the case of State v. Bonn, which shares all morally
relevant features with Adams. Given Adams’s sentence, formal equality

10. Or. Rev. Stat. Ann. §§ 127.800–127.897 (2006).

260 Limits of Legality: Part II

gives Jack a pro tanto reason also to sentence Bonn to five months in prison.
Whereas Jack needed no reason to give Adams six months, formal equality
entails that he needs a reason to give Bonn six months. Formal equality
requires judges to treat like cases alike when the first decision was one of
a number of equally permissible decisions. Formal equality requires a
judge to make the same decision in the second case, unless he has a good
reason to do otherwise. It forbids deciding differently for arbitrary reasons
or bad reasons or morally irrelevant reasons. That is the function of formal
equality.11
The story continues. Upon subsequent reflection, Jack concludes that
he was mistaken, that in fact Adams and Bonn really deserved six-month
sentences. He then hears the case of State v. Carlo, which shares all morally
relevant features with Adams and Bonn. A fortiori Carlo deserves a six-
month sentence and Jack has a pro tanto reason to impose it. But formal
equality entails that Jack also has a pro tanto reason to give Carlo the same
five months that Adams and Bonn received. This is a suboptimal result ex
hypothesi, but it is only slightly suboptimal. There is only one month of
difference. So Jack’s reason to give Carlo six months is weak. It could be
weaker than Jack’s formal-equality reason to give him five months,
although I need not explore this.
Adams, Bonn, and Carlo differ in many ways. The defendants have
different names and birthdays. They committed their crimes on different
days at different locations. But none of these cases has a morally relevant
feature, intrinsic or extrinsic, that distinguishes it from the other two.
Having given Adams and Bonn five months each, Jack has a reason to give
Carlo five months. But Jack also has a reason to give Carlo six months.
Without knowing more, we cannot say whether either sentence would be
wrong, all things considered. Either might be morally permitted. It would,
however, be wrong for Jack to treat the fact that he dislikes Carlo’s haircut,
or that Carlo’s ex-wife has bribed him, as reasons to give Carlo six months,
rather than five. These are morally irrelevant factors.
Formal equality applies to Adams, Bonn, and Carlo because they are
optimal-result cases in which the judge has some discretion. By contrast,
Fred’s and George’s are suboptimal-result cases. The law gives Judge Susan
no discretion to reach optimal results in these cases. Two suboptimal-result
cases can have different, morally relevant, nonlocal features even if they
share all local features. Fred’s and George’s cases have different, morally
relevant, nonlocal features. Specifically, Fred’s case comes before Judge
Susan when she has deviation credits left, whereas George’s case arises
when she has no credits left. So Susan has a good, nonarbitrary reason to

11. Judges are, in fact, disciplined for “abuse of discretion” when they act within their
discretion, but do so on an arbitrary or inappropriate basis. See, e.g., In re Brown, 662 N.W.2d
773 (Mich. 2003) (awarding temporary custody of children during Christmas to father on
basis of coin flip).
Treating Like Cases Alike 261

adhere in George’s case. Deviation credits are a scarce resource. Susan

must spend them if she is to advance the impartially defensible enterprise
of minimizing suboptimal results. As she spends them, she must allocate
them fairly. But these credits are finitely divisible. She faces a binary choice
between Class A and Class B misdemeanors. She cannot give Fred and
George each “half of a Class A and half of a Class B,” Solomonic as this
might sound. Under these conditions, her choice to give Fred a Class B and
George a Class A is arbitrary but not unfair. It is not as though Susan
decides to disfavor George because she dislikes his red hair, or because his
ex-girlfriend bribes her. Choosing on those bases would be arbitrary and
unfair. Biased decisions do not further the enterprise of minimizing subop-
timal results. But deviating in Fred’s case but not George’s for the reasons
I have specified is not biased. Nor does it infringe formal equality.
I concede that there are versions of comparative justice that Susan
infringes if she deviates in Fred’s case but not in George’s: legal comparative
justice (conditional or unconditional) and moral comparative justice. But I
have argued that these versions provide extremely weak reasons to adhere.
There is also a version that provides very strong reasons to adhere: formal
equality. However, Susan does not infringe formal equality if she deviates in
Fred’s case, but not George’s, in order to conserve deviation credits.
No argument can change the fact that token-selective deviation is an
unfamiliar and probably unsettling idea. I must emphasize that this section
serves primarily philosophical purposes. True, individual policy does not
forbid token-selective deviation as a matter of principle. But I do not
believe that we have any reason to oppose it as a matter of principle. Nor
can I easily predict how often real judges will have an all-things-considered
reason to engage in token-selective deviation according to individual
policy. But token-selective deviation often goes unnoticed. I have deliber-
ately concocted stark and artificial juxtapositions of identical cases for the
purpose of explaining what token-selective decisions are. If individual
policy entailed frequent episodes of token selectivity as conspicuous as
those I have described, then that might give cause for alarm. In the real
world, however, the type identity of cases is often camouflaged behind
distinguishing features that seem, at least to some people, to have practical
relevance.12 Much token selectivity takes place “under the radar.”
I mentioned earlier that first priority rule permits judges to practice
type-selective deviation—deviating in some but not all suboptimal-result
cases at a certain level of suboptimality.13 I also mentioned that there is a
comparative justice argument against this practice but I declined to present
it. Now you can see why. If two suboptimal-result cases are tokens of the
same type, then the legally required results are equally suboptimal in both.

12. See the discussion of legal mirages in §16.9.

13. See §14.5.
262 Limits of Legality: Part II

All instances of token-selective deviation are also instances of type-

selective deviation. If an argument condemns type-selective deviation,
then it condemns token-selective deviation. I have attempted to refute
comparative justice objections to token-selective deviation. My arguments,
if successful, refute parallel objections to type-selective deviation.

15.3 INTEGRITY

As I have noted, judges following individual policy in real legal systems

rarely have reason to make a conspicuous display of token-selective
deviation. But I am willing to bite this particular bullet and admit that
individual policy presents no principled objection to token-selective devi-
ation, and that engaging in it inconspicuously can be justified under certain
conditions. In this respect, my position appears to conflict with integrity—
the central value of Dworkin’s jurisprudence. Dworkin opposes deviation,
selective or not.14 But integrity seems to provide a further reason to
oppose token-selective deviation, in particular, with special vehemence.
To what extent can my position reconcile with Dworkin’s?
Dworkin argues that integrity, in his special sense, is a basic value that
everyone accepts.15 He illustrates this by having us imagine legislatures
that respond to electoral disagreement about moral issues by enacting
legislative schemes in which “each body of opinion is represented, to a
degree that matches its numbers, in the final result.”16 He gives three
examples of such checkerboard laws:17

Do the people of North Dakota disagree whether justice requires compensation

for product defects that manufacturers could not reasonably have prevented?
Then why should their legislature not impose this “strict” liability on manufac-
turers of automobiles but not on manufacturers of washing machines? Do the
people of Alabama disagree about the morality of racial discrimination? Why
should their legislature not forbid racial discrimination on buses but permit it
in restaurants? Do the British divide on the morality of abortion? Why should
Parliament not make abortion criminal for pregnant women who were born in
even years but not for those born in odd years?18

Dworkin argues convincingly that checkerboards offend neither justice,

nor fairness, nor expedience. But he insists that everyone nonetheless con-
demns checkerboard solutions. The best explanation for our condemnation,
he infers, is that integrity—a value distinct from justice, fairness, and

14. Except, perhaps, to avoid extreme injustices. See §1.2.

15. Dworkin, Law’s Empire, pp. 182–83.
16. Ibid., p. 178.
17. Ibid., p. 179.
18. Ibid., p. 178.
Treating Like Cases Alike 263

expedience—is basic to our political morality.19 The checkerboard state

lacks integrity because it is unprincipled: “it must endorse principles to
justify part of what it has done that it must reject to justify the rest.”20
Now imagine a checkerboard statute requiring judges to convict shop-
lifters of Class A misdemeanors in odd months but allowing them to
convict shoplifters of Class B misdemeanors in even months. Dworkin
would surely oppose such statutes so he would probably oppose how Judge
Susan decides Fred’s and George’s cases (beyond the fact that she deviates).
However, I shall argue that Dworkin provides no reason to embrace a
conception of integrity that actually condemns Susan’s selectivity.
Dworkin would point out that because Fred’s and George’s cases are
tokens of the same type, no case-type-specific principle could justify
deciding them differently. In this respect, Susan’s policy resembles
legislation that criminalizes abortion only for women born in even years.
In the abortion case, however, Parliament also has the option of enacting
a better law (whatever that is). By contrast, Judge Susan’s alternatives to
selectivity are worse. If she adheres in both cases, then she inflicts more
undeserved disadvantages. If she deviates in both, then she rides free on
her fellow judges or participates in an excessive-deviation enterprise.
Susan simply lacks the ability to prevent mimetic failure. She cannot
control the decisions of future judges who hold erroneous moral opinions.
So it is unreasonable to expect her to eschew token selectivity but
reasonable to demand that Parliament eschew checkerboard legislation.21
Interestingly, Dworkin even permits some forms of token selectivity.
He admits that “arbitrarily” choosing to rescue some prisoners of tyranny
from torture, when one cannot rescue them all, makes good sense and is
morally permissible.22 Susan, similarly, chooses to “rescue” one of the two
shoplifters from an unjustified fate when she cannot rescue both.
Dworkin also supports token selectivity in the guise of prosecutorial
discretion: “If a prosecutor’s reason for not prosecuting one person lies in
policy—if the prosecution would be too expensive, for example, or would
for some reason not contribute effectively to deterrence—integrity offers
no reason why someone else should not be prosecuted when these reasons
of policy are absent or reversed.”23 Susan’s reason for deviating selectively
is, likewise, a reason of policy, rather than principle. If principle alone
mattered, then Susan would deviate in both cases. But deviating in too

19. Ibid., pp. 179–83.

20. Ibid., p. 184.
21. On a related point see Frank Easterbrook, “Ways of Criticizing the Court,” Harvard
Law Review 95 (1982): 802–32 (concluding that we may reasonably expect each justice of
the Supreme Court to develop a consistent, principled jurisprudence, but we must not ask
the Court as whole to do so).
22. Dworkin, Law’s Empire, pp. 181, 184.
23. Ibid., p. 224.
264 Limits of Legality: Part II

many cases will cause too much mimetic failure. Dworkin gives us no
reason to reject Susan’s reasoning.
Dworkin mentions, however, that some kinds of selectivity cannot be
justified by reasons of policy: “Obviously integrity would also condemn
prosecutors’ decisions that discriminate, even for reasons of ostensible
policy, on grounds that violate rights otherwise recognized, as if our
prosecutors saved expense by prosecuting only blacks for a kind of crime
that was particularly prevalent in mainly black communities.”24 Similarly,
he asserts that “[a]n American legislature could not decide that no Catholic
farmer should receive subsidies even if, incredibly, there were sound reasons
of policy for this discrimination.”25 Dworkin does not tell us why he singles
out racial and religious categories in these passages, but it matters not. Pre-
sumably he objects to laws that relegate a “discrete and insular minority” to
second-class status. But Susan’s reason for favoring Fred over George is not
racial, or religious, or anything remotely similar. I can agree that she should
not decide on such grounds. In the next section I shall argue that deciding
randomly in such situations may be appropriate, in theory.

15.4 RANDOMIZING

The previous sections illustrate that formal equality permits Judge Susan,
in principle, to engage in token-selective deviation. But individual policy
provides her no reason to decide among the suboptimal-result cases that
second priority rule classifies as equally eligible for deviation. How should
she decide? One possible method involves a simple randomizing device.
This is what game theorists call a mixed strategy.26
Randomizing becomes appropriate only when all nonarbitrary selec-
tion criteria are exhausted. As mentioned, judges should give priority to
the more suboptimal of the suboptimal-result cases and to suboptimal-
rule cases over gap cases. Whenever possible, Susan should group together
gap-case tokens of a single type and deviate in all of them, rather than
deviating in some tokens of a type but not others. But this is rarely possible.
Usually Susan will find herself choosing between (1) leaving some of her
deviation allotment unused and (2) spending her remaining allotment on
a set of case tokens that includes some but not all of the tokens of the
type. After these priorities have been applied an argument can be made
for a randomizing strategy. From the set of all case tokens that are equally
eligible for deviation, Susan could randomly choose a subset to fill out her
deviation allotment. She could opt for the following:

24. Ibid.
25. Ibid., p. 223.
26. See, e.g., R. Duncan Luce and Howard Raiffa, Games and Decisions (New York:
Dover, 1989), p. 70.
Treating Like Cases Alike 265

Default Rule: If final moderate rule permits deviation in only one of two
suboptimal-result case-tokens, which cannot be distinguished under first and
second priority rules, then choose between the cases on the basis of any mor-
ally permissible criteria, including random choice.

Although the optimizing power of mixed strategies has been under-

stood for decades, few philosophers have entertained the possibility of
using them as solutions to moral-moral prisoner’s dilemmas. Goldman
comes close. He briefly entertains the idea of using “some randomizing
procedure” to optimize, but he promptly rejects the idea as facing
“insuperable” problems.27 He “take[s] it as obvious that designing and
implementing such a procedure would be neither feasible nor acceptable
to the majority of citizens.” In a subsequent paper he states that
randomizing would not be feasible because “it would be too difficult to
check whether the individuals allowed not to comply with the rule were
donating their money to moral projects.”28
Consider Goldman’s claim that it would be too difficult to monitor
individual compliance. Regarding taxpayers, he may be correct. Taxpayers
have powerful incentives to reroute tax funds to their own selfish pur-
poses rather than moral purposes. Monitoring them would be imperative,
but there are millions of taxpayers. Tracking their activities is costly and
infringes on their privacy. Perhaps this is what Goldman means when he
writes that a randomizing procedure would not be “acceptable to the
majority of citizens.” Monitoring would be too invasive and costly, but
without it citizens would not trust that their peers were using the exemp-
tion for moral purposes. For these reasons taxpayers would also have
difficulty coordinating their behavior, as Goldman suggests. A taxpayer
has little information about what his fellow taxpayers are doing, and he
knows that others have little information about him.
Judges, by contrast, are relatively few and easy to monitor. They can
obtain information about one another, as can others. Although their delib-
erations are private, their rulings and opinions are matters of public record
in which they have no privacy interest whatsoever. Moreover, a judge
cannot benefit himself in any direct way by deviating from the law if checks
on judicial corruption and self-dealing are vigorously enforced. If anything,
he has reputational incentives to adhere even when he believes that devi-
ation is morally warranted. Anyone who wants to know how often a certain
judge deviates from the law can examine his judicial record. We must also
remember Posner’s claim that judges enjoy adhering to the law.29
Goldman also states that “considerations of fairness tell against allowing
some individuals, but not others, to make their own decisions in this

27. Goldman, Practical Rules, p. 44.

28. Alan H. Goldman, “The Rationality of Complying with Rules: Paradox Resolved,”
Ethics 116 (2006): 453–70,” p. 456.
29. See §4.6.
266 Limits of Legality: Part II

regard.” He concludes that all citizens should simply pay their taxes rather
than rerouting to charity. He takes a parallel position regarding judicial
deviation: all judges should simply adhere to the law.30 But Goldman’s
envisioned randomizing procedure has an unnecessary feature that dimin-
ishes its appeal. His procedure arbitrarily divides the population into two
castes: those who may deviate and those who may not. There are, of
course, unfair ways of distributing scarce privileges. Consider a system
that allowed only Protestants or only right-handed citizens to redirect
their tax payments to charities. Creating this hierarchy may be wrong in
itself. It is likely to foster resentment between the castes and suspicion
that the sorting was based on morally irrelevant attributes even if it was
not. Moreover, if the activity of defecting exhibits diminishing marginal
utility for defectors, then Goldman’s way of allocating the privilege is still
unfair as well as Pareto-inferior. Rather than allowing some to reroute all
while others reroute none, it would be fairer, and Pareto-superior, to allow
everyone (or the largest possible randomly selected subpopulation) to
reroute some fraction of his tax payment to charity. That is the proper
analog to a randomizing procedure that permits each judge to deviate in
some percentage of his suboptimal-result cases.
Nevertheless, the mere fact that default rule authorizes judges to
decide randomly under certain conditions will alarm some readers, so I
must explain what it does and does not mean. Judges are blameworthy for
making decisions randomly if they disregard relevant legal or moral con-
siderations. Decision by coin flip “obviously constitutes a complete
abdication of the judicial function, which is . . . the duty to make reasoned
decisions according to law.”31 Judges have been censured for deciding
cases by flipping coins,32 for arbitrarily dismissing criminal charges,33 and
for deciding based on the odor of a defendant’s hair.34 These judges acted
unethically, and default rule does not come to their defense. Default rule
should be understood in light of other special contexts in which we allow
random factors to influence legal outcomes. Consider some familiar
examples. The complaints filed with a court are assigned randomly to
sitting judges. Different judges apply the law differently and exercise dis-
cretion differently, so these random assignments affect outcomes. The
same is true, of course, of jury selection, another random process. The
military draft in the United States operates by lottery. From a set of income

30. Goldman, Practical Rules, pp. 147–48.

31. Steven Lubet, “Judicial Discipline and Judicial Independence,” Law and Contempo-
rary Problems 61 (1998): 59–74, p. 74.
32. In re Daniels, 340 So.2d 301 (La. 1976); In re Brown. Another judge was sanctioned
for giving the appearance of having decided on the basis of a coin flip. Turco, Stipulation
(Washington Commission on Judicial Conduct, Oct. 2, 1992).
33. In re DeRose, 1980 Annual Report 181 (New York State Commission on Judicial
Conduct, Nov. 13, 1979).
34. Aaron (California Commission on Judicial Performance, July 8, 2002).
Treating Like Cases Alike 267

tax returns with equally significant indicators, the Internal Revenue

Service randomly chooses some for tax audits. Highway patrol officers
perform random traffic stops. Indeed, whether someone is arrested and
prosecuted for any crime depends substantially on random factors such as
the proximity of law enforcement, the availability of witnesses, and the
caliber of detective work.
Scholars, too, have defended random decision in various settings,
including some legal ones.35 Deciding randomly between two or more
options is not unjust absent a moral basis for choice. Default rule authorizes
random decision only under those conditions. It permits judges to appeal
to random factors for the purpose of deciding in which of two subopti-
mal-result case tokens to deviate when final moderate rule permits
deviation in only one of them and they cannot be distinguished under first
and second priority rules.
I can now state the entire scheme of guidance rules that I call selective
optimization:

Final Moderate Rule: Over the course of your judicial career, deviate in no more
than n percent of suboptimal-result cases.

First Priority Rule: If final moderate rule permits you to deviate in only one of
two suboptimal-result cases, then deviate in the one in which the legally
required result is more suboptimal.

Second Priority Rule: If final moderate rule and first priority rule permit devia-
tion in only one of two sets, each composed of the same number of cases,
where the members of the first set fall under more than one rule and the mem-
bers of the second set fall under a single rule, and are otherwise as apt for
deviation as are the members of the first set (regarding suboptimality, etc.),
then deviate in the second set.

Default Rule: If final moderate rule permits deviation in only one of two
suboptimal-result case tokens, which cannot be distinguished under first and
second priority rules, then choose between the cases on the basis of any
morally permissible criteria, including random choice.

Final moderate rule is lexically prior to first priority rule, which is prior to
second priority rule, which is prior to default rule.

35. The most thorough, sympathetic treatment of randomization in the law is

Neil Duxbury, Random Justice: On Lotteries and Legal Decision-Making (Oxford: Clarendon
Press, 1999). See also Barbara Goodwin, Justice by Lottery (Hemel Hempstead: Harvester
Wheatsheaf, 1992); David Lewis, “The Punishment that Leaves Something to Chance,
” Philosophy and Public Affairs 18 (1989): 53–67; John Harris, Violence and Responsibility
(London: Routledge & Kegan Paul, 1980), pp. 66–84.
268 Limits of Legality: Part II

15.5 BLAMEWORTHY ADHERENCE

I have suggested using evidentiary standards similar to those used in law

in order to determine whether a judge has sufficient evidence of subopti-
mality to warrant blameless deviation in a given case, pursuant to selective
optimization.36 Observers could use these standards to evaluate judges.
Judges could use them to evaluate themselves. But what about a judge
who adheres more often than selective optimization requires, obeying
restrictive rule or a moderate rule more restrictive than selective optimi-
zation? Is he blameworthy for reaching suboptimal results that selective
optimization permits him to avoid?
Wrongdoers can sometimes avoid blame by demonstrating that they
lack the requisite mental state. However, this moral defense is unavailable
to judges. A judge knows that his decision threatens the losing party with
undeserved force.37 If he adheres in what he believes to be an optimal-
result case, then he believes that the threat is morally justified and that it
would be justified even if the law did not require it. If his beliefs are rea-
sonable, then he is blameless. However, if he adheres in what he believes
to be a suboptimal-result case, then he believes that the threat would not
be morally justified if the law did not require it. Because he knows that his
decision threatens force under these conditions, he cannot claim to be
merely reckless or negligent. His mens rea is, at least, “knowledge.”
Agents who knowingly act wrongly can sometimes avoid blame if they
act under duress. Again, this defense rarely applies to judges. If someone
credibly threatens to maim the judge if she deviates, then she can claim
duress.38 However, deviation in the United States carries no risk of criminal
or civil penalties and minimal risk of formal sanctions.39 Realistically
speaking, a judge who deviates runs some reputational risks at most. He
may lose some esteem in the eyes of professional colleagues, commenta-
tors, and the general public.40 How should we morally evaluate a judge
who adheres to the law in a suboptimal-result case in order to avoid rep-
utational risks? I suggest that his desire to avoid damaging his reputation
does not undermine his subjective moral reasons to deviate. It does not
subjectively justify adhering. One’s reputation is precious, to be sure, but
I cannot think of another situation in which protecting one’s reputation

36. See §14.8.

37. Notwithstanding the possibilities of reversal on appeal, legislative intervention,
defiance by subordinate officials, executive pardon, et cetera. See the discussion of “institu-
tions of amelioration” in Kent Greenawalt, Conflicts of Law and Morality (New York: Oxford
University Press, 1987), pp. 271–376.
38. Lyons discusses such an example in David Lyons, “Derivability, Defensibility,
and the Justification of Judicial Decisions,” in Moral Aspects of Legal Theory (Cambridge:
Cambridge University Press, 1993) p. 138.
39. See §4.3.
40. See §4.6.
Treating Like Cases Alike 269

undermines a moral reason against using force. A teenager has an all-

things-considered moral reason not to vandalize windows, even if his
friends will ostracize him for refusing. At most, we might mitigate his
blameworthiness, blaming him somewhat less than we would blame a
solitary, thrill-seeking vandal. Analogously, perhaps a judge who adheres
in a suboptimal-result case in order to protect his reputation is less blame-
worthy than a judge with no reputational concerns—one who adheres in
the case simply because he enjoys enforcing the law. But reputational
concerns cannot eliminate blame.
Nevertheless, judges who knowingly adhere in suboptimal-result cases
are frequently blameless. In reasonably just legal systems many suboptimal-
result cases are ones about which reasonable people could disagree.
Therefore a judge may be sincerely and reasonably uncertain that the instant
case is, in fact, a suboptimal-result case. He is blameless for adhering in such
a case. Many cases in modern legal systems will fall in this category.
Even if a judge is confident that a case is a suboptimal-result case, he
may be blameless for adhering. Selective optimization requires judges not
to deviate in more than n percent of their suboptimal-result cases. They
must adhere in the remainder. A judge is blameless if she reasonably
adheres in a suboptimal-result case in order to conserve deviation credits.
She adheres because she aims to contribute to the enterprise of minimizing
suboptimal results.
One can, however, imagine special situations in which conserving
credits would be unreasonable. If a judge knows that his judicial career is
almost over and that he has deviation credits left and he is sufficiently
confident that the case before him is a suboptimal-result case, then
adhering could be blameworthy even if he adheres in order to conserve
credits. Even before the end of his career a judge might become blame-
worthy if he deviates too infrequently in cases that he confidently believes
to be suboptimal-result cases. But an observer will rarely be in a position
to blame a judge for any particular adherent decision if the judge
reasonably adheres in order to conserve credits. Blame for adhering
becomes warranted only if two conditions are met. First, the judge exhibits
a long-term pattern of adhering in cases that he does not reasonably
believe to be optimal-result cases. Second, the judge does not reasonably
believe that selective optimization requires adhering as often as he does. I
do not know how often these conditions are satisfied.
16

Implementation

Does selective optimization have anything to do with the real world?

Perhaps. Judges may already be using various elements of selective opti-
mization, although I shall not test that hypothesis. Instead, I shall address
the prospects for implementing selective optimization. Lawmakers could,
in theory, codify it in legislation or codes of judicial conduct. However, no
lawmaker today would want to be associated with the claim that deviation
is sometimes permissible. Therefore, I shall concentrate on the feasibility
of selective optimization serving as a set of uncodified guidance rules:
rules not formally enacted or announced by lawmakers. Although not
backed by formal sanctions, they can still guide action and serve as the
basis for instruction, criticism, praise, and blame.1 They can function as
social norms for particular groups,2 as do rules of etiquette, or as personal
policies for particular individuals.
In this chapter I address two basic questions about implementing
selective optimization: (1) to what extent should judges disclose or conceal
the fact that they selectively optimize (§§16.1–16.6), and (2) are judges
psychologically capable of internalizing selective optimization (§§16.7–
16.9)?

16.1 CANDOR

Deviation involves either advancing fallacious legal arguments or reaching

results for which no sound legal argument can be given. When judges
deviate knowingly, not mistakenly, they almost never express in advance
their intention to deviate or admit it when they do so. Instead they make

1. Larry Alexander and Emily Sherwin, The Rule of Rules (Durham, N.C.: Duke Univer-
sity Press, 2001), p. 39 (positive rules need not be written); Frederick Schauer, Playing by the
Rules: A Philosophical Examination of Rule-Based Decision-Making in Law and in Life (Oxford:
Oxford University Press, 1991), p. 71 (rules can exist without “canonically inscribed formu-
lations”).
2. See, e.g., Eric Posner, Law and Social Norms (Cambridge, Mass.: Harvard University
Press, 2000).

270
Implementation 271

legal assertions—in filed opinions or statements from the bench—that

they do not believe to be true. They deliberately foster the impression of
adherence. Such surreptitious deviation raises ethical questions because
judges have a pro tanto moral duty to be candid about the law in their
public statements.3 Published opinions are supposed to persuade readers
that any applicable law has been applied correctly and to inform readers
of the state of the law (which may or may not have changed). If a judge
does not believe a certain legal proposition to be true, then she has a pro
tanto duty not to assert it publicly.4 In this section I argue that, neverthe-
less, judges who knowingly deviate have reasons to do so surreptitiously—
reasons that compete with and sometimes outweigh their reasons to be
candid.
A judge who plans to selectively optimize could, in theory, announce his
intention to do so or announce it whenever he deviates. However, such
announcements hurt the judge’s career in realistic societies. An elected
judge who makes such an announcement lowers his chance of reelection.
An unelected judge increases his chance of impeachment, removal, or non-
promotion. So a judge has prudential reasons not to announce his deviation.
However, prudential reasons do not typically outweigh moral reasons.
Consider, instead, the fact that a candid judge might be replaced by—or
passed over in favor of—a judge who reaches suboptimal results more often
than he would. If his announcement has that effect, then it actually increases
the number of suboptimal results reached, thereby defeating the purpose of
selective optimization. Candor also exacerbates mimetic failure, of course,
encouraging subpar judges to deviate in optimal-result cases. The chance of
any given judge’s announcement having this effect is probably small, but it
still provides a pro tanto reason against making the announcement.
For the reasons given, selective optimization comprises esoteric rules for
the members of Group O in realistic societies: the rules fail to achieve
their purpose if it is generally known that they are followed.5 Judges who
choose to selectively optimize have prudential reasons not to announce
that they plan to do so and not to announce when they actually deviate, at

3. See, e.g., Scott C. Idleman, “A Prudential Theory of Judicial Candor,” Texas Law
Review 73 (1995): 1307–417; Deborah Hellman, “The Importance of Appearing Principled,”
Arizona Law Review 37 (1995): 1108–51; David Shapiro, “In Defense of Judicial Candor,”
Harvard Law Review 100 (1995): 731–50; Phototron Corp. v. Eastman Kodak Co., 687 F.
Supp. 1061 (N.D. Tex. 1988) (“The requirement that the judiciary be candid is perhaps
absolute . . .”).
4. She probably also has a duty to volunteer reasons for her decisions and a duty to
avoid actions that could lead reasonable people to infer that she believes false propositions.
5. On esoteric theories see Ben Eggleston, “Self-Defeat, Publicity, and Incoherence:
Three Criteria for Consequentialist Theories,” Ph.D. diss., University of Pittsburgh, 2001;
Derek Parfit, Reasons and Persons (Oxford: Clarendon Press, 1984), p. 40–43; Peter Railton,
“Alienation, Consequentialism, and the Demands of Morality,” Philosophy and Public Affairs
13 (1984): 134–71; Henry Sidgwick, The Methods of Ethics, 7th ed. (Indianapolis, Ind.:
Hackett, 1981).
272 Limits of Legality: Part II

least until deviation gains broader acceptance (which may never happen).
Judges in Group O also have moral reasons not to be candid about their
deviation.6
Some readers will be suspicious of esoteric rules. Rules that could be
publicized seem preferable, ceteris paribus. Some philosophers even treat
publicity as a “formal constraint” on principles of right.7 They categorically
reject esoteric rules. However, the total package of rules that I support is
not esoteric in the sense contemplated by the publicity condition. The pub-
licity condition rules out principles that could not be publicized to everyone.
It does not require that principles be safely capable of publicity under all
nonideal conditions. No principle passes that test. The only reason that the
members of Group O should not publicly admit that they are deviating is
that many judges in realistic societies hold incorrect opinions about which
results are optimal. Under these conditions mimetic failure remains a
danger. The larger Group O becomes, the more openly can its members
optimize. At some point the danger of mimetic failure disappears and total
candor—total publicity—becomes feasible. Total candor is unjustified,
however, if it might encourage someone to deviate in an optimal-result case.
Group O must interact with many such individuals in realistic societies.
Therefore, judges have moral reasons to persuade readers that they have
adhered, whether or not they have. Imagine that Justice Douglas had writ-
ten the following opinion for the Court in Griswold v. Connecticut,8 without
offering any additional reasoning: “Because we find §53–32 of the General
Statutes of Connecticut to be an uncommonly silly law, appellants’
convictions are hereby reversed.” 9 Such an opinion might, in fact, repre-
sent accurately why the majority wished to invalidate the law. But even
lawyers who believe that Connecticut’s contraception laws were pro-
foundly unjust and unconstitutional will agree that this argument represents
fallacious legal reasoning. Not just any argument for an optimal result is a
sound legal argument. Even Panglossians know this. The justices have no
legal authority to invalidate a statute simply because they think it “uncom-
monly silly.” The hypothetical opinion is deviant. More important, it fails

6. Compare Meir Dan-Cohen’s argument that legal systems maintain “acoustic

separation” between decision rules and conduct rules in criminal law and that this is often
appropriate. Meir Dan-Cohen, “Decision Rules and Conduct Rules: On Acoustic Separation
in Criminal Law,” Harvard Law Review 97 (1984): 625–77.
7. See, e.g., Bernard Williams, Ethics and the Limits of Philosophy (Cambridge, Mass.:
Harvard University Press, 1985), pp. 101–2, 108–9; John Rawls, A Theory of Justice
(Cambridge, Mass.: Harvard University Press, 1971), p. 130. I actually reject publicity as a
formal constraint, for reasons advanced by Brink and others, but I set aside my general
objection here. See David O. Brink, Moral Realism and the Foundations of Ethics (New York:
Cambridge University Press, 1989), p. 428; Eggleston, “Self-Defeat, Publicity, and Incoher-
ence: Three Criteria for Consequentialist Theories.”
8. Introduced in §6.1.
9. Griswold v. Connecticut, 381 U.S. 479, 527 (1965) (Stewart, J., dissenting) (Con-
necticut’s law “uncommonly silly”).
Implementation 273

to disguise its deviance well.10 It runs a greater risk of causing mimetic

failure than would an opinion containing more persuasive legal reasoning
in support of the result, or an opinion at least appearing to reflect a good-
faith effort to adhere. A judge who deviates has a pro tanto moral reason to
write an opinion that causes minimal mimetic failure. Assume arguendo
that Justice Douglas actually agreed with the critics of Griswold that his
argument was fallacious. On that assumption, one can plausibly infer that
he wrote the opinion as he did—dutifully citing the Bill of Rights and
precedent—in order to disguise its deviance as best he could. This was,
apparently, the best legal argument he could devise for declaring
Connecticut’s statute unconstitutional as applied to married couples.
On these assumptions, restrictive rule entails that it was morally imper-
missible for Justice Douglas to file the Griswold opinion. However, if
selective optimization is true and surreptitious deviation is permissible, then
a judge is morally permitted to advance a fallacious legal argument in order
to avoid a suboptimal result, if he cannot find a sound legal argument for
avoiding it and his system is below threshold. Therefore, the Griswold Court
was morally permitted to advance a fallacious argument if the following
three conditions obtained in 1965: the Court could not find a sound legal
argument for reversing Griswold’s conviction; upholding her conviction was
a suboptimal result; and the U.S. legal system was below threshold.
Although surreptitious deviation attracts less attention than overt
deviation, it is still risky. Disingenuous opinions never persuade everyone
that the judge has adhered. Observers may infer that she is incompetent
or corrupt. Even if they reach a less derogatory conclusion—that she is
using consistent adherence rules that permit some deviation—they may
conclude that these rules are indefensible. Someone who reaches negative
conclusions about too many judges loses faith in the judiciary and the rule
of law. If too many people lose faith, then the system fails. Therefore,
judges have pro tanto moral reasons not to engage in such ruses,11 but in
some cases they may be justified in doing so.12
10. I am not sure whether other public officials would even defer to such opinions if the
Court were crazy enough to file them. Perhaps justices who filed such opinions would be
impeached and removed from office.
11. See Charles Fried, “Impudence,” Supreme Court Review 1992 (1992): 155–94 (dis-
tinguishing between open judicial “defiance” of the law and covert “impudence,” vehemently
condemning the latter).
12. Dworkin envisions situations in which a judge believes himself to have a moral duty
to advance fallacious legal arguments: “If the judge decides that the reasons supplied by
background moral rights are so strong that he has a moral duty to do what he can to support
these rights, then it may be that he must lie, because he cannot be of any help unless he is
understood as saying, in his official role, that the legal rights are different from what he
believes they are.” Ronald Dworkin, “A Reply to Critics,” in Taking Rights Seriously
(Cambridge, Mass.: Harvard University Press, 1978), p. 327. See also Paul Butler, “When
Judges Lie (and When They Should),” Minnesota Law Review 91 (2007): 1785–828; Martin
Shapiro, “Judges as Liars,” Harvard Journal of Law and Public Policy 17 (1994): 155–56,
p. 156 (because judges “must always deny their authority to make law, even when they are
making law. . . . [c]ourts and judges always lie”).
274 Limits of Legality: Part II

I hope it is obvious that selective optimization by itself does not entail

that invalidating Connecticut’s law was permissible. It merely entails that,
if banning contraceptives is unjust and the system is below threshold, then
courts may be morally permitted to invalidate such bans even if the
Constitution does not authorize them to do so. This particular result will
please opponents of such bans. However, selective optimization does not
entail that upholding Connecticut’s law would necessarily have been
impermissible, either. On the contrary, it entails that if allowing married
couples to use contraceptives was a suboptimal result and the system was
below threshold, then the Court was morally permitted to uphold such
bans even if constitutional law actually required invalidation. In fact,
selective optimization does not entail that a court today is morally
obligated to invalidate such bans. Imagine that a state today criminalizes
contraceptive use. A woman convicted under the new law challenges her
conviction, citing the Griswold line of cases. The trial judge, however,
believes that there are strong reasons of morality and public policy to ban
contraceptives. He affirms the conviction, writing a fallacious opinion that
purports to distinguish the new law from those overturned in Griswold
and its progeny. Nonetheless, selective optimization entails that he has
acted permissibly if he is correct that acquittal is a suboptimal result and
his system is below threshold.
The preceding conclusion may disappoint those seeking an uncondi-
tional defense of birth control rights, but they cannot have it both ways. If
one agrees that the Griswold Court deviated, but rejects that fact as a
reason to disapprove of the decision, then one cannot consistently treat
deviation itself as a reason to disapprove of other decisions. Conversely, if
the optimality of the Griswold holding justified deviation, then the
(ostensible) optimality of the contrary result could justify deviation in the
opposite direction. Of course, I am not questioning the overwhelming
moral and policy arguments against banning birth control. I merely
demand consistent criteria for permissible deviation. Because selective
optimization incorporates no one’s opinions about optimal results—
neither yours nor mine—we cannot expect it to permit and forbid devia-
tion in just the cases in which you or I would deviate or adhere, respectively.
If the general enabling conditions for deviation are met, but you would
prefer the court to adhere, then your only honest arguments for adherence
are nonlegal. You could argue that the legally required result is, in fact,
optimal—using moral or policy considerations that have not yet been in-
corporated into the law. Or you could argue that the judge is likely to hear
many other cases in which the legally required results are more suboptimal
than in this case, such that spending deviation credits now would be
foolish. According to selective optimization, the fact that the law requires
a certain result constitutes a disjunctive reason: a reason either to reach
that result or to “debit one’s deviation account.”
Consider, in light of the previous discussion, two landmark cases con-
cerning sexual orientation. In Romer v. Evans a 6–3 majority of the Supreme
Implementation 275

Court used the Equal Protection Clause to strike down a provision of the
Colorado Constitution that permanently banned legislation making sex-
ual orientation into a civil rights category.13 In Lawrence v. Texas a 5–4
majority held that a criminal prohibition on same-sex sodomy, along with
most state sodomy laws,14 deprived individuals of liberty under the Due
Process Clause of the Fourteenth Amendment.15 Justice Antonin Scalia
dissented vigorously in both cases. He believes that Romer and Lawrence
were incorrectly decided as matters of constitutional law. If I understand
him, then this is what he thinks happened. The Romer and Lawrence
majorities believed the challenged laws to be unjust and were determined
to invalidate them. They did their best to write opinions supporting inval-
idation using accepted forms of legal argument.16 This was impossible:
there was no way to support invalidation with accepted legal arguments.
So the majorities deviated, knowingly or not. Believing that judges must
always adhere, Scalia concludes that the majorities acted impermissibly,
all things considered.
The most direct response to Scalia is to defend Romer and Lawrence as
legally sound opinions. This is the standard response of liberal lawyers
who believe that criminalizing sodomy is unjust and that sexual orienta-
tion should be a proscribed classification under civil rights laws.17 But
those of us who hold these moral opinions must be especially careful
when evaluating Romer and Lawrence. Moral convictions can make
fallacious legal arguments seem sound even to good lawyers. Socially con-
servative lawyers object that the arguments in Lawrence and Romer tacitly
rest upon premises that they reject and that are not part of the law: (1)
the premise that gay men and lesbians have a moral right to greater sexual
liberty (Lawrence) or greater social equality (Romer) than they currently
enjoy, and (2) the premise that the state should vindicate these rights
under law.
In response to this objection, liberals have three options: first, deny
that the arguments actually rely upon these premises; second, argue that
these premises are, in fact, part of the law; third, offer different arguments
for the holdings. Perhaps one of these options will yet succeed, but selec-
tive optimization offers a fourth alternative. It allows one to defend Law-
rence and Romer as justifiable, all things considered, while admitting—at
least to oneself—that they were quite possibly deviant18 decisions: maybe

13. Romer v. Evans, 517 U.S. 620 (1996).

14. Whether or not they singled out same-sex sodomy.
15. Lawrence v. Texas, 539 U.S. 558, 578 (2003). This case explicitly overruled Bowers
v. Hardwick, 478 U.S. 186 (1986).
16. See, e.g., Philip Bobbitt, Constitutional Interpretation (Oxford: Basil Blackwell,
1991), pp. 12–13.
17. Or at least that state constitutions should not permanently proscribe such
legislation.
18. Le mot juste.
276 Limits of Legality: Part II

there was no sound constitutional argument for invalidating these laws. If

surreptitious deviation is permissible, then one can even defend Lawrence
and Romer on the assumption that the majorities knowingly deviated. In
order to minimize the impression of deviation, they wrote opinions in the
style of constitutional argument. These arguments were flawed, but the
justices did their best.
However, announcing that Lawrence and Romer are “legal mistakes”
could confuse people, for the following reason. Treating restrictive rule as
true is a widely accepted social norm governing public discourse about
the law. Restrictive rule states that reaching legally incorrect results is
morally impermissible in reasonably just societies, so people reasonably
assume that anyone who publicly calls Lawrence a “legal mistake” also
believes that the decision was morally impermissible, all things consid-
ered. They will severely misunderstand anyone who makes this statement
unless they realize that she also rejects restrictive rule, which they will
not understand unless it is carefully explained to them. Therefore, I am
not suggesting that it makes political sense for liberals to stop defending
Lawrence and Romer as legally sound.

16.2 SELECTIVITY

Scalia underscores his objection to Lawrence and Romer by suggesting that

his brethren do not really endorse the broad constitutional principles
from which they claim to reason. In his Lawrence dissent he argues as
follows:
State laws against bigamy, same-sex marriage, adult incest, prostitution,
masturbation, adultery, fornication, bestiality, and obscenity are . . . sustain-
able only in light of Bowers’ validation of laws based on moral choices. Every
single one of these laws is called into question by today’s decision; the Court
makes no effort to cabin the scope of its decision to exclude them from its
holding. . . . The impossibility of distinguishing homosexuality from other
traditional “morals” offenses is precisely why Bowers rejected the rational-
basis challenge.19
Scalia makes a similar point in Romer:
The constitutions of [five states] to this day contain provisions stating that
polygamy is “forever prohibited.” . . . Polygamists, and those who have a polyg-
amous “orientation,” have been “singled out” by these provisions for much more
severe treatment than merely denial of favored status. . . . The Court’s disposi-
tion today suggests that these provisions are unconstitutional, and that
polygamy must be permitted in these States . . . unless, of course, polygamists
for some reason have fewer constitutional rights than homosexuals.20

19. Lawrence v. Texas at 586 (Scalia, J. dissenting).

20. Romer v. Evans at 648 (Scalia, J. dissenting).
Implementation 277

Acknowledging Scalia’s sarcasm, we can reconstruct his argument in

these dissents as the following reductio ad absurdum:

1. If the Texas sodomy statute and Amendment 2 of the Colorado

Constitution are unconstitutional, then at least some other morals
laws are unconstitutional.21
2. The Texas sodomy statute and Amendment 2 of the Colorado
Constitution are unconstitutional.
3. Therefore, at least some other morals laws are unconstitutional.

Scalia rejects the third proposition, intending his argument as a reductio

of the second premise.22 The majorities draft their opinions in ways that
suggest that they, too, reject the third proposition. One could, of course,
take a libertarian position against all morals laws, but the Lawrence and
Romer majorities do not. They do not mention other morals laws. Nor do
they mention, much less reverse, any precedents upholding such laws.
Nor do they attempt to distinguish the Texas and Colorado laws. Scalia
concludes that the second premise is false and that Romer and Lawrence
are, accordingly, constitutional mistakes, even in terms of the majority
justices’ own commitments.
The following hypothetical—similar to ones in the literature—
reinforces Scalia’s point.23 It is 2004, the year after Lawrence was decided.
An adult brother and sister are involved in a long-term romantic
relationship that includes consensual sexual intercourse.24 The sister has
had a hysterectomy, so there is no danger of pregnancy. They are convicted
of criminal incest and challenge the constitutionality of the state law,
citing Lawrence. Could the court distinguish the incest case from Lawrence?
As Scalia emphasizes, Lawrence does not contain any argument, much less
a sound one, for distinguishing such cases.25 Yet the majorities do not even
suggest that other morals laws are unconstitutional or that any prior
cases26 are inconsistent with Lawrence.

21. Morals laws encompass, inter alia, state laws against sodomy, bigamy, same-sex mar-
riage, consensual adult incest, prostitution, adultery, fornication, bestiality, and obscenity.
22. This argument should not be confused with Scalia’s more familiar opposition to
nonoriginalist modes of constitutional deliberation and the notion of a “living constitution.”
Cf. Antonin Scalia, “Originalism: The Lesser Evil,” University of Cincinnati Law Review 57
(1989): 849–65.
23. Such hypotheticals are discussed in John Corvino, “Homosexuality and the PIB
Argument,” Ethics 115 (2005): 501–34.
24. See Arthur P. Wolf and William H. Durham, eds., Inbreeding, Incest, and the Incest
Taboo: The State of Knowledge at the Turn of the Century (Stanford, Calif.: Stanford University
Press, 2005). A sexual relationship between half-siblings is the dramatic subject of Sam
Shepard, Fool for Love (New York: Dramatists Play Service, 1984).
25. Lawrence v. Texas at 590, 599 (Scalia, J., dissenting); Romer v. Evans at 636 (Scalia,
J. dissenting).
26. Other than Bowers v. Hardwick, of course.
278 Limits of Legality: Part II

Liberals who do not wish to challenge incest laws could look for
constitutionally relevant distinctions between the incest hypothetical and
the facts of Lawrence.27 The cases differ in two basic respects. The brother
and sister are siblings; John Lawrence and his partner were unrelated by
blood. The siblings engaged in sexual intercourse; Lawrence and his
partner engaged in sodomy. However, it is hard to see any constitutional
significance in these distinctions. Incest has long been criminalized in all
states, but sodomy was once criminalized in all states and was still crimi-
nalized in many as of 2003. Incest is widely regarded as immoral, but so
was sodomy. Certainly, sodomy is enjoyed by millions and plays a vital
role in millions of intimate relationships, whereas adult incest is probably
rarer by several orders of magnitude. But why should that matter? If
sodomy loses popularity, will banning it become constitutional again?
Siblings who copulate may have serious psychological problems, for all I
know, but will banning sodomy become constitutional again if sodomy
starts causing such problems?
Perhaps incest between two infertile adults has more negative effects
on other people than sodomy has. The parents of incestuous adult siblings
usually experience emotional distress, for example, if they learn of the
relationship. But so do parents who learn that their son is gay, in many
cases. So do parents whose offspring marry outside their faith or race. This
cannot be the issue.
This is not the place to analyze Justice Kennedy’s majority argument in
Lawrence. If you believe that it is legally sound, then I suggest the following
thought experiment: imagine that you consider decriminalizing incest
between infertile adults to be an important matter of social justice—as
important as you now consider decriminalizing sodomy to be. I submit
that you would not, under those conditions, find a parallel legal argument
for invalidating incest laws (as applied to infertile adults) notably weaker
than you presently find the Lawrence argument.28 The only reason I can
see why someone might want to uphold the incestuous siblings’ conviction
while reversing Lawrence’s is if she accepts a certain comparative moral
judgment: that the moral right of any two unrelated adults to practice
sodomy is more compelling than the right of any two adult siblings
to copulate. I cannot formulate a sound constitutional argument for

27. Writing before Lawrence was decided, Cass Sunstein tried to distinguish between
Amendment 2 and antipolygamy laws by emphasizing that Amendment 2 is based on status,
not conduct. Cass R. Sunstein, “The Supreme Court 1995 Term: Foreword: Leaving Things
Undecided,” Harvard Law Review 110 (1996): 6–101, p. 63. But “being polygamous” is no
less a status than is “being gay.” In any case, Sunstein’s distinction is useless in Lawrence, in
which the challenged statute proscribes conduct and makes no reference to status.
28. In reality, very few consider decriminalizing incest between infertile adults to be a
high political priority, and those who do are politically disorganized. If the day comes when
enough intelligent, informed people agree with them, then courts will probably invalidate
incest statutes, so applied.
Implementation 279

invalidating sodomy laws but not incest laws (as applied to infertile adults)
without simply assuming that the comparative moral judgment is true
and legally dispositive. Perhaps someone will produce a legal argument
that does not tacitly make this assumption, but I am pessimistic.
My point is that supporters of the Lawrence and Romer holdings who
favor other morals laws need not challenge the major premise of Scalia’s
reductio: that some other morals laws would be unconstitutional if the
Texas and Colorado laws were. Only if they accept restrictive rule do they
bear a burden to challenge Scalia’s premise. If, instead, they favor selective
optimization, then they can defend Lawrence and Romer as morally
permissible (albeit deviant) decisions without challenging Scalia’s premise.
They can concede that the majorities never explain how the Texas and
Colorado laws differ from other morals laws that they do not wish to
invalidate. They can admit that there may be no constitutional argument
for invalidating the Texas and Colorado laws without also invalidating
some other morals laws.
The Romer and Lawrence majorities could have confessed their inability
to draw a constitutional distinction between sodomy laws and incest laws,
while declaring their intention to uphold the latter and invalidate the
former. However, such confessions draw undesirable attention to the fact
that the court has deviated. Scalia’s dissents already drew attention to
that fact. Once a judge on a panel realizes that a majority of his brethren
are going to endorse a fallacious legal argument, despite his efforts to
persuade them otherwise, he has a moral reason to avoid drawing attention
to their deviation, as doing so may provoke mimetic failure. Judges have
reasons to keep quiet about deviation—their own and that of other judges.
Of course, they may also have reasons to point out deviation—judicial
candor is, after all, a virtue, and pointing out deviation keeps judges “hon-
est.” These reasons may sometimes be strong enough to justify announc-
ing that someone has deviated. Scalia apparently believed that his reasons
for candor in Romer and Lawrence were strong enough.

16.4 FINAL MODERATE RULE AND DEFAULT RULE

I have argued that judges who deviate should not usually announce it,
partly because deviation is widely condemned. Two other aspects of
selective optimization would also encounter resistance from lawyers and
the general public: final moderate rule and default rule. These rules make
the disposition of a case sometimes depend on factors that are not case-
type-specific. In final moderate rule the relevant factor is the number of
other cases in which the judge has deviated or anticipates deviating.
Default rule makes the outcome sometimes depend on random factors.
Both rules have an alarming implication—the outcome of a case could
depend on factors over which the parties have no control. No one likes
the idea that his fate depends on such factors. A crucial function of the
280 Limits of Legality: Part II

rule of law is to reduce the dependence of fate on circumstances beyond

our control. I can easily imagine what an outraged party might say upon
learning that the judge assigned to his case obeyed final moderate rule and
default rule:
Each case should be judged on its merits and nothing else. The fact that my
judge has already bent the rules for other folks doesn’t make my case any
weaker. He shouldn’t punish me for my position in the “line.” If anything,
the fact that the judge has bent the rules for other people entitles me to
have them bent in my case, too. And a coin flip certainly shouldn’t have any
bearing on the outcome of my case!

Understandable as this reaction is, no one who has fully internalized

selective optimization would react in this way. It is usually unfair when
someone’s fate is affected by forces beyond his control, but so is it usually
unfair to the losing party when the judge adheres in a suboptimal-result
case. The first source of unfairness does not always outweigh the second.
Selective optimization takes both sources into account and strives to
minimize the aggregate level of unfairness in the system. The outraged
party quoted above fails to appreciate that selective optimization could
also benefit him—in suboptimal-result cases in which the law disfavors
him. Perhaps he never expects to find himself on the losing end of a
suboptimal-result case.
I think selective optimization comprises a fair set of guidance rules. My
point in this chapter so far is that until enough people agree with me,
acquiring a reputation for selectively optimizing will only hurt a judge’s
career and increase the relative influence of other judges. For the foresee-
able future, therefore, judges should not publicize selective optimization.

16.5 PRECEDENT

I have made the unorthodox suggestion that judges who deviate should
sometimes file fallacious opinions. How should other courts treat such
opinions, given the doctrine of stare decisis? 29 Before answering this
question I must distinguish between deviation from the law and judicial
revision of legal doctrine. Consider cases in which the judge has legal
authority to revise legal rules created by lower courts or courts of coordi-
nate jurisdiction. Although judges in such cases have the legal authority to
revise rules that dictate suboptimal results, they should not always do so.
First, the rule might be an optimal rule, the present result notwithstand-
ing. In gap cases, by definition, rule revision is unjustified: the judge should
either adhere or deviate.

29. See Barbara B. Levenbook, “The Meaning of a Precedent,” Legal Theory 6 (2000):
185–240; Larry Alexander, “Constrained by Precedent,” Southern California Law Review 63
(1989): 1–64; Frederick Schauer, “Precedent,” Stanford Law Review 29 (1987): 571–605.
Implementation 281

Second, even if the rule should be revised, a decision to revise it has

systemic effects. Revising rules—optimal or suboptimal—can encourage
other judges to revise optimal rules. This is a form of mimetic failure. A
judge who revises rules bears some moral responsibility for it, just as she
bears responsibility for mimetic failure when she deviates. The norm of
stare decisis discourages an excessive rate of rule revision.30
When a judge in a common-law jurisdiction announces a new rule of
law or revises an existing rule, she intends to set precedent. She intends for
inferior judges to follow the new rule. She probably also intends to follow
it herself and to have courts of coordinate jurisdiction follow it. However,
she need not intend anyone—including herself—to follow the new rule in
every case. If she is moderately rational and informed, then she believes
that her new rule will dictate some suboptimal results. Gap cases can arise
under any legal rule. She may not even believe that her new rule is the
optimal rule—the one with the fewest gap cases of all those possible. She
probably just believes that her new rule dictates fewer suboptimal results
than the old rule. Otherwise, why would she have revised it? So she
intends to follow the new rule at least as often as she would have followed
the old rule, and she intends other judges to do likewise. This, I suggest, is
what it ordinarily means to set a precedent—a standard precedent.
By contrast, a judge who deviates does not intend to change the rule
from which she deviates. She need not be committed to deviating in all
similar cases, even those that are factually type identical to the case at bar.
Nor, a fortiori, does she intend that other judges should deviate in every
type-identical case. Selective optimization forbids judges to follow a policy
of deviating in every token of this case type if too many such tokens arise.
Therefore, a rational judge never intends to set a standard precedent by
deviating. However, she might intend to set a weaker kind of precedent.
She might intend to deviate in some similar cases and to increase the
probability that other judges will deviate in some similar cases. After all,
many similar cases will be, for the same kinds of reasons, suboptimal-
result cases—cases in which posterior judges have pro tanto moral reasons
to deviate that are just as strong as the anterior judge’s reasons. The ante-
rior judge will want posterior judges to heed those reasons and to deviate
as well, unless the system is at threshold. These intentions do not entail an
intention to deviate in every similar case or to have others do so.
Picture an appellate court announcing a new interpretation of a statute
or constitutional provision. The judges intend that other judges at their
level and below should follow the new interpretation. The new interpre-
tation will still generate gap cases, so rational judges do not intend that

30. There is, we might say, a “revision density threshold” akin to the deviation density
threshold. With any luck, the norm of stare decisis prevents courts from crossing the revision
density threshold.
282 Limits of Legality: Part II

others should always follow it. In fact, they may even anticipate that their
new interpretation will generate more gap cases than the old. They may
have aimed for an accurate reading of text or legislative purpose rather
than aiming to minimize gap cases with their interpretation. But the
judges intend to follow their interpretation at least as often as they follow
comparable precedents: precedents of similar age that generate a similar
number of suboptimal-result cases (adjusted, perhaps, for the severity of
the suboptimal results). They intend for others to do likewise.
By contrast, when an appellate court deviates in a suboptimal-result
case, the judges cannot rationally intend to set a standard precedent.
Rational judges cannot intend that posterior judges should deviate in sim-
ilar suboptimal-result cases as often as they follow other precedents. The
judges can rationally want posterior judges to deviate in similar subopti-
mal-result cases, but to do so subject to selective optimization. The judges
might also intend to increase the probability of such warranted deviation.
What sort of precedent did the Griswold majority intend to set, a
standard one or a weak one? The tempting answer is: a standard one. I
think the tempting answer is mistaken, but I must first explain why it is
tempting. The Griswold majority certainly intended courts at every level
to treat the opinion as controlling with respect to similarly situated parties
and other state bans on the use of contraceptives as applied to married
couples. The decision was not limited to Estelle Griswold or the State of
Connecticut. As to similar bans on the use of contraceptives by married
couples, the majority intended to set a typical precedent and did so.
The important question, however, is whether the Griswold majority
could have rationally intended other courts to consistently apply their
reasoning. Suppose judges in Group O were to imitate Griswold consis-
tently: for any challenged state law, if they can find a legal argument for
invalidation as strong as the Griswold argument, then they invalidate the
law, using the corresponding argument in their opinion. Being Group O,
they reach only optimal results. However, using too many fallacious argu-
ments encourages judges outside Group O to use them. Eventually an
idiosyncratic outsider (inspired by judges who were inspired by judges
who were inspired by Group O) would use a fallacious argument to reach
a suboptimal result: perhaps invalidating laws against animal cruelty,
counterfeiting, or criminal solicitation when committed privately in the
home. That is why selective optimization does not support a policy of
deviating in every suboptimal-result case. The Griswold majority could
not rationally intend its reasoning to serve as a standard precedent, even
for Group O. Selective optimization permits only occasional forays into
fallacious legal reasoning.
The same goes for the Romer and Lawrence Courts. Perhaps they did
not intend for other courts to adopt their reasoning. They merely intend-
ed to invalidate statutes criminalizing sodomy between adults and state
constitutional amendments banning the recognition of sexual orientation
as a civil rights category, leaving other morals legislation undisturbed.
Implementation 283

Selective optimization does not forbid such selectivity in principle and

actually provides a principled rationale for it.

16.6 MINIMALISM

Selective optimization entails that judges are morally permitted to offer

fallacious legal arguments for optimal results when they cannot formulate
sound arguments, provided they exercise moderation. Selective
optimization also provides reasons for courts sometimes to write what
Cass Sunstein calls “minimalist” opinions. These reasons are distinct from
Sunstein’s own. First, I shall explain minimalism and review one of
Sunstein’s arguments for it.
Sunstein draws two distinctions: between narrow and broad opinions
and between shallow and deep opinions. A broad opinion announces a
general legal rule that could be applied to different cases. A narrow one
does not. A deep opinion articulates, as the bases for the holding, an un-
derlying principle that could be applied to different cases. A shallow one
does not. Courts often must choose between multiple arguments in
support of a given result. They could write a broad opinion or a narrow
one, a deep opinion or a shallow one.
Reasonable people can sometimes agree on a holding while disagreeing
about the correct principle upon which the decision should rest or the
best statement of the applicable rule of law. Similarly, a judge may find
that she is confident about a holding but less so about the underlying
principle or the best rule to announce. Judges should not have great con-
fidence in their own moral insights in difficult cases that raise novel issues,
Sunstein insists. He argues that in such cases, the Court should render
decisions that are narrow and shallow rather than wide and deep.31 An
ambitious opinion that offers wide and deep reasoning can do great good
as a precedent or great harm. A less ambitious opinion affords a greater
margin for error and revision in future generations.32 The opinion should
offer just enough reasoning to support the holding, but no more. It need
not offer deep rationales or explain how to distinguish similar cases.33
Sunstein’s judge intends to reach a result that she believes the law per-
mits and to write an opinion that reflects the actual state of the law or
announces a new rule. She faces a choice between minimalist and non-
minimalist opinions, but she believes that both opinions contain sound
legal arguments. Sunstein believes that judges write minimalist opinions
when they lack confidence in the underlying moral judgments. He sug-
gests that the Romer majority declined to articulate the differences

31. Cass R. Sunstein, One Case at a Time: Judicial Minimalism on the Supreme Court
(Cambridge, Mass.: Harvard University Press, 1999), pp. 10–14.
32. Ibid., ch. 3.
33. Ibid., chs. 1–4.
284 Limits of Legality: Part II

between Amendment 2 and other morals legislation—such as the

entrenched antipolygamy provisions of several state constitutions—be-
cause although they agreed that relevant differences existed, they could
not agree on what those differences were.34 He could explain Lawrence in
the same way.35
However, these explanations would not satisfy Scalia, nor do they sat-
isfy me. The justices could have authored separate opinions, as they often
do, each explaining how he or she distinguishes between the invalidated
law and other morals laws. No justice did so. So I still think it reasonable
to infer that they did not know how to draw a constitutional distinction
between sodomy laws and, for example, laws banning adult incest. Sun-
stein neglects the possibility that Romer and Lawrence were suboptimal-
result cases. I cannot tell whether he believes that judges should ever
deviate in suboptimal-result cases. In any event, I shall argue that subop-
timal-result cases provide another reason for judges to write minimalist
opinions. Although courts should always offer the best legal arguments
they can find for their conclusions, there are no sound arguments—
minimalist or otherwise—for the optimal result in suboptimal-result
cases. A judge who chooses to deviate in such a case could, in theory,
present no argument whatsoever, as I imagined the Griswold Court doing
earlier, but this approach draws unwanted attention to her deviation. On
the other hand, if she writes a nonminimalist opinion, then she propa-
gates fallacious legal reasoning. Less enlightened judges may try to imitate
this reasoning and be led to suboptimal results.
A minimalist opinion could represent a happy compromise between
these unattractive extremes. Of course the argument will still be fallacious—
ex hypothesi, no sound argument for the holding exists. But a well-written
minimalist opinion, even a fallacious one, has at least the superficial appear-
ance of legal reasoning. It disguises its own fallaciousness. But precisely
because its reasoning is so shallow and narrow, it gives other judges less
material with which to work and thus makes them less likely to treat it as a
standard precedent. Therefore, minimalist arguments may be the best kind
for a judge to use if she wants to reach an optimal result in a suboptimal-
result case. Perhaps the justices joined minimalist opinions in Lawrence
and Romer because they could not think of any sound arguments for the
holdings.
Sunstein does not condone courts making fallacious legal arguments,
minimalist or otherwise. He advocates minimalist opinions that contain
“just enough” reasoning to support the holding. But lawyers disagree about
how much reasoning is enough. Critics of the opinions praised by Sun-
stein could object that these opinions do not, in fact, contain enough

34. Ibid., ch. 7.

35. In fact, he appeals primarily to the infrequency of enforcement of sodomy laws in
Texas. Cass R. Sunstein, “What Did Lawrence Hold? Of Autonomy, Desuetude, Sexuality,
and Marriage,” Supreme Court Review 2003 (2003): 27–74.
Implementation 285

reasoning to support the holdings. I shall not ask whether Sunstein or his
critics have the stronger argument with respect to any particular case. My
point is that there are two situations in which a judge might want to use
minimalism. In the situation described by Sunstein, the judge is confident
that his intended result is legally permissible but not confident about any
given argument. In my situation, by contrast, the judge is not confident
that his intended result is legally permissible. He may even be confident
that it is not. In that situation he may want to avoid raising similar doubts
in other people and to minimize the quantity of fallacious legal reasoning
that he introduces into the jurisprudential corpus. That is a different
reason to write a minimalist opinion.
Opponents of minimalism believe that a sound, nonminimalist argu-
ment for a legally defensible conclusion is always possible and should be
used instead of minimalist arguments. They accuse judges of using mini-
malism to reach results for which no strong legal arguments exist. I suggest
that even if these critics are correct, we should not conclude that judges
act impermissibly if they write minimalist opinions to reach deviant
results: if they use minimalism pursuant to selective optimization, then
their actions may be permissible.
In addition to writing minimalist opinions, judges who plan to selec-
tively optimize might consider filing unpublished opinions—ones not
intended for citation by other courts—or publishing only selected parts of
certain opinions. These techniques could be especially useful in gap cases,
in which the court wants to avoid suboptimal results without changing
optimal rules. There are notable arguments against filing unpublished
opinions, but doing so could facilitate selective optimization and may be
worth contemplating.36

16.7 COGNITIVE AND PSYCHOLOGICAL BURDENS

I turn now to the second question of this chapter: are judges psychologi-
cally capable of internalizing selective optimization? Insofar as selective
optimization allows judges to deviate more frequently than does restric-
tive rule, it seems to impose fewer demands. However, it also imposes
greater cognitive and psychological burdens on judges. Restrictive rule
merely requires judges to find and apply the law. Of course this can be
challenging if the law is complicated or unclear, and if the law explicitly

36. See, e.g., Arthur J. Jacobson, “Publishing Dissent,” Washington and Lee Law Review
62 (2005): 1607–36; Elizabeth M. Horton, “Selective Publication and the Authority of Prec-
edent in the United States Courts of Appeals,” UCLA Law Review 42 (1995): 1691–778;
William L. Reynolds and William M. Richman, “The Non-Precedential Precedent—Limited
Publication and No-Citation Rules in the United States Courts of Appeals,” Columbia Law
Review 78 (1978): 1167–208.
286 Limits of Legality: Part II

grants the judge discretion, as it usually does in criminal sentencing, for

example, then she faces some additional work after she finds the law. But
once she finds it, restrictive rule does not invite her to consider deviating.
In that respect it is easier to obey than selective optimization. Selective
optimization also requires the judge to find the law when there is law to
find, but it assigns her some additional jobs. She must determine the best
result notwithstanding the law. She must monitor her own deviation rate
and decide when to deviate. These tasks may be intellectually demanding.
Furthermore, a judge who consciously optimizes lives with some unset-
tling self-knowledge. He knows that he sometimes deviates from the law
and that some of his filed opinions misrepresent his actual reasoning—he
publishes lies. He knows that his decision patterns are type selective,
maybe even token selective. They may even involve random decision. For
all these reasons some judges will find that consciously following selective
optimization would lower their self-esteem or cause them mental distress.
Some will be unable to reflectively endorse selective optimization for
these reasons.
These tasks and this self-knowledge will weigh more heavily upon
some judges than others, according to variations in psychological profile.
Bearing these burdens requires a combination of self-confidence, humility,
intelligence, initiative, and willingness to risk one’s reputation for the sake
of others. There may be judges who, because of their individual psychol-
ogies, will reach decisions that approximate selective optimization less
closely if they try to optimize than if they use some other method. Perhaps
the cognitive and psychological demands overwhelm them: they make
too many mistakes or become too distressed or frustrated to function. For
them, selective optimization may be a self-effacing method: one that can-
not serve its purpose if they consciously attempt to follow it.37 Such judg-
es should not consciously obey selective optimization. Not everyone is
cut out to be a double agent, or a Marine officer, or a surgeon, or a crimi-
nal defense attorney, either. To demand that judges self-consciously opti-
mize may be asking too much. I am not confident that judges are
blameworthy if they shrink from the demands of optimization. I contend
only that they are morally permitted to optimize and are rarely blame-
worthy for doing so if they use selective optimization. I believe that judg-
es with the right psychological profile should consciously try to optimize,
perhaps surreptitiously, and to rationalize their deviation and selectivity
as best they can in their opinions.

16.8 MAINTAINING EQUILIBRIUM

A critic of selective optimization might object that even if typical judges

have the cognitive skills to consciously pursue it, we cannot count on

37. This usage of self-effacing originates in Parfit, Reasons and Persons, pp. 23–24.
Implementation 287

them to do so. Suppose publication of this book leads judges to weaken

their commitment to restrictive rule. If they decide to selectively opti-
mize instead, then my critic’s objection fails. But the objection succeeds if
the book misfires: if its effect is that many judges become less committed
to restrictive rule, but they also snub my critique of permissive rule and
my arguments for selective optimization. They proceed to “overindulge”
in deviation, damaging the rule of law. Maybe judicial commitment to
restrictive rule is like a fence that keeps the herd from heading for the
edge of the cliff. Even if the fence seems needlessly far from the edge,
removing it would be disastrous if no other fence lies beyond it and the
herd is ready to stampede.
The preceding objection makes an empirical claim that may be cor-
rect, for all I know. I shall merely offer some reasons for skepticism.
My first reason is that judicial behavior already reflects less than perfect
compliance with restrictive rule. Everyone agrees that there have been
some deviant rulings over the years, despite disagreement about which
ones were deviant. So far judges in the United States have not destroyed
the rule of law with their deviation, although they may have damaged
it.38 Most judges claim allegiance to restrictive rule, but even those who
endorse it sometimes deviate in practice. At least some of the deviation
in U.S. history has surely been self-conscious, even ignoring the extremely
rare cases of candid deviation. No one can read a judge’s mind, but the
more intelligent and experienced the judge and the more obvious her
deviation, the more confident one can be that she knew she was deviat-
ing, even without a signed confession. Perfect compliance with restric-
tive rule has never existed and is not necessary for the rule of law to
survive.39
My critic could revise her objection. She need not make the hyperbolic
claim that anything less than perfect compliance with restrictive rule
destroys the rule of law. Rather, she could argue that reducing current
levels of commitment to restrictive rule as a social norm will destroy the
rule of law. At least some judges would prefer to deviate more often than
they do in suboptimal-result cases. Incentives surely provide at least part
of the explanation for their relative self-restraint. Formal sanctions for
deviation hardly exist, but judges have several other incentives to adhere,
as discussed in chapter 4. Researchers have only recently begun creating
testable models of the incentive mechanisms that bear on judicial behavior,
using resources from economics, game theory, and the theory of social
norms. Little research has studied adherence incentives in general, as
opposed to stare decisis, in particular.

38. See generally Brian Z. Tamanaha, Law as a Means to an End: Threat to the Rule of
Law (New York: Cambridge University Press, 2006).
39. See Margaret Jane Radin, “Reconsidering the Rule of Law,” Boston University Law
Review 69 (1992): 781–819 (discussing how the rule of law survives without “formalism”).
288 Limits of Legality: Part II

Most judges want to maintain good reputations. Lower court judges

may also want to avoid reversal and to be promoted. Each of these incentives
depends on social norms condemning deviation. Reputational incentives
do not exist unless judges believe that they are more likely to suffer repu-
tational losses if they deviate. The other incentives do not exist unless lower
court judges believe that they are more likely to be reversed or to lose
potential promotions if they deviate. As the social norm of restrictive rule
weakens, tolerance for deviation increases, with a commensurate drop in
the associated risk of reversal and cost to one’s reputation and career.
Judges who adhere mainly for prudential reasons begin deviating more
often, eventually pushing the system over threshold.
Again, it is an empirical question whether wider rejection of restrictive
rule as a social norm would have these effects. It is possible. But it is also
possible that judges will still have incentives to adhere. There is no reason
to anticipate that incentive mechanisms will disappear as restrictive rule
becomes less popular. At least, there is no reason to predict this outcome
provided that enough morally motivated people embrace selective opti-
mization instead. Just as those who accept restrictive rule currently criti-
cize judges who deviate, so will those who adopt selective optimization
criticize judges who deviate more often, or more conspicuously, than the
theory allows. Appellate courts that accept selective optimization will
freely reverse deviant decisions by such judges. Observers who accept
selective optimization will think less highly of such judges and will
promote them less often. For these reasons, approximate conformity to
selective optimization could emerge as a stable equilibrium, even for
judges whose motivations are entirely prudential and who reject selective
optimization themselves.
However, there is a substantial difference between restrictive rule and
selective optimization. Determining whether someone has violated
selective optimization is more difficult than determining whether she has
violated restrictive rule. For that reason alone, the fraction of judges who
violate selective optimization and meet with criticism, if selective optimi-
zation is widely accepted, will probably be smaller than the fraction of
judges who violate restrictive rule and meet with criticism, if restrictive
rule is widely accepted. Perhaps there will be so little criticism of judges
who violate selective optimization that adherence incentives will diminish
too much and the system will cross threshold.
I admit that this is a conceptual possibility. The publication of my book
could, in theory, push our legal system over the cliff. But this seems un-
likely. The scenario just described occurs only if some special conditions
obtain. The only dangerous individual, as I see it, rejects restrictive rule
because of my book, but never adequately internalizes selective
optimization. He is too lazy or cognitively impaired. He criticizes deviation
less often than selective optimization prescribes. If he is a judge, then he
is also too unresponsive to adherence incentives. He deviates more often
than selective optimization permits. If there were too many people like
Implementation 289

that—judges, lawyers, scholars, commentators, and politicians—then my

book could send them over the cliff.
I see no reason to believe, however, that many people fit this profile. I
argued that judges have no moral or legal obligation to obey restrictive
rule. No judge who would choose to slog through this book could be
obtuse enough to conclude that I had shown his adherence incentives to
be weaker than he had believed them to be. Similarly, any observer of the
courts who rejects restrictive rule because of this book probably has the
cognitive skills to use selective optimization criteria for praise and blame.
The only type of person who worries me is a critic of the bench who
rejects restrictive rule because of this book, but is lazy and simply stops
criticizing judges for deviating in suboptimal-result cases, even when they
deviate more often than selective optimization permits. I hope there are
few such people, but they might exist.

16.9 CONFORMING UNREFLECTIVELY TO DEFAULT RULE

I recognize that some judges—perhaps most—lack a psychological profile

that facilitates conscious optimization. If too many reject restrictive rule
without adopting selective optimization, then the rule of law could be
destroyed. Therefore, a judge should not consciously reject restrictive rule
if the cognitive and psychological burdens of selective optimization seem
too great for him.
However, I am curious about mechanisms for enabling judges to inch
closer to conformity with selective optimization without consciously
following it. Consider a set of cases heard by Judge Janet during her career.
In each of these cases the law requires results that Janet believes to be
suboptimal and equally so. If Janet recognizes this fact, then she will face
a conundrum. She believes herself to have a pro tanto reason, and an equally
strong one, to reach the optimal result in each case. If she recognizes that
the law forbids this, then she will conclude that she has an equally strong
pro tanto reason to deviate in each case. However, she also believes (cor-
rectly) that it would be wrong to deviate in so many cases and that doing
so would ruin her reputation. In such situations, default rule recommends
randomly choosing a permissible fraction of the cases for deviation.40
However, Janet believes that random deviation is wrong as a matter of
principle. She believes that judges should not “dice” with justice.
I have argued that random deviation is not wrong as a matter of princi-
ple. I still believe that, but because Janet disagrees with me, I want to know
if she can decide correctly without consciously violating any of her own
principles. I think she can. She can achieve the same results as random

40. See §15.4.

290 Limits of Legality: Part II

deviation without having to consciously randomize. She can do so by

acquiring the false belief that the law permits her to reach optimal results
in a certain fraction of her suboptimal-result cases. This is psychologically
possible because the fact patterns of the cases differ from one another in
various ways. If we stipulate a set of suboptimal-result cases, then none of
these differences actually justifies reaching the optimal result, by defini-
tion. But Janet might easily reach the mistaken conclusion that some are
optimal-result cases. She might perceive a legal mirage: the appearance of
legally relevant factors where none actually exists. Judicial perception of
legal mirages can serve as a proxy for conscious randomizing. In some frac-
tion of the cases, the judge perceives some “special features” that seem to
her to justify reaching the optimal result as a matter of law. Recall the
statutes that banned hate speech and nude dancing, respectively.41 When
deciding these cases, Judge Janet might find that hate speech “seems” sim-
ilar to certain other speech, the banning of which has been previously
upheld, whereas nude dancing “seems” less similar to her, although she
may not be able to put the perceived difference into perspicuous
language.
Some people believe that experienced lawyers possess special capac-
ities that enable them to ascertain with confidence that an answer to a
legal question is legally correct or legally superior to the alternatives when
equally intelligent individuals, who have access to the same source mate-
rials but lack comparable training and experience, reach the opposite con-
clusions, or reach their conclusions with less confidence. These capacities
are sometimes described in perceptual or visual terms (“seeing the right
answer”), sometimes in cognitive terms (“thinking like a lawyer”). They
are sometimes said to be irreducible to propositional knowledge. In this
respect, the experienced lawyer’s putative capacities invite comparison to
certain forms of “knowledge how,” or aspect seeing,42 or gestalt percep-
tion.43
Believing that one possesses these capacities enables one to conform to
default rule without consciously obeying it. Imagine a judge who is com-
mitted to avoiding suboptimal results and to deciding cases only on the
basis of case-type-specific factors. A judge with these commitments will
experience psychological pressure to conclude that she never hears sub-
optimal-result cases. There are two logical ways for her to reach this
conclusion. First, she could conclude that the legally required result is not,

41. See §14.6.

42. Malcolm Budd, “Wittgenstein on Seeing Aspects,” Mind 96 (1987): 1–17.
43. See, e.g., Scott Brewer, “Exemplary Reasoning: Semantics, Pragmatics, and the Ratio-
nal Force of Legal Argument by Analogy,” Harvard Law Review 109 (1996): 923–1028;
Brian Leiter, “Heidegger and the Theory of Adjudication,” Yale Law Journal 106 (1996):
253–82; Anthony Kronman, The Lost Lawyer: Failing Ideals of the Legal Profession
(Cambridge, Mass.: Belknap, 1993), pp. 170–85; Charles Fried, “The Artificial Reason of the
Law Or: What Lawyers Know,” Texas Law Review 60 (1981): 35–58.
Implementation 291

in fact, suboptimal. Alternatively, she could conclude that there are legally
relevant features of the case type that justify the optimal result.
A judge who thinks she possesses the aforementioned capacities is
more likely to perceive legal mirages in one or both case tokens that
help her to justify to herself her decision to deviate in one token but not
the other. These mirages emerge from intrinsic features of the case
token. We should not assume that the judge’s belief in her special
capacities is insincere. On the contrary, the more sincerely held the
belief, the more effectively can it ameliorate cognitive dissonance for
judges who believe that adjudication should ignore features that are
not case-type specific, but who also want to avoid suboptimal results.
Legal mirages provide judges with resources for rationalizing selective
deviation, even token-selective deviation, in suboptimal-result cases,
thereby facilitating conformity to selective optimization without con-
sciously obeying it. Judges who make proper “use” of the mirages they
see can achieve these goals without suffering the cognitive dissonance
associated with consciously adopting default rule. A judge who uses
mirages believes that she sees legally relevant, intrinsic factors that dis-
tinguish the case tokens. This belief enables her to conclude that she
has both adhered to the law and reached a morally optimal result.
In fact, she has done only the latter, but she may sleep more soundly if
she believes otherwise.44
With these regrettably speculative remarks, I must leave the topic of
implementation.

44. Compare Henry S. Richardson, discussing the “evicted widow” case: “Theoretically,
an optimal intermediate solution may exist that would evict only a certain percentage of
such widows—either on the basis of an arbitrary randomizing device or (more likely within
the law) on the basis of a more finely specified principle.” Henry S. Richardson, “Review of
Alan H. Goldman, Practical Rules,” Notre Dame Philosophical Reviews (2002).
17

Theoretical Implications

17.1 FORMALISM

Several scholars have recently defended the view that judges should treat
legal rules as “serious rules.” This means applying the rule without looking
behind it to the reasons that support it or, a fortiori, to other reasons.
Frederick Schauer calls his version of this view presumptive positivism,
which assigns presumptive priority to the most “local” rule.1 He characterizes
positivism as entailing “systemic isolation” and claims that a positivist
system is the systemic analog of a rule.2 Alexander and Sherwin defend a
related view, formalism.3 Unlike these three theorists, Alan Goldman does
not believe that judges should treat legal rules as serious rules, but he
argues that judges should treat as a serious rule what he calls the “funda-
mental rule of the legal system,” which is what I call restrictive rule.
My primary objection to Schauer’s position is that he assumes that
authority to deviate includes authority to define what constitutes deviation.
On this assumption, deviation invades the jurisdiction of the rule maker,
just as Schauer concludes. But I think judges can deviate without transfer-
ring jurisdiction from the rule-maker to the judge. They can do this if they
obey selective optimization and the rule-maker retains jurisdiction to
define what constitutes deviation in the first place. The judge has no juris-
diction to make that decision and that decision is still binding on the
judge, albeit in an indirect way.
I mentioned earlier that obeying rebuttable presumption is equivalent
to taking a rule-sensitive particularist or presumptive positivist approach
to restrictive rule.4 Since then I have defended selective optimization.
Selective optimization is like permissive rule and unlike restrictive rule in

1. Frederick Schauer, Playing by the Rules: A Philosophical Examination of Rule-Based

Decision-Making in Law and in Life (Oxford: Oxford University Press, 1991), p. 191.
2. Ibid., p. 199.
3. See also Antonin Scalia, “The Rule of Law as a Law of Rules,” University of Chicago
Law Review 56 (1989): 1175–88 (connecting formalism to textualism and the rule of
law); Frank Easterbrook, “Ways of Criticizing the Court,” Harvard Law Review 95 (1982):
802–32.
4. See §14.3.

292
Theoretical Implications 293

that it allows judges to deviate in some suboptimal-result cases. But it is

also rule-sensitive, unlike permissive rule and like restrictive rule, in that
it aims to preserve the rule of law and thus does not allow judges to follow
a policy of deviating in all suboptimal-result cases. The question remains
whether objections that others have raised against rule-sensitive particu-
larism and presumptive positivism apply to selective optimization. As a
quasi-particularist view, does selective optimization inherit the flaws of
rule-sensitive particularism?
Discussions of rule-sensitive particularism in the literature have tended
to address only the crudest possible version of the view, which I call causal
rule-sensitive particularism. This is what Goldman, Schauer, Alexander,
and Sherwin criticize.5 Alexander and Sherwin conclude that rule-sensitive
particularism is self-defeating if publicized (i.e., it is esoteric):

[R]ule-sensitive particularism in regard to coordination rules narrows the

gap only when most people avoid particularism and treat legal rules as seri-
ous rules. . . . [R]ule-sensitive particularism cannot be publicized as the cor-
rect approach to rules. It can only be effective when rules are presented, and
generally accepted, as serious rules.6

They argue that rules can only serve their “settlement function” if we obey
them blindly, without consulting the moral reasons behind them in each
case. Therefore, they conclude that legal rules must be treated as serious
rules, which entails obeying restrictive rule.
Alexander and Sherwin seem to have in mind an alternative to restric-
tive rule that instructs each judge to take into account the value of having
rules and the harm that deviating will cause to that value. Call this causal
moderate rule. Causal moderate rule permits judges to deviate whenever
the net benefits to the parties outweigh the marginal systemic effects.
Alexander and Sherwin’s position entails that judges should not obey
causal moderate rule and that it should not be publicized as the correct
adherence rule. I agree. Causal moderate rule licenses so much deviation
that it would be collectively self-defeating for the judges in Group O to
obey it. A larger number of suboptimal results will be reached if they
obey causal moderate rule than if they obey restrictive rule. Causal mod-
erate rule fails because the only reason to adhere that it recognizes is harm
actually caused by a deviant decision. This is a defect because the morally
relevant relationship between a deviant decision and the systemic effects
of deviation is not so direct. A viable adherence rule must assign relevance
to the overall systemic effects of the pattern to which a deviant decision
contributes. Causal moderate rule fails to do so.

5. Alan H. Goldman, Practical Rules: When We Need Them and When We Don’t (Cambridge:
Cambridge University Press, 2002); Schauer, Playing by the Rules; Larry Alexander and Emily
Sherwin, The Rule of Rules (Durham, N.C.: Duke University Press, 2001).
6. Alexander and Sherwin, The Rule of Rules, p. 67.
294 Limits of Legality: Part II

Alexander and Sherwin might also conclude that just as “most people”
must avoid particularism and treat legal rules as serious rules, so must
“most judges” obey restrictive rule. But their reasoning does not support
this conclusion. It supports only the conclusion that most judicial decisions
in a given system must follow legal standards. One way to maintain this
pattern is for most judges to obey restrictive rule, perhaps knowing that
there will always be a few particularist judges in their midst—the “irre-
sponsible renegades.” But other divisions of labor could maintain the same
pattern. An alternative division has everyone following the same rule:
usually adhere to legal standards, but not always. This is what selective
optimization prescribes.
We should distinguish the idea that judges must obey serious rules,
which is a sensible idea, from the idea that judges must treat first-order
legal rules as serious rules, which I have called into question. Alexander
and Sherwin’s claim that judges must obey restrictive rule conflates these
ideas. I contend that judges should treat selective optimization as a serious
rule. They should never disobey selective optimization even if it requires
them to reach suboptimal results in some cases.7 Unlike permissive rule,
selective optimization does not guarantee that a judge who obeys it will
never knowingly reach suboptimal results. Selective optimization merely
attempts to minimize the number of suboptimal results that the system
as a whole produces. However, selective optimization does not require
judges to treat first-order legal rules as serious rules. It permits judges to
deviate from such rules on some occasions.
I should reiterate that lawmakers may have conclusive reasons to codify
and promulgate restrictive rule rather than selective optimization. But
judges do not necessarily have an all-things-considered reason to obey
restrictive rule rather than an uncodified rule such as selective optimiza-
tion. Judges should treat selective optimization as a set of serious rules.

17.2 CONSTRAINTS AND DECISIONS

In this section, I shall consider an objection to selective optimization based

on Scott Shapiro’s work on the nature of rule following. Shapiro distin-
guishes between the decision model of rule following and the constraint
model, defending the latter. He argues that when a judge genuinely commits
herself to the law, she makes subsequent deviation “infeasible.” Shapiro
does not intend infeasible as a metaphor. He means that it is literally
impossible for a judge who is committed to the law to deviate for a reason.
One who deviates for a reason is not, ipso facto, committed to the law.8

7. Obeying selective optimization as a serious rule is extensionally equivalent to “obey-

ing” restrictive rule and defining obedience in terms of a noncausal version of rule-sensitive
particularism.
8. Scott J. Shapiro, “Judicial Can’t,” Noûs 35, Supp. 1 (2001): 530–57.
Theoretical Implications 295

Shapiro suggests two mechanisms by which commitment to the law

renders deviation infeasible. First, adopting a rule activates the psycholog-
ical mechanism of repression. The judge becomes “unaware of reasons for
breaking the rule,” rendering her “unable to break the rule for a reason.”
Second, even a judge who remains “aware of important reasons for
breaking the rule” becomes “unable to withstand certain emotional pres-
sures, such as guilt and shame,” after she adopts the rule.9
The following implication appears to follow from the constraint model:
if a judge commits herself to the law, then she has no choice but to obey
restrictive rule. Breaking restrictive rule becomes infeasible for her. “By
committing to the legal system’s rule of recognition,” Shapiro writes, “a
judge will thus be constraining her future self to implement the rules that
share these authoritative features.”10 He makes other similar remarks:
It is rational for the committed judge to apply the law in any given instance
because that is the only option available. . . . [I]f judges are constrained to apply
the law, they must apply it regardless of its consequences.11
If the Constraint Model of rules is correct, then judges cannot be morally obli-
gated to decide individual cases according to the balance of reasons. Nor can
they be legally obligated to do so.12

I am not yet prepared to endorse the constraint model, but it has some
plausibility. Regardless, I do not believe that the constraint model entails
that judges who commit themselves to the law have no choice but to
obey restrictive rule. The constraint model is a generic model of rule guid-
ance. It does not entail any particular adherence rule such as restrictive
rule. Suppose the constraint model and selective optimization are both
true. If so, then a judge who commits himself to the law has no choice but
to obey selective optimization. Disobeying selective optimization becomes
infeasible for him, but selective optimization does not forbid all deviation
in suboptimal-result cases as restrictive rule does.
Now Shapiro could object that the constraint model entails that selec-
tive optimization cannot guide judges. According to his model, rules guide
by two mechanisms: repression and the disabling of normal inhibitions.
Shapiro could claim that selective optimization is incompatible with these
mechanisms. Selective optimization, unlike restrictive rule, sometimes
permits judges to consider the full range of reasons to deviate. Selective
optimization also requires each judge to monitor his own deviation record,
but if he has deviation credits remaining, then selective optimization
permits him to act on reasons to deviate. In other words, selective optimi-
zation does not always require judges to repress these reasons from their
consciousness. Deviating for reasons remains a feasible option.

9. Ibid., pp. 552–55.

10. Ibid., p. 550.
11. Ibid., p. 551.
12. Ibid., p. 555.
296 Limits of Legality: Part II

Similarly, the constraint model entails that rules can guide only if the
agent’s “normal psychological inhibitions” are disabled and she anticipates
too much guilt or shame to consider breaking the rule. By contrast, selec-
tive optimization sometimes allows the judge’s normal psychological
inhibitions against mistreating other human beings to remain active and
to override any guilt or shame that might be associated with being a “law-
less judge.” Again, deviating for reasons remains feasible. Shapiro’s claim
would be that selective optimization cannot guide judges because the
content of selective optimization interferes with the only two mecha-
nisms by which rules can guide agents.
It is true that a judge does not make deviating infeasible for himself by
committing to selective optimization. That would defeat the purpose of
selective optimization. But this commitment could make it infeasible for
him to treat, as all-things-considered reasons to deviate, certain facts that
would otherwise constitute such reasons. For a judge who is committed
only to permissive rule, the fact that Yasmin does not deserve to be evicted
constitutes an all-things-considered reason not to evict her. However, for
one who is committed to selective optimization this fact constitutes a con-
ditional reason: a reason that must be enabled by certain conditions. It is a
reason for Judge Lucas not to evict Yasmin if selective optimization permits
him to act upon that reason—if he has deviation credits left, et cetera. I can
accept Shapiro’s theory of rule guidance and assert that committing one-
self to selective optimization makes it infeasible to treat facts about
Yasmin’s situation as all-things-considered reasons to deviate in her favor.
Now Shapiro might reply that this is psychologically untenable. If, as I
stipulate, Judge Lucas is sufficiently aware of Yasmin’s plight to treat it as
a conditional reason to deviate, then how can he be unaware of it as an
all-things-considered reason to deviate? My reply is that this is precisely
what selective optimization entails, and that Shapiro’s own constraint
model cannot achieve what he wants unless judges ordinarily sustain
similar psychological compartmentalization. Shapiro cannot, and need
not, claim that a judge who adheres to the law in Yasmin’s case is, in every
sense, unaware of the facts that favor ruling for her. Imagine that we ask
Judge Lucas, “Does Yasmin deserve to be evicted as a moral matter?” If he
is committed to the law, then what honest responses are available to him,
according to the constraint model? Here are some possible responses:
“I have no idea whether she deserves to be evicted as a moral matter. All I
know is what the law tells me: she must be evicted.”
“I think I have an opinion about whether she deserves to be evicted as a
moral matter, but I can’t presently remember what my opinion is.”

“I’m deciding her case, so I can’t think about that right now. Ask me later.”

Each of these responses comports with the constraint model, but none is
necessary. This is fortunate, because none is especially realistic as a matter
of human psychology. Here is a more realistic reply:
Theoretical Implications 297

“No, she doesn’t deserve eviction morally speaking. I know that perfectly
well, but as a judge I can’t treat it as a dispositive reason to rule in her favor.
Therefore, I refuse to think about it when I deliberate in chambers.”13

That sounds like a more honest answer to me. But if the constraint model
allows the judge to offer this reply, then I think it also allows him to offer
a reply that reflects a commitment to selective optimization:
“No, she doesn’t deserve eviction, morally speaking. I know that perfectly well.
But I’ve bent the law an awful lot this month. As a judge, I can’t treat Yasmin’s
plight as a reason to rule in her favor if I’ve been deviating too often recently.
So I refuse to think about her plight when I deliberate in chambers.”

It is commonly assumed that a judge who deviates in a suboptimal-result

case does so because he has treated the suboptimality of the legally
required result as a self-sufficient reason to deviate. Only if this is true
does the constraint model entail restrictive rule. But it is not necessarily
true. A judge can treat suboptimality as one of several individually insuf-
ficient but jointly sufficient conditions on deviation. Suboptimality may
require enablers before it constitutes a reason to deviate. It is only partly
accurate to say that a judge who obeys selective optimization deviates
“because the law requires a suboptimal result.” A more accurate statement
is that he deviates “because the law requires a suboptimal result and he
has deviation credits left.” The latter phrase denotes what Jonathan Dancy
calls a combination of a “favorer” and an “enabler.”14

17.3 THE PRACTICAL DIFFERENCE THESIS

The question then arises whether, according to selective optimization,

first-order legal standards can actually guide judges. Consider Shapiro’s
conception of rule guidance: “An agent is being ‘guided’ by a rule when
that agent takes the fact that the rule is applicable as a conclusive reason to
follow it” (emphasis in original).15 Elsewhere Shapiro asserts that “A legal
rule R guides a person P to do A only if P might not have done A if he had
not appealed to R as a legal rule.”16

13. “[J]udges may adhere to precedent without introspective analysis of the effects of
the instant case as a means of avoiding distress and guilt over issuing difficult holdings.” Eric
Talley, “Precedential Cascades: An Appraisal,” Southern California Law Review 73 (1999):
87–137,” p. 109. See also Scott Altman, “Beyond Candor,” Michigan Law Review 89 (1990):
296–351, p. 305.
14. Jonathan Dancy, Ethics without Principles (Oxford: Oxford University Press, 2004),
pp. 38–43. Roger Crisp calls these combinations “ultimate reasons.” Roger Crisp, “Particular-
izing Particularism,” in Moral Particularism, eds. Brad Hooker and Margaret Olivia Little
(Oxford: Oxford University Press, 2000), p. 37.
15. Shapiro, “Judicial Can’t,” p. 535.
16. Scott J. Shapiro, “On Hart’s Way Out,” in Hart’s Postscript, ed. Jules Coleman
(Oxford: Oxford University Press, 2000).
298 Limits of Legality: Part II

This conception of rule guidance undergirds Shapiro’s defense of the

practical difference thesis, which states, “Legal rules must in principle be
capable of securing conformity by making a difference to an agent’s prac-
tical reasoning.”17 Shapiro claims that many positivists, notably H. L. A.
Hart, are committed to both the practical difference thesis and the conven-
tionality thesis—the view that the master rule of the legal system is a
conventional rule.18 Shapiro argues that one cannot consistently endorse
inclusive positivism, the practical difference thesis, and the conventionality
thesis. Although Shapiro does not claim that all versions of inclusive posi-
tivism are indefensible, he favors retaining the practical difference thesis
and conventionality and abandoning inclusive for exclusive positivism.19
Even those who reject the practical difference thesis recognize that
legal rules sometimes secure conformity by making a difference to an
agent’s practical reasoning. So regardless of whether one accepts the
practical difference thesis, it is important to get clear what sort of practical
difference legal rules make when, as often, they make one.
Here is an illustration of rule guidance as Shapiro conceives it. Imagine
a master rule that specifies the following: “All statutes enacted by the
Virginia General Assembly are valid.” The assembly enacts a statute crim-
inalizing the purchase of beer on Sundays. Meg, a Virginia resident, is
guided by this provision of Virginia’s master rule and she is aware of the
statute. As a result she refrains from buying beer on Sunday. However, it
was physically possible for the state legislature not to enact (or to repeal)
the beer statute. In that case, even though Meg was guided by the master
rule she would not have appealed to the statute as a legal rule, for the
simple reason that the statute would not have been a legal rule in that
case. She might in that case have bought beer on Sunday. Therefore, the
statute guides Meg not to buy beer.
If Shapiro means “all-things-considered reason” by “conclusive reason” in
his statement of the practical difference thesis, then selective optimization
entails that the fact that a legal standard is applicable is not always a con-
clusive reason for a judge to apply it. If a standard requires a suboptimal
result and the judge has deviation credits left, then she could have a stronger

17. Ibid., p. 129.

18. See, e.g., Jules Coleman, The Practice of Principle (Oxford: Oxford University Press,
2001), p. 68.
19. Inclusive positivists have offered several responses to Shapiro’s arguments. Some
have disavowed the practical difference thesis. See Ibid., pp. 137–48; Matthew Kramer,
“How Moral Principles Can Enter into the Law,” Legal Theory 6 (2000): 83–108. Others
retain some version of the thesis but argue, against Shapiro, that inclusive positivism is con-
sistent with the practical difference thesis, properly understood. See, e.g., Kenneth Einar
Himma, “Judicial Discretion and the Concept of Law,” Oxford Journal of Legal Studies 19
(1999): 71–82; W. J. Waluchow, “Authority and the Practical Difference Thesis,” Legal Theory
6 (2000): 45–81. Shapiro has responded imaginatively to many of these criticisms. Scott J.
Shapiro, “Law, Morality, and the Guidance of Conduct,” Legal Theory 6 (2000): 127–70.
Theoretical Implications 299

reason to deviate than to adhere. On this reading of Shapiro, selective

optimization entails that first-order legal rules and other legal standards do
not, in his sense, guide judges. However, I shall argue that selective optimi-
zation allows first-order legal rules to guide judges in a different way.
Shapiro believes that an agent is guided by law only if she obeys legally
pedigreed rules. This belief entails that judges who deviate from pedi-
greed rules are not guided by the law. This is correct if one accepts
Shapiro’s conception of rule guidance. As Shapiro defines rule guidance, a
rule guides an agent to Φ only if she might not have Φ-ed had she herself
not appealed to that rule. This is a familiar and straightforward way for
rules to influence agents. But it is not the only way. When we think of a
rule guiding an agent we naturally think of cases that are regulated by the
rule, as Meg’s decision to refrain from purchasing beer is regulated by the
statute. In regulated cases the rule guides by discounting certain reasons
that would otherwise apply. Unconditional discounting reasons are second-
order reasons to treat specified first-order reasons as having less force than
they otherwise would. Discounting reasons can themselves be discounted.
When an agent defers to a conflicting rule, she discounts the discounting
reasons supplied by the first rule, thereby allowing the first-order reasons
again to play their usual role.
Shapiro concentrates on cases of unconditional discounting. But rules
can also discount conditionally. A conditional discounting reason is a sec-
ond-order reason to accept a conditional according to which, if some
condition is met, then one will treat specified first-order reasons as having
less force than they otherwise would. The specified condition could
concern virtually anything, including the conformity of oneself or other
agents to other rules.
The following scenario illustrates all the previous points at once. It
shows how a rule can guide an agent, in a case not regulated by that rule,
by providing conditional discounting reasons, where the specified condi-
tion concerns the conformity of another agent to another rule. Cordelia
and Goneril are teenage sisters. Goneril, the elder sister, has an 11 P.M.
curfew, whereas Cordelia’s curfew is earlier. Goneril sits at home at 10:30
P.M. one night. She wants to attend a late-night concert that would require
her to violate her 11 P.M. curfew. She is considering leaving the house
when she notices Cordelia coming home. Goneril does not know whether
Cordelia has missed her curfew, however, because Goneril cannot remem-
ber if her parents have changed Cordelia’s curfew from 10 P.M. to 10:30
P.M. Goneril has special reason to care about Cordelia’s curfew because
Goneril is eagerly anticipating a family vacation and she knows from
experience that her parents will cancel the vacation if both sisters violate
their respective curfews, but not if only one of them does so. If Cordelia
has missed her curfew then Goneril must stay home, or lose the vacation.
If, however, Cordelia’s curfew has been moved to 10:30 P.M., then Gon-
eril has the option to attend the concert, thereby violating her own
curfew without jeopardizing the vacation.
300 Limits of Legality: Part II

Goneril’s own curfew remains 11 P.M., regardless of whether Cordelia’s

has changed. Goneril could decide to guide her conduct exclusively on
the basis of her own curfew rule and take no account of Cordelia’s rule or
behavior. We can analogize Goneril’s situation to that of a judge deliber-
ating in a suboptimal-result case. Much as Goneril could decide to consult
only her own curfew rule, so the judge could decide to consult only pedi-
greed legal standards. But Shapiro does not just claim that judges could
decide to consult only pedigreed legal standards. He denies that judges
who ignore such standards are guided by the law. Presumably Shapiro
would also deny that Cordelia’s rule guides Goneril, which is true in one
sense but false in another. Cordelia’s curfew rule influences Goneril’s de-
liberations in an obvious way. Although Goneril could decide to ignore
Cordelia’s rule and behavior, she could also decide to respond strategi-
cally to Cordelia. Suppose Goneril takes the strategic route. In that case
Cordelia’s curfew rule provides Goneril with a conditional discounting
reason, which is a second-order reason to accept the following condi-
tional: if Cordelia comes home later than the hour specified in Cordelia’s
curfew rule, then Goneril shall discount her reasons to attend the concert.
Cordelia’s rule addresses Cordelia directly, but it also addresses Goneril
indirectly, by providing conditional discounting reasons. If Goneril is try-
ing to obey the rules that apply to her, then as the content of Cordelia’s
rule fluctuates so does Goneril’s behavior in response. I take this to sug-
gest that Cordelia’s rule is guiding Goneril, albeit indirectly. So a rule can
indirectly guide an agent even in a case that is not regulated by that rule.
Shapiro, however, assumes that a rule cannot guide an agent unless it
provides her with unconditional discounting reasons. In one sense, this is
correct. If r is a legal rule, then there could, in principle, exist an agent for
whom r provides unconditional discounting reasons. However, Shapiro’s
argument depends on a stronger condition. His argument presupposes
that r cannot constitute a legal rule for a certain agent unless r provides
unconditional discounting reasons for that very agent. This is not so. Indeed,
no agent for whom r provides unconditional discounting reasons must
actually exist in order for r to constitute a legal rule for other agents, by
providing them with conditional discounting reasons.
This last point becomes clearer if we modify our scenario. Suppose
Cordelia and Goneril are twins, each with an 11 P.M. curfew. They encoun-
ter one another downtown at 10:45 P.M. Still assuming their parents’ vaca-
tion rule to be in effect, Cordelia’s rule provides Goneril with a conditional
reason and Goneril’s provides Cordelia with a conditional reason. Cord-
elia’s rule guides Goneril and Goneril’s rule guides Cordelia—indirectly
in both cases. Neither rule provides either teenager with an unconditional
discounting reason. Yet the curfew rules clearly play as important a role as
ever in their deliberation. The parents have effectively issued this com-
mand: “Girls, you mustn’t both break curfew!”
I suggest that what Shapiro describes as rule guidance is really a special
case of a more general phenomenon, generalized rule guidance:
Theoretical Implications 301

A legal rule r guides (in the generalized sense) a person p to Ф only if p

might not have done Ф if a person q had not appealed to r as a legal rule
(where q might or might not be the same person as p and q actually does
appeal to r as a legal rule).
In terms of my condition on generalized rule guidance, Shapiro describes
the special case in which q is the same person as p. I call this direct rule
guidance, in contrast to indirect rule guidance, where q and p are different
persons. It is true that Cordelia’s curfew rule cannot guide Goneril directly.
Shapiro’s position, again in my terms, is that a legal rule cannot make a
practical difference for a certain agent unless it guides her directly.
Someone might object that what guides Goneril is not Cordelia’s rule
simpliciter but rather the conjunction of Cordelia’s rule and the parents’
vacation-canceling rule. True, Cordelia’s rule guides Goneril only if the
vacation rule is also in force. But if the vacation rule is in force, then
Cordelia’s rule does, indeed, guide Goneril. The vacation rule simply makes
it the case that Cordelia’s rule guides Goneril. It does not supplant
Cordelia’s rule, but rather extends to Goneril that rule’s capacity to guide.
Alternatively, Shapiro could object that Cordelia’s rule does not guide
Goneril because Goneril does not appeal to it as an authoritative rule.
Indeed, Goneril does not “follow” Cordelia’s rule. If Cordelia’s curfew is
10 P.M., that fact will not motivate Goneril to get home by 10 P.M. Never-
theless, Goneril appeals to Cordelia’s rule as a rule that is directly
authoritative for Cordelia and indirectly authoritative for Goneril.
Compare Goneril’s attitude toward a rule announced by her little brother
(who has no authority over Cordelia): “Cordelia must return by 8 P.M.!”
Goneril does not appeal to this rule as authoritative, whereas she does
appeal to Cordelia’s rule as authoritative.
Now consider how my generalized conception of rule guidance
applies to the judiciary. Some exclusive positivists portray legal rules as
supplying judges with unconditional discounting reasons. These reasons,
moreover, are supposed to reference only case-type-specific facts. I con-
test both points. First, I suggest that we need not understand legal rules
as providing real judges with anything more than conditional discounting
reasons. Second, I suggest that the conditions can reference facts that are
not case-type-specific, such as the judge’s own deviation record, or even
(in principle) the proximity of the legal system to its threshold.
Shapiro believes that judges who deviate from pedigreed rules are not
guided by law. He might be correct that they are not guided directly and
unconditionally. But I disagree if he also means that judges who deviate
from pedigreed rules are not guided by the law in any way. Suppose
Goneril takes the strategic route. She decides to understand her parents
as having established a violation threshold: two curfew violations will
cancel the vacation. In that case, Goneril would understand Cordelia’s
rule as providing her with a special kind of conditional discounting reason,
one that references the violation threshold. Here is the parallel strategy in
the judicial case. According to selective optimization, valid legal rules
302 Limits of Legality: Part II

provide judges with conditional discounting reasons that reference the

judge’s own deviation rate. Much as Goneril has reason both to attend the
concert and to protect the vacation, so our judge has reason to deviate,
thus avoiding a suboptimal result in the case at bar without violating se-
lective optimization. On my view a valid legal rule gives judges reasons of
two kinds in cases regulated thereby: first, an unconditional reason to
discount private reasons to deviate; and second, a conditional reason to
discount moral reasons to deviate when required by selective optimiza-
tion. However, if selective optimization permits deviation in a certain
suboptimal-result case, then the suboptimality of the legally required
result discounts the reason to adhere that is provided by the legal rule.
This parallels the case in which Cordelia has not actually violated her
own curfew. In that case, given Cordelia’s conduct and her curfew rule,
the condition specified in the parents’ vacation rule is not met. Therefore,
Goneril’s own curfew rule gets discounted. Rather than providing Gon-
eril with a second-order reason to discount her first-order reasons to
attend the concert, Goneril’s curfew rule gives her, at most, an ordinary
first-order reason to stay home that must compete with her reasons to
attend the concert. I conclude that legal standards can indirectly guide
judges who deviate in suboptimal-result cases.
I agree with exclusive positivists that there is an important difference
between deviating from pedigreed rules and adhering to them. Deviation
contributes to an enterprise that could, in excess, weaken the rule of law.
Adherence does not. But exclusive positivists are incorrect if they also
believe that deviating from a particular legal standard must involve a
failure to be guided by the law as a whole. Legal standards can guide and
constrain the judiciary in more ways than that.

17.4 EQUITY

Readers familiar with the concept of equity in legal history will rightly
wonder if selective optimization should be understood as a form of equi-
table decision making. A comparison with various historical conceptions
of equity would be fascinating but must await another occasion.20 Here

20. Equity has multiple meanings in legal history that I shall not bother to differentiate.
See, e.g., John F. Manning, “Textualism and the Equity of the Statute,” Columbia Law Review
101 (2001): 1–127; John R. Kroger, “Supreme Court Equity, 1789–1835, and the History of
American Judging,” Houston Law Review 34 (1998): 1425–86; John J. Farley, III, “Robin
Hood Jurisprudence: The Triumph of Equity in American Tort Law,” St. John’s Law Review
65 (1991): 997–1021 (conceding that equitable notions still exist in American tort law);
Peter Charles Hoffer, “Principled Discretion: Concealment, Conscience, and Chancellors,”
Yale Journal of Law and the Humanities 3 (1991): 53–82 (analysis of American equity juris-
diction); Calvin Woodard, “Joseph Story and American Equity,” Washington and Lee Law
Review 45 (1988): 623–44; Gary McDowell, Equity and the Constitution: The Supreme Court,
Equitable Relief, and Public Policy (Chicago: University of Chicago Press, 1982); Walter
Wheeler Cook, Cases and Other Authorities on Equity (St. Paul: West, 1923).
Theoretical Implications 303

I limit myself to a prominent contemporary treatment. Lawrence Solum

defines equity as “the practice of doing particularized justice, when the
just result is not required by, or is contrary to, the result required by the
set of applicable legal rules.”21 Solum’s equity differs from all moderate
rules, including selective optimization, in several respects. First, Solum’s
equity applies only to negative-gap cases:
Equity tailors the law to the requirements of the particular case. Under-
standing equity as a particularized practice allows us to distinguish it from
other practices that involve a departure from legal rules. For example, equity
is not identical to the resolution of conflicts between law and morality in
favor of the latter. Judges might nullify a statute that legalized the practice
of slavery on the ground that slavery is always morally wrong. This is not an
example of the practice of equity, because such a decision would not involve
a departure from the rule on the basis of the facts of the particular case.
Rather the decision would be based on a general moral principle—for
example, that wicked or immoral statutes should not be enforced.22

By contrast, moderate rules apply to all suboptimal-result cases, includ-

ing both negative-gap cases and suboptimal-rule cases.
Second, Solum’s equity involves doing particularized justice. Solum
permits deviation from legal rules only when necessary to avoid unjust
results, whereas moderate rules permit deviation to avoid the broader
category of suboptimal results. All unjust results are suboptimal, but not
all suboptimal results are unjust.
Third, moderate rules allow deviation from the law simpliciter, whereas
Solum’s equity, as described, allows deviation only from legal rules. It is
not clear whether Solum also allows deviation from the law simpliciter.
Perhaps he regards equitable decisions as legally correct and believes that
deviation from the law simpliciter is never necessary or justified.23
Despite these differences between moderate rules and Solum’s equity,
Solum’s view remains a relative of selective optimization and serves as an
instructive foil for it. Selective optimization can even be seen as a general-
ization of Solum’s equity with some added safeguards and qualifications.
In addition to these generic contrasts between Solum’s equity and any
moderate rule, there are two additional differences between Solum’s equity
and selective optimization, specifically. First, Solum appears to accept a
strong comparative justice principle that rules out token-selective deviation.

21. Lawrence B. Solum, “Equity and the Rule of Law,” in The Rule of Law, ed. Ian
Shapiro (New York: New York University Press, 1994), p. 123. See also Gerald J. Postema,
Bentham and the Common Law Tradition (Oxford: Oxford University Press, 1986), pp. 200,
407–8, 416–17, 429.
22. Solum, “Equity and the Rule of Law,” p. 124.
23. The idea that equitable deviation from legal rules does not constitute deviation
from the law, simpliciter, is reinforced by assertions such as this: “An adjudicator with judicial
integrity cares about the coherence of the law and is motivated to ensure that her departure
from the letter of the rules accords with their spirit.” Ibid., p. 136 (emphasis added).
304 Limits of Legality: Part II

Although he notes that equity can involve treating differently two case
tokens that the legal rules classify as identical,24 thereby infringing legal
comparative justice, the differential treatment must be itself justified by
differences between the factual case types. The “particulars” of the case to
which Solum demands judicial sensitivity are particulars of the moral case
type.25 Cases apt for equitable treatment are those in which the particular
fact pattern was not anticipated by the lawmaker, contracting party, testator,
et cetera. For Solum, if the particular facts entail that the result required by
a generally good legal rule is unjust, then the judge should deviate. Other-
wise, she should adhere. These facts are both necessary and sufficient to
warrant deviation, so there is no possibility of deviation being appropriate in
one case token but inappropriate in another token of the same type. Selec-
tive optimization could, in principle, permit this.
Solum’s theory is indifferent to systemic factors, including deviation
density, which brings me to the second contrast with selective optimiza-
tion. Chapter 11 demonstrates that indifference to deviation density is
acceptable only if suboptimal-result cases are very rare.26 This is true in
modern legal systems only on certain conceptions of justice—those
according to which it is very rare for the good laws of such systems to
dictate unjust results. Perhaps Solum accepts such a conception of justice.
In that case his position would be coherent, but his theory of equity would
then depend upon his conception of justice. I am trying to construct a
theory of permissible deviation that is compatible with a wider range of
conceptions of justice—those that classify many cases in modern legal
systems as suboptimal-result cases, as well as those that place fewer cases
in that category. Therefore, selective optimization conditions the permis-
sibility of deviation on the proximity of the system to threshold. Solum’s
theory has no counterpart to this condition.

17.5 LEGAL PRAGMATISM

Profitable as it would be to compare selective optimization with various

incarnations of legal pragmatism, a brief note will have to suffice. Judge
Richard Posner, the preeminent legal pragmatist of our time, believes that
when judges have the authority to formulate legal rules they should craft
rules that they predict will produce the best consequences. They should
use the best available social science for that purpose, as well as common
sense, and they should explain their use of these sources in their opinions.
Departing from bad horizontal precedent is permissible, Posner thinks,
although judges should give it some weight, recognizing the values of

24. Ibid., p. 125.

25. “A judge doing equity will give reasons for his decision that are based on the facts of
the particular dispute. . . .” Ibid., p. 126.
26. See §11.9.
Theoretical Implications 305

stability, predictability, and reliance. He appears to believe that judges

have the legal and moral authority to choose optimal rules more often
than mainstream legal scholars think they have it.27 So Posner and the se-
lective optimizer both believe that judges are morally permitted to reach
the optimal result more often than some scholars consider permissible.
None of Posner’s prescriptions seems to conflict with selective optimi-
zation. However, the two positions address different types of cases. In the
cases that interest Posner, judges have legal authority to choose or modify
rules. Posner expresses opinions about how they should decide such cases,
what sources they should use, and how they should justify their choices
in print. By contrast, I have considered cases in which binding legal stan-
dards apply: the text is unambiguous or vertical precedent controls. Judges
have no legal authority to deviate in such cases. I cannot tell if Posner
believes that judges must always adhere in such cases. But I think he could
consistently endorse restrictive rule without retracting his forward-looking
criteria for the best rules, his recommendation that judges write empirical
evidence into their opinions, or his belief that they should often reverse
horizontal precedents with bad consequences. To this extent, Posner’s
pragmatism and selective optimization complement each other, but their
distinctive features do not substantially overlap.

17.6 POLITICAL NEUTRALITY

Finally, I return to the issue of political neutrality in adjudication theory,

first raised in chapter 1.28 Commentators of all political persuasions com-
plain about judges who ignore the law, but American conservatives have
been more likely than liberals to make such complaints, at least for the
past few decades. The correlation has loosened somewhat as the federal
bench has become more conservative. Of course, most liberals do not
openly advocate ignoring the law, but it remains reasonable to infer that a
writer in the early twenty-first century who advocates ignoring the law is
probably not conservative. I do not claim that political conservatives
should welcome my conclusions in this book. No interesting prescription
concerning adjudication will, at a particular moment in history, be equally
appealing to everyone in complex pluralistic societies. If political neutrality
requires pleasing everyone, then adjudication theory should not aspire to
such neutrality.
There is, however, another sort of political neutrality toward which we
should aspire. It is rational for conservatives to assume, upon encountering
an argument for deviation, not just that its author is liberal, but that it

27. Richard A. Posner, The Problematics of Moral and Legal Theory (Cambridge, Mass.:
Belknap, 1999), pp. 240–52.
28. See §1.8.
306 Limits of Legality: Part II

rests upon liberal premises. Close scrutiny by clever conservatives usually

reveals these premises, depriving the argument of any persuasive force for
conservative readers who reject those assumptions. Some writers on
adjudication acknowledge that their arguments proceed from politically
controversial premises. They assume, for example, that their reader is a
fellow liberal, and they advocate a particular theory of adjudication that
enables them to defend various rulings and legal arguments favored by
liberals. I do not oppose this kind of scholarship. I would be happy to
discover a coherent theory of adjudication that enabled judges to imple-
ment all of my political views, while preventing them from implementing
anyone else’s. I have not found such a theory, nor did I formulate selective
optimization with that goal in mind. Instead, I have tried to avoid any
premises that are closely identified with the political left or right. Part I
uses widely accepted normative premises with no distinctive political
associations. Part II uses premises that remain controversial in academic
philosophy, but they are not distinctively creatures of the political right or
left, either. I hope I have offered arguments for selective optimization that
people with a wide range of political opinions can accept.
Nevertheless, I must make a confession: I believe that legalizing contra-
ception was morally correct and that Justice Douglas acted permissibly in
Griswold, even if there was no sound legal argument for his ruling, as
many lawyers believe. Selective optimization is consistent with these
beliefs, but not because my arguments for it depend on my political con-
victions. Selective optimization authorizes judges to deviate only if the
legally required result is actually suboptimal. If I am wrong about the
injustice of birth control bans and no sound legal argument existed for
the Griswold holding, then selective optimization entails that Justice
Douglas acted impermissibly. Selective optimization cannot guarantee
that anyone is, in fact, morally enlightened. It merely assures me that
morally enlightened judges are sometimes morally permitted to apply
their wisdom, even when the law disagrees. It gives precisely the same
assurance to those who reject my politics. In fact, I offer an uncomfort-
able prediction: if judges in the United States today were faithfully to
obey selective optimization, then they would reach fewer results that I
favor as a political citizen than if they were to obey restrictive rule. This
prediction disappoints me as a political citizen, but it strangely reassures
me as a theorist. It provides some circumstantial evidence that my argu-
ments do not presuppose my idiosyncratic opinions about justice and
public policy.
At the same time, if selective optimization is correct, then we must
reexamine popular attacks on lawless judging. Commentators often state
that a court has “misread,” “misstated,” “misinterpreted,” or “ignored” the
law. The literal content of such statements is theoretical, not practical: the
judge has made a legally incorrect assertion. A professor could make
exactly the same theoretical criticism of a legal assertion in a paper sub-
mitted by a law student. However, when a commentator makes such a
Theoretical Implications 307

statement about a judicial decision, he usually intends to imply much

more than a theoretical criticism. He intends to imply practical claims
such as one or more of the following conversational implicatures:29
1. The judge had an all-things-considered reason to decide differently.
2. If the judge knowingly did what the commentator claims she did
(i.e., deviate from the law), then she is blameworthy.
3. The judge acted unethically, at least if she deviated knowingly,
and perhaps even if she deviated unknowingly.
If selective optimization is true, then such implicatures are usually
false. If judges have an all-things-considered duty to obey restrictive rule,
then it is not especially important to distinguish between the theoretical
claim that a judge has deviated and practical claims such as the three
above. If restrictive rule is correct, then the theoretical claim implies at
least some of these practical claims. However, if selective optimization (or
any moderate or permissive rule) is correct, then we must carefully distin-
guish between the theoretical claim and the practical claims because the
former does not imply the latter. Selective optimization entails that if a
judge deviates in more than a certain fraction of his cases, then he creates
an impermissible decision pattern. But selective optimization also entails
that very few, if any, of these decisions is morally impermissible in itself if
it avoids a suboptimal result.

29. For an introduction to conversational implicature see Paul Grice, “Logic and Con-
versation,” in Studies in the Way of Words (Cambridge, Mass.: Harvard University Press,
1989).
18

Conclusion

I have called into question three adjudication principles that many knowl-
edgeable people accept, uncritically:
1. Judges in reasonably just legal systems are morally obligated to
apply the law correctly in all cases in which the law requires a
certain result or range of results.
2. Judges are morally permitted to deviate from the law only, if ever,
when the legally required result is extremely unjust.
3. Only the fact pattern of the case is relevant to the moral permissi-
bility of deviation from the law.
Popular as these principles are, they are rarely defended. I considered and
rejected various arguments for them. Whereas most discussions of judicial
deviation concentrate on its political legitimacy, I asked whether it is rea-
sonable for society to ask judges, as autonomous moral agents, to refrain
from deviation in suboptimal-result cases as often as the conventional
wisdom demands. I considered not just the legitimacy of the judicial role,
but the moral rights of the individuals who occupy it. I shall now review
my lengthy argument and offer some concluding reflections about the
implications of selective optimization.
My investigation began with an account of judicial authority based on
the natural rights and duties of judges. In a state of nature, individuals
have duties of nonmaleficence, samaritan rights, and rights of justice.
In civil society they retain their duties of nonmaleficence and their
samaritan rights, but their rights of justice are partially undermined as the
state largely takes over the function of pursuing justice. Judges are agents
of the state. When they decide cases they create, withdraw, or block
threats of force. They have legal authority to do so. They are also morally
permitted to use force in certain situations in which the law authorizes
them to use it.
Some people appear to accept the undermining principle, which states
that if the law requires a public official to use force in a given situation,
then he has no moral reason not to use it. The undermining principle has
implications for adjudication in suboptimal-result cases: those in which
the law, properly understood, requires a result that the judge would have
an all-things-considered reason to avoid if the law permitted her to avoid

308
Conclusion 309

it. If the undermining principle is true, then judges never have moral
reasons to deviate from the law. I considered many arguments for the
undermining principle, including arguments from legal positivism, ordi-
nary discourse, role morality, formal legality, political legitimacy, legal
authority, intention, and means. I found none of these arguments to be
sound. In the absence of a sound argument for the undermining principle,
I concluded that judges have pro tanto moral reasons to deviate from the
law in suboptimal-result cases.
I then turned to the topic of adherence rules, which require judges to
adhere to the law. I distinguished between restrictive, moderate, and
permissive rules. I noted the centrality of adherence rules to the rule of
law and considered various reasons for lawmakers to promulgate restrictive
rule. These included reasons of settlement, predictability, coordination,
reliance, efficiency, error, stability, legitimacy, autonomy, respect, and
fairness. I concluded that most of these are good reasons to promulgate
restrictive rule.
My next question was whether judges have reasons to obey adherence
rules. After concluding that judges have excellent pro tanto moral reasons
to obey at least permissive rule, I asked whether they have moral reasons
to obey restrictive rule. I found that the arguments for promulgating
restrictive rule do not support obeying it in suboptimal-result cases.
Neither do consequentialist or contractualist arguments. In fact, these
arguments fail to support a moral duty to obey anything stronger than
permissive rule. Chapter 9 evaluated and rejected the claim that judges
must adhere to the law in suboptimal-result cases because they swore a
solemn oath of office to uphold the law. Chapter 10 evaluated and rejected
the claim that judges have a moral duty to adhere because they are legally
obligated to do so. I considered how several familiar arguments for a duty
to obey the law apply to the issue of judicial obligation: arguments from
consent, fair play, natural duty, gratitude, and samaritanism. I also argued
that judges are morally permitted to decide suboptimal-result cases rather
than self-recusing.
Part I reached three main conclusions. First, we have no reason to
accept the undermining principle. Second, lawmakers have good reasons,
perhaps all-things-considered reasons, to promulgate restrictive rule.
Third, the obvious arguments for the claim that judges have moral reasons
to obey nonpermissive rules are fallacious. If that were the end of the
story, then we would be left with the unpopular conclusion that judges
have no moral reason to obey anything but permissive rule.
Part II described the normative structure that one must, I think, endorse
if one wishes to maintain that judges have a moral reason to obey a non-
permissive rule. The needed normative structure is intricate and contains
some philosophically controversial elements. I did not thoroughly defend
all of these elements. I simply showed why they are necessary and how
they fit together. I invite the reader to investigate the literatures defending
these elements and to decide if she is prepared to accept them as the price
310 Limits of Legality: Part II

of believing that judges have an all-things-considered moral reason to

obey a nonpermissive rule.
The argument of part II proceeds from the premise that the judges
who possess good moral judgment constitute a group—Group O—the
members of which share two collective intentions: to minimize suboptimal
results throughout their legal system and to avoid reaching suboptimal
results themselves. They can fulfill the second intention by deviating from
the law in suboptimal-result cases, but a pattern of deviating from the law,
even in suboptimal-result cases, causes mimetic failure—other judges will
imitate Group O and deviate in optimal-result cases, thereby reaching
suboptimal results. At some point the rate of deviation by Group O could
encourage so much deviation by other judges that the suboptimal results
reached by those judges would outweigh the suboptimal results avoided
by Group O. That point is what I call the deviation density threshold.
However, as I discussed in chapter 12, there are four agent-relative
principles that, if true, entail that mimetic failure does not give Group O
any reason to adhere to the law in suboptimal-result cases, even if
additional deviation will transgress threshold. The principles that entail
this result are ones that assign moral significance to intention, means,
proximity, and agential mediation. I argued that we must abandon, qualify,
or demote all of these principles if we wish to support the principle that
judges have moral reasons to obey a nonpermissive rule. Adjudication
theory should be foundationally agent-neutral.
The next question was whether the members of Group O have
individual reasons to contribute to the group’s efforts by adhering in at
least some suboptimal-result cases. I argued that if enough of them are
actually contributing, then they all have moral reasons to contribute in
order to avoid riding free on one another. If too few of them are contrib-
uting, however, then the case for contributing becomes more difficult to
make. I suggested three possible moral principles, any one of which would
support contributing under conditions of general defection. The first
principle holds individuals responsible for the imperceptible effects of
their actions. The second holds them responsible for subjecting others to
minimal, but unjustified, risks of harm. The third holds them responsible
for participating in harmful or wrongful actions even when they cause no
harm and otherwise do no wrong. Each of these principles is controversial.
But if one or more of them is true, then we can argue that judges have
reasons to adhere in at least some suboptimal-result cases, even when
most other judges are deviating too frequently. This discussion culminated
in the defense of individual policy, which specifies permissible deviation
rates for judges based in part on how often other judges deviate.
In chapters 14, 15, and 16, I asked how judges might actually imple-
ment the rules defended in the previous chapters of part II (i.e., combined
restriction). The discussion shifted from objective appraisal rules to
subjective guidance rules. I responded to Alan Goldman’s objections to
rules that permit judges sometimes to deviate in suboptimal-result cases.
Conclusion 311

I defended guidance rules that permit each judge to deviate in a certain

percentage of the suboptimal-result cases that she decides over the course
of her career. I defended two priority rules for judges. The first rule assigns
deviation priority to cases based on how suboptimal the legally required
results are. The second rule assigns priority to suboptimal-rule cases over
gap cases. I also defended a default rule that permits judges, after they
obey the priority rules, to use any morally permissible criteria as the basis
for their selection of suboptimal-result cases for deviation. I made the
surprising claim that even random selection is not ruled out as a matter of
principle. I defended this claim against arguments from comparative
justice and Dworkinian integrity. The emerging theory is selective
optimization.
In chapter 16, I examined some basic concerns about implementing
selective optimization. I considered some familiar court cases and showed
how selective optimization allows one coherently to make a claim that
sounds paradoxical: that the judges were justified, all things considered, in
reaching results that could not be defended as legally correct. I argued
that judges are often in such a position, even in reasonably just legal
systems.
In chapter 17, I examined more restrictive theories of rule guidance,
including those of some exclusive positivists and legal formalists. I showed
that their arguments do not impugn selective optimization. I also briefly
contrasted selective optimization with pragmatism and equity.
I now offer some concluding reflections. Much judicial rhetoric is legal-
istic: it expresses the principle that judges have an all-things-considered
moral obligation to obey restrictive rule while they remain on the bench.
When a judge says “the law requires result x,” you can be sure that he is
going to reach x. However, “the law requires result x” can express at least
two distinct propositions. One is that reaching a result inconsistent with
x would violate the conventions of legal argument. Another is that judges
have an ethical duty to reach x. When someone asserts “the law requires
result x,” without qualification, he usually means to assert both proposi-
tions and it is reasonable to infer as much. The absence of this distinction
in legal discourse is understandable, given the traditional commitment to
restrictive rule. If judges must obey restrictive rule, then the distinction
serves no practical purpose. But equating “what the law requires” with
“what judges must do” makes it impossible even to state selective
optimization.
I do not know how many judges actually believe that they are morally
obligated to obey restrictive rule, but I assume that many of them do.
A morally conscientious judge who sincerely believes this will make one
of two opposing mistakes. First, her desire to avoid suboptimal results
may turn her into a kind of Panglossian. She desperately wants to believe
that the law permits every result that she considers optimal and she is
prepared to warp her theory of law in whatever direction enables her to
embrace that harmonious conclusion. She distorts her perception of the
312 Limits of Legality: Part II

law until it no longer functions in her mind as a set of standards that can
dictate determinate results. Alternatively, if she avoids the Panglossian
trap, then her desire to be moral will lead her to reach an unnecessarily
large number of suboptimal results.
In realistic legal systems it is impossible to guarantee, a priori, that a
judge will never hear a case in which he has an all-things-considered moral
obligation to reach a suboptimal result. A judge in such a case faces an
uncomfortable dilemma (ignoring the recusal option): he must either
reach a suboptimal result or deviate. The only way to preclude such
dilemmas, a priori, is to loosen the conventions of legal argument to such
a degree that the law always permits reaching optimal results. That would
sacrifice the advantages of a legal system over a system of unguided judicial
discretion. Judges are fated to face some of these dilemmas even if my
defense of selective optimization is successful. However, they face more
such dilemmas if they obey restrictive rule than if they selectively
optimize.
The rhetoric of lawyers is legalistic, too. An essential part of a litigator’s
job is to persuade the judge that the law at least permits a ruling in favor
of her client. Even when a litigator presents arguments from justice or
policy, she presents them as arguments about what the law says or should
say. She encourages the judge to understand existing law in a certain way
or to develop the law in a certain direction. She knows that judges some-
times deviate and she will encourage deviation if the law disfavors her
client, but she will do so by persuading the judge to believe, or at least to
assert, that the law permits what it actually forbids. Urging a judge to
deviate in so many words is a rare act of desperation. If judges actually
have an all-things-considered moral obligation to obey restrictive rule,
then encouraging them—whether explicitly or implicitly—to deviate is
probably unethical, too, even in suboptimal-result cases. By contrast, if
selective optimization is permissible, then encouraging judges to deviate
in suboptimal-result cases is not unethical.
Selective optimization also has implications for critics of lawless judging.
It entails that, in realistic legal systems, the mere fact that a decision is
deviant never provides a reason to conclude that the judge has acted im-
morally. The result may be suboptimal, but that is a separate matter. The
only way to determine whether a judge has behaved unethically in virtue
of having deviated is to assess her entire record for a pattern of excessive
deviation as defined by selective optimization. The fact that a judge has
deviated is not, in itself, a reason for censure, impeachment, removal, or
any form of retaliation against her. It is not even a reason to vote against a
judge when she runs for reelection, or a reason for a senator to vote against
confirming a presidential nominee to the federal bench. Only a pattern of
excessive deviation provides such reasons. Nor should judges generally
face consequences for reaching suboptimal results that are required by
law. Whether judges should face consequences for reaching suboptimal
results not required by law is a larger issue that I cannot address here.
Conclusion 313

Of course, avoiding criticism and disciplinary action is an important

incentive for many judges. If everyone stopped criticizing deviation, then
judges would deviate more often, perhaps eventually crossing threshold.
Such criticism should continue, in order to guard against that possibility.
However, critics of the bench should heed the distinction between the
proposition that a judge has deviated and the judgment that he has acted
wrongly. They should recognize that calling a decision “lawless” is generally
understood to express both. Until the linguistic conventions of legal
discourse change, commentators should seek creative ways to convey the
difference between these propositions and to avoid being misunderstood.
At least, they should draw the distinction in their own minds. Perhaps
they have been privately drawing it all along and I am not telling them
anything new. If they have been drawing this distinction, however, then
they have been hiding it very well.
Although selective optimization challenges the ideal of judicial adher-
ence to the law, it is not skeptical or nihilistic about legal discourse. It is
entirely compatible with the claim that some legal questions, or even all
of them, have legally correct answers, although it does not entail this
claim. Selective optimization does not call into question the objectivity of
legal reasoning or the autonomy of legal reasoning from other kinds of
discourse. Neither does it affirm legal objectivity and autonomy. It provides
a principled alternative to the largely discredited, radical indeterminacy
claims commonly ascribed to critical legal studies. Whereas radical inde-
terminacy implies that legal standards never preclude reaching morally
optimal results, selective optimization avoids this indefensible claim.
Selective optimization entails, nevertheless, that evaluating the work of
judges involves more than asking whether their legal reasoning is correct,
however broadly we understand legal reasoning. It provides an ethical
justification for reaching the morally optimal result, even when the law
precludes it, in at least some cases. The rule of law cannot survive if judges
deviate from the law too frequently, thereby violating selective optimiza-
tion. However, traditional conceptions of judicial duty and the rule of law
constrain adjudication to an extent that cannot be easily justified.
This page intentionally left blank
List of Authorities

Aaron, California Commission on Judicial Performance (July 8, 2002)

Albrecht v. Herald Co., 390 U.S. 145 (1968)
Alden, United States v., 141 Fed. Appx. 562 (9th Cir. 2005)
Ashcroft v. Free Speech Coalition, 535 U.S. 234 (2002)
Barnes v. Glen Theatre, Inc., 501 U.S. 560 (1991)
Bayless, United States v., 913 F. Supp. 232 (S.D.N.Y. 1996)
Bayless, United States v., 921 F. Supp. 211 (S.D.N.Y. 1996)
Bennis v. Michigan, 116 S. Ct. 994 (1996)
Bethlehem Steel Co. v. New York State Labor Relations Bd., 330 U.S. 767
(1947)
Bittaker v. Enomoto, 587 F.2d 400 (9th Cir. 1978)
Boumediene v. United States, 128 S. Ct. 2229 (2008)
Bowers v. Hardwick, 478 U.S. 186 (1986)
Boy Scouts of America v. Dale, 530 U.S. 640 (2000)
Bracy v. Gramley, 520 U.S. 899 (1997)
Bradley v. Fisher, 80 U.S. 335 (1872)
Bradley v. Richmond School Board, 416 U.S. 696 (1974)
Brignoni-Ponce, United States v., 422 U.S. 873 (1975)
Broadman v. Commission on Judicial Performance, 959 P.2d 715 (Cal. 1998)
Brown v. Board of Education, 347 U.S. 483 (1954)
Bush v. Gore, 531 U.S. 98 (2000)
Callender, United States v., 25 F. Cas. 239 (C.C.D. Va. 1800)
Chickasaw Nation v. United States, 534 U.S. 84 (2001)
Connecticut National Bank v. Germain, 503 U.S. 249 (1992)
District of Columbia v. Heller, 128 S. Ct. 2783 (2008)
Dred Scott v. Sandford, 60 U.S. 393 (1857)
Ekwunoh, United States v., 888 F. Supp. 369 (E.D.N.Y. 1994)
Elk Grove Unified School District v. Newdow, 542 U.S. 1 (2004)
Everson v. Board of Education, 330 U.S. 1 (1947)
Florida v. Rodriguez, 469 U.S. 1 (1984)
Furman v. Georgia, 408 U.S. 238 (1972)
Glassroth v. Moore, 335 F.3d 1282 (11th Cir. 2003)
Gonzales v. Carhart, 550 U.S. 124 (2007)
Gonzales v. Raich, 545 U.S. 1 (2005)
Griswold v. Connecticut, 381 U.S. 479 (1965)
Harrod v. Illinois Courts Commission, 372 N.E.2d 53 (Ill. 1977)
Henchey v. City of Chicago, 41 Ill. 136 (1866)
Henningsen v. Bloomfield Motors, 32 N.J. 358 (1960)
Hutto v. Davis, 454 U.S. 370 (1982)

315
316 List of Authorities

Illinois v. Wardlow, 528 U.S. 119 (2000)

In Re Benoit, 487 A.2d 1158 (Me. 1985)
In Re Brown, 662 N.W.2d 773 (Mich. 2003)
In Re Cargill, 66 F.3d 1256 (1st Cir. 1995)
In Re Cooks, 694 So.2d 892 (La. 1997)
In Re Curda, 49 P.3d 255 (Alaska 2002)
In Re Daniels, 340 So.2d 301 (La. 1976)
In Re Derose, 1980 Annual Report 181 (New York State Commission on Judicial
Conduct, Nov. 13, 1979)
In Re Duckman, 699 N.E.2d 872 (N.Y. 1998)
In Re Friess, 1984 Annual Report 84 (New York State Commission on Judicial
Conduct, Mar. 30, 1983)
In Re Friess, 1984 Annual Report 84 (New York State Commission on Judicial
Conduct, Mar. 30, 1983)
In Re King, 568 N.E.2d 588 (Mass. 1991)
In Re Labelle, 591 N.E.2d 1156 (N.Y. 1992)
In Re Mattera, 168 A.2d 38 (N.J. 1961)
In Re Mckinney, 478 S.E.2d 51 (S.C. 1996)
In Re Quirk, 705 So.2d 172 (La. 1997)
In Re Spencer, 798 N.E.2d 175 (Ind. 2003)
Inquiry Concerning a Judge No. 52 (Hampton) (Texas Commission on Judicial
Conduct, 1989)
Isgro, United States v., 974 F.2d 1091 (9th Cir. 1992)
Jackson v. Lykes Bros. S.S. Co., No. 575, 386 U.S. 731 (1967)
Kelo v. City of New London, 545 U.S. 469 (2005)
Khan v. State Oil Co., 93 F.3d 1358 (7th Cir. 1996)
Kloepfer v. Commission on Judicial Conduct, 782 P.2d 239 (Cal. 1989)
Lawrence v. Texas, 539 U.S. 558 (2003)
Lewis v. Green, 629 F. Supp. 546 (DC Dist. Col. 1986)
Lockyer v. Andrade, 538 U.S. 63 (2003)
Lopez, United States v., 514 U.S. 549 (1995)
Marbury v. Madison, 5 U.S. 137 (1803)
Maryland St. Dept. Of Educ., Div. Of Rehabilitation Servs. v. United States Dept.
Of Veterans Affairs, 98 F.3d 165 (4th Cir. 1997)
McCartney v. Commission on Judicial Quality, 526 P.2d 268 (Cal. 1974)
McCulloch v. Maryland, 17 U.S. 316 (1819)
McCullough v. Commission on Judicial Performance, 776 P.2d 259 (Cal. 1989)
Mirsky, United States v., 17 F.2d 275 (S.D.N.Y. 1926)
Model Code of Judicial Conduct (1990, 2007)
Morrison, United States v., 529 U.S. 598 (2000)
Morrow v. Hood Communications, Inc., 69 Cal. Rptr. 2d 489 (Cal. Ct. App.
1997)
Mosley, United States v., 965 F.2d 906 (10th Cir. 1992)
Murtagh v. Maglio, 9 A.D.2d 515 (N.Y. 1960)
Neary v. Regents of the University of California, 834 P.2d 119 (Cal. 1992)
Newdow v. U.S. Congress, 292 F.2d 597 (9th Cir. 2002)
Newdow v. U.S. Congress, 328 F.3d 466 (9th Cir. 2003)
Oberholzer v. Commission on Judicial Performance, 975 P.2d 663 (Cal. 1999)
Oregon Waste Systems, Inc. v. Department of Environmental Quality, 511 U.S. 93
(1994)
List of Authorities 317

Orjuela, United States v., 809 F. Supp 193 (E.D.N.Y. 1992)

Payne v. Tennessee, 501 U.S. 808 (1991)
People ex rel. Harrod v. Illinois Courts Commission, 372 N.E.2d 53 (Ill. 1977)
People v. Zaring, 10 Cal. Rptr. 2d 263 (5th Dist. 1992)
Phototron Corp. v. Eastman Kodak Co., 687 F. Supp. 1061 (N.D. Tex. 1988)
Planned Parenthood v. Casey, 505 U.S. 833 (1992)
Prima Paint Corp. v. Flood & Conklin Mfg. Co., 388 U.S. 395 (1967)
R.A.V. v. City of St. Paul, 505 U.S. 377 (1992)
Restatement First of Contracts (1932)
Riggs v. Palmer, 115 N.Y. 506 (1889)
Rogers v. Tennessee, 532 U.S. 451 (2001)
Romer v. Evans, 517 U.S. 620 (1996)
Roper v. Simmons, 543 U.S. 551 (2005)
San Antonio v. Rodriguez, 411 U.S. 1 (1973)
Santa Fe Independent School District v. Doe, 530 U.S. 290 (2000)
Shonubi, United States v., 895 F. Supp. 460 (E.D.N.Y. 1995)
Sokolow, United States v., 490 U.S. 1 (1989)
Sparf v. United States, 156 U.S. 51 (1895)
State Oil Co. v. Khan, 118 S. Ct. 275 (1997)
States v. Schooner Peggy, 1 Cranch 103 (1801)
Stump v. Sparkman, 435 U.S. 349 (1978)
Troen, State v., 786 P.2d 751 (Or. Ct. App. 1990)
Tropiano, United States v., 898 F. Supp. 90 (E.D.N.Y. 1995)
Turco, Stipulation (Washington Commission on Judicial Conduct, Oct. 2, 1992)
Weber v. Kaiser Aluminum & Chemical Corp., 611 F.2d 132 (5th Cir. 1980)
Williams, United States v., 504 U.S. 36 (1992)
Wilson v. State, 279 Ga. App. 459 (Ga. Ct. App. 2006)
This page intentionally left blank
Bibliography

Abramson, Leslie W. “Appearance of Impropriety: Deciding When a Judge’s

Impartiality ‘Might Reasonably Be Questioned.’” Georgetown Journal of Legal
Ethics. 14 (2000): 55–102.
Ackerman, Bruce. “Anatomy of a Constitutional Coup.” London Review of Books,
Feb. 8, 2001.
———. We the People: Foundations. Cambridge, Mass.: Belknap, 1991.
Alexander, Larry. “Constrained by Precedent.” Southern California Law Review 63
(1989): 1–64.
———. “Deontology at the Threshold.” San Diego Law Review 37 (2000): 893–
912.
———. “The Gap.” Harvard Journal of Law and Public Policy 14 (1991): 695–
701.
———. “Law and Exclusionary Reasons.” Philosophical Topics 18 (1990): 5–22.
———. “‘With Me, It’s All Er Nuthin’: Formalism in Law and Morality.” University
of Chicago Law Review 66 (1999): 530–65.
Alexander, Larry, and Kimberly Kessler Ferzan. Crime and Culpability: A Theory of
Criminal Law. Cambridge: Cambridge University Press, 2009.
Alexander, Larry, and Ken Kress. “Against Legal Principles.” Law and Interpretation:
Essays in Legal Philosophy. Ed. Andrei Marmor. Oxford: Clarendon, 1995. 279–
327.
Alexander, Larry, and Emily Sherwin. The Rule of Rules. Durham, N.C.: Duke
University Press, 2001.
Alexy, Robert. “A Defence of Radbruch’s Formula.” Recrafting the Rule of Law: The
Limits of Legal Order. Ed. David Dyzenhaus. Oxford: Hart, 1999.
Alfini, James J., et al. Judicial Conduct and Ethics. 4th ed. Newark, N.J.: LexisNexis,
2007.
Almeida, Michael J., ed. Imperceptible Harms and Benefits. Dordrecht: Kluwer, 2000.
Alstyne, William Van. “Notes on a Bicentennial Constitution: Part II, Antinomial
Choices and the Role of the Supreme Court.” Iowa Law Review 72 (1987):
1281–99.
Altman, Andrew. Critical Legal Studies: A Liberal Critique. Princeton, N.J.:
Princeton University Press, 1990.
———. “Legal Realism, Critical Legal Studies, and Dworkin.” Philosophy and
Public Affairs 15 (1986): 205–35.
Altman, Scott. “Beyond Candor.” Michigan Law Review 89 (1990): 296–351.
Anderson, Lisa R., and Charles A. Holt. “Information Cascades in the Laboratory.”
American Economic Review 87 (1997): 847–62.
Applbaum, Arthur Isak. Ethics for Adversaries: The Morality of Roles in Public and
Professional Life. Princeton, N.J.: Princeton University Press, 1999.

319
320 Bibliography

Aquinas, Thomas. Summa Theologiae. Trans. Fathers of the English Dominican

Province. London: Washbourne, 1915.
Aristotle. Nicomachean Ethics. Trans. W. D. Ross. Oxford: Oxford University Press,
1980.
Atiyah, P. S. Promises, Morals, and Law. Oxford: Clarendon, 1981.
Austin, J. L. “A Plea for Excuses.” Proceedings of the Aristotelian Society 57 (1956–
57): 1–30.
Austin, John. The Province of Jurisprudence Determined. Ed. Wilfrid E. Rumble.
New York: Cambridge University Press, 1995.
Baier, Kurt. “The Justification of Governmental Authority.” Journal of Philosophy
69 (1972): 700–16.
Bales, R. E. “Act Utilitarianism: Account of Right-Making Characteristics or Decision-
Making Procedure.” American Philosophical Quarterly 8 (1971): 257–65.
Balkin, J. M., ed. What Roe v. Wade Should Have Said. New York: New York Uni-
versity Press, 2005.
Balkin, J. M., and Bruce A. Ackerman, eds. What Brown v. Board of Education
Should Have Said. New York: New York University Press, 2001.
Balkin, Jack M. “Bush v. Gore and the Boundary between Law and Politics.” Yale
Law Journal 110 (2001): 1407–58.
Banerjee, Abhijit. “A Simple Model of Herd Behavior.” Quarterly Journal of Eco-
nomics 107 (1992): 797–818.
Barak, Aharon. Judicial Discretion. Trans. Yadin Kaufmann. New Haven, Conn.:
Yale University Press, 1989.
Baron, Marcia, Philip Pettit, and Michael A. Slote. Three Methods of Ethics. Oxford:
Wiley-Blackwell, 1997.
Bassett, Debra Lyn. “Judicial Disqualification in the Federal Courts.” Iowa Law
Review 87 (2002): 1213–56.
———. “Recusal and the Supreme Court.” Hastings Law Journal 56 (2005): 657–98.
Bauer, Joseph P. “Addressing the Incoherency of the Preemption Provision of the
Copyright Act of 1976.” Vanderbilt Journal of Entertainment & Technology Law
10 (2007): 1–119.
Bayles, Michael. Procedural Justice: Allocating to Individuals. Dordrecht: Kluwer
Academic, 1989.
Benn, S. I., and R. S. Peters. The Principles of Political Thought. New York: Free Press,
1965.
Bennett, Jonathan. The Act Itself. Oxford: Oxford University Press, 1995.
Bentham, Jeremy. A Fragment on Government. Eds. J. H. Burns and H. L. A. Hart.
Cambridge: Cambridge University Press, 1988.
Berger, Raoul. Government by Judiciary: The Transformation of the Fourteenth
Amendment. Cambridge, Mass.: Harvard University Press, 1977.
Bikhchandani, Sushil, David Hirshleifer, and Ivo Welch. “Learning from the
Behavior of Others: Conformity, Fads, and Informational Cascades.” Journal of
Economic Perspectives 12 (1998): 151–70.
———. “A Theory of Fads, Fashion, Custom and Cultural Change as Informational
Cascades.” Journal of Political Economy 100 (1992): 992–1026.
Bix, Brian. Law, Language, and Legal Determinacy. Oxford: Clarendon, 1993.
Black, Hugo L. A Constitutional Faith. New York: Knopf, 1968.
Bloom, Frederic M. “State Courts Unbound.” Cornell Law Review 93 (2008):
501–54.
Bobbitt, Philip. Constitutional Interpretation. Oxford: Basil Blackwell, 1991.
Bibliography 321

Bonfield, Arthur E. “The Abrogation of Penal Statutes by Nonenforcement.” Iowa

Law Review 49 (1964): 389–440.
Bork, Robert H. The Tempting of America. New York: Simon & Schuster, 1990.
Bradley, Gerard V. “Plea Bargaining and the Criminal Defendant’s Obligation to
Plead Guilty.” South Texas Law Review 40 (1999): 65–82.
Brand-Ballard, Jeffrey. “Contractualism and Deontic Restrictions.” Ethics 114
(2004): 269–300.
———. “Innocents Lost: Proportional Sentencing and the Paradox of Collateral
Damage.” Legal Theory 15 (2009): 65–105.
Bratman, Michael E. “I Intend That We J.” Contemporary Action Theory. Eds. Ghita
Holmström-Hintikka and Raimo Tuomela. Vol. 2. Dordrecht: Kluwer, 1997,
49–63.
———. “Shared Intention.” Ethics 104 (1993): 97–113.
———. “What Is the Accordion Effect?” Journal of Ethics 10 (2006): 5–19.
Brennan, Samantha. “Thresholds for Rights.” Southern Journal of Philosophy 33
(1995): 143–68.
Brewer, Scott. “Exemplary Reasoning: Semantics, Pragmatics, and the Rational
Force of Legal Argument by Analogy.” Harvard Law Review 109 (1996): 923–
1028.
Bright, Stephen B. “Let’s Try Brian Nichols Properly the First Time.” Atlanta Journal-
Constitution, Nov. 7, 2007, 19A.
Brink, David O. “Legal Positivism and Natural Law Reconsidered.” Monist 68
(1985): 364–87.
———. Moral Realism and the Foundations of Ethics. New York: Cambridge
University Press, 1989.
Brown, Darryl K. “Jury Nullification within the Rule of Law.” Minnesota Law
Review 81 (1997): 1149–200.
Budd, Malcolm. “Wittgenstein on Seeing Aspects.” Mind 96 (1987): 1–17.
Burton, Stephen J. Judging in Good Faith. Cambridge: Cambridge University Press,
1992.
———. “Law as Practical Reason.” Southern California Law Review 62 (1989):
747–93.
Bushnell, Eleanore. Crimes, Follies and Misfortunes: The Federal Impeachment Trials.
Urbana: University of Illinois Press, 1992.
Butler, Paul. “When Judges Lie (and When They Should).” Minnesota Law Review
91 (2007): 1785–828.
Calabresi, Guido. A Common Law for the Age of Statutes. Cambridge, Mass.:
Harvard University Press, 1982.
Caminker, Evan H. “Why Must Inferior Courts Obey Superior Court Precedents?”
Stanford Law Review 46 (1994): 817–73.
Cardozo, Benjamin N. “Jurisprudence.” Selected Writings of Benjamin Nathan
Cardozo. Ed. Margaret E. Hall. New York: Fallon, 1947.
———. The Nature of the Judicial Process. New Haven, Conn.: Yale University
Press, 1921.
Carter, Stephen L. “The Religiously Devout Judge.” Notre Dame Law Review 64
(1989): 932–44.
Christopher, Russell L. “Deterring Retributivism: The Injustice of ‘Just’
Punishment.” Northwestern University Law Review 96 (2002): 843–976.
Cohen, Mark A. “The Motives of Judges: Empirical Evidence from Antitrust
Sentencing.” International Review of Law and Economics 12 (1992): 13–30.
322 Bibliography

Colby, Paul L. “Two Views of the Legitimacy of Nonacquiescence in Judicial

Opinions.” Tulane Law Review 61 (1987): 1041–69.
Coleman, Jules. The Practice of Principle. Oxford: Oxford University Press,
2001.
Coleman, Jules L., and Brian Leiter. “Determinacy, Objectivity, and Authority.”
University of Pennsylvania Law Review 142 (1993): 549–637.
Conkle, Daniel O. “Religiously Devout Judges: Issues of Personal Integrity and
Public Benefit.” Marquette Law Review 81 (1998): 523–32.
Cook, Walter Wheeler. Cases and Other Authorities on Equity. St. Paul, Minn.:
West, 1923.
Cooter, Robert D. “The Objectives of Private and Public Judges.” Public Choice 41
(1983): 107–32.
Copp, David. “Introduction: Metaethics and Normative Ethics.” Oxford Handbook
of Ethical Theory. Ed. David Copp. New York: Oxford University Press, 2006.
Corvino, John. “Homosexuality and the PIB Argument.” Ethics 115 (2005):
501–34.
Cover, Robert M. Justice Accused: Antislavery and the Judicial Process. New Haven,
Conn.: Yale University Press, 1975.
———. “Violence and the Word.” Yale Law Journal 95 (1986): 1601–29.
Crisp, Roger. “Particularizing Particularism.” Moral Particularism. Eds. Brad Hooker
and Margaret Olivia Little. Oxford: Oxford University Press, 2000.
Cullity, Garrett. The Moral Demands of Affluence. Oxford: Oxford University
Press, 2004.
———. “Moral Free Riding.” Philosophy and Public Affairs 24 (1995): 3–34.
D’Amato, Anthony. “Aspects of Deconstruction: Refuting Indeterminacy with
One Bold Thought.” Northwestern University Law Review 85 (1990): 113–27.
Dagger, Richard. “Playing Fair with Punishment.” Ethics 103 (1993): 473–88.
Damasio, Antonio R. Descartes’ Error: Emotion, Reason, and the Human Brain.
New York: Harper & Row, 1995.
Dan-Cohen, Meir. “Decision Rules and Conduct Rules: On Acoustic Separation in
Criminal Law.” Harvard Law Review 97 (1984): 625–77.
———. “In Defense of Defiance.” Philosophy and Public Affairs 23 (1994): 24–51.
Dancy, Jonathan. Ethics without Principles. Oxford: Oxford University Press,
2004.
Daniels, Norman. Justice and Justification: Reflective Equilibrium in Theory and
Practice. Cambridge: Cambridge University Press, 1996.
Darwall, Stephen L. “Agent-Centered Restrictions from the Inside Out.”
Philosophical Studies 50 (1986): 291–319.
DePaul, Michael R. Balance and Refinement: Beyond Coherence Methods of Moral
Inquiry. New York: Routledge, 1993.
Dershowitz, Alan M. Supreme Injustice: How the High Court Hijacked Election
2000. Oxford: Oxford University Press, 2001.
De Sousa, Ronald. The Rationality of Emotion. Cambridge, Mass.: MIT Press,
1987.
Dobbs, Dan B. The Law of Torts. Vol. 2. St. Paul, Minn.: West, 2001.
Drahos, Peter, and Stephen Parker. “Rule Following, Rule Scepticism and Indeter-
minacy in Law: A Conventional Account.” Ratio Juris 5 (1992): 109–19.
Dressler, Joshua. Understanding Criminal Law. 3rd ed. New York: Lexis, 2001.
Dubber, Markus Dirk. “The Pain of Punishment.” Buffalo Law Review 44 (1996):
545–611.
Bibliography 323

Duncan, William C. “Goodridge and the Rule of Law Same-Sex Marriage in

Massachusetts [sic]: The Meaning and Implications of Goodridge v. Depart-
ment of Public Health.” Boston University Public Interest Law Journal 14
(2004): 42–55.
Duxbury, Neil. “Faith in Reason: The Process Tradition in American Jurispru-
dence.” Cardozo Law Review 15 (1993): 601–705.
———. Random Justice: On Lotteries and Legal Decision-Making. Oxford:
Clarendon, 1999.
Dworkin, Ronald. “A Badly Flawed Election.” New York Review of Books, Feb. 8, 2001, 1.
———. Freedom’s Law: The Moral Reading of the American Constitution. Cambridge,
Mass.: Harvard University Press, 1996.
———. “Hard Cases.” Taking Rights Seriously. Cambridge, Mass.: Harvard University
Press, 1977. 81–130.
———. Law’s Empire. Cambridge, Mass.: Harvard University Press, 1986.
———. Life’s Dominion: An Argument about Abortion, Euthanasia, and Individual
Freedom. New York: Knopf, 1993.
———. A Matter of Principle. Cambridge, Mass.: Harvard University Press, 1985.
———. “The Model of Rules I.” Taking Rights Seriously. Cambridge, Mass.: Harvard
University Press, 1977. 14–45.
———. “The Model of Rules II.” Taking Rights Seriously. Cambridge, Mass.:
Harvard University Press, 1977.
———. “A Reply to Critics.” Taking Rights Seriously. Cambridge, Mass.: Harvard
University Press, 1978. 291–368.
———. Sovereign Virtue: The Theory and Practice of Equality. Cambridge, Mass.:
Harvard University Press, 2000.
———. Taking Rights Seriously. Cambridge, Mass.: Harvard University Press,
1977.
Dyzenhaus, David. Hard Cases in Wicked Legal Systems: South African Law in the
Perspective of Legal Philosophy. Oxford: Clarendon, 1991.
———. “Positivism’s Stagnant Research Programme.” Oxford Journal of Legal
Studies 20 (2000): 703–22.
Easterbrook, Frank. “Ways of Criticizing the Court.” Harvard Law Review 95
(1982): 802–32.
Edlin, Douglas E. Judges and Unjust Laws: Common Law Constitutionalism and the
Foundations of Judicial Review. Ann Arbor: University of Michigan Press,
2008.
Edmundson, William A. “State of the Art: The Duty to Obey the Law.” Legal
Theory 10 (2004): 215–59.
———. Three Anarchical Fallacies. Cambridge: Cambridge University Press,
1998.
Edwards, Harry T. “To Err Is Human, but Not Always Harmless: When Should Legal
Error Be Tolerated?” New York University Law Review 70 (1995): 1167–213.
Eggleston, Ben. “Self-Defeat, Publicity, and Incoherence: Three Criteria for
Consequentialist Theories.” Ph.D. diss., University of Pittsburgh, 2001.
Eisenberg, Melvin Aron. An Introduction to Agency, Partnerships, and LLCs. 3rd ed.
New York: Foundation, 2000.
Ellis, Anthony. “Deontology, Incommensurability and the Arbitrary.” Philosophy
and Phenomenological Research 52 (1992): 855–75.
Ely, John Hart. “The Wages of Crying Wolf: A Comment on Roe v. Wade.” Yale Law
Journal 82 (1973): 920–49.
324 Bibliography

Endicott, Timothy A. O. “The Impossibility of the Rule of Law.” Oxford Journal of

Legal Studies 19 (1999): 1–18.
Enoch, David. “Intending, Foreseeing, and the State.” Legal Theory 13 (2007):
69–99.
Epstein, Richard A. “Blind Justices: The Scandal of Kelo v. New London.” Wall
Street Journal, July 3, 2005.
Eskridge, William N., Jr. Dynamic Statutory Interpretation. Cambridge, Mass.:
Harvard University Press, 1994.
Estreicher, Samuel. “Judicial Nullification: Guido Calabresi’s Uncommon Common
Law for a Statutory Age.” New York University Law Review 57 (1982): 1126–73.
Estreicher, Samuel, and Richard L. Revesz. “Nonacquiescence by Federal Admin-
istrative Agencies.” Yale Law Journal 98 (1989): 679–772.
Fallon, Richard H., Jr. “‘The Rule of Law’ as a Concept in Constitutional Discourse.”
Columbia Law Review 97 (1997): 1–56.
Farley, John J., III. “Robin Hood Jurisprudence: The Triumph of Equity in American
Tort Law.” St. John’s Law Review 65 (1991): 997–1021.
Farnsworth, Ward. “‘To Do a Great Right, Do a Little Wrong’: A User’s Guide to
Judicial Lawlessness.” Minnesota Law Review 86 (2001): 227–66.
Farrell, Daniel M. “Punishment without the State.” Noûs 22 (1988): 437–53.
Feinberg, Joel. “Causing Voluntary Actions.” Metaphysics and Explanation. Eds.
W. H. Capitan and D. D. Merrill. Pittsburgh: University of Pittsburgh Press, 1966.
———. “Collective Responsibility.” Doing and Deserving. Princeton, N.J.: Princeton
University Press, 1970.
———. “Noncomparative Justice.” Philosophical Review 83 (1974): 297–338.
Fisher, William W., III, Morton J. Horwitz, and Thomas Reed, eds. American Legal
Realism. Oxford: Oxford University Press, 1993.
Fitzpatrick, William J. “The Intend/Foresee Distinction and the Problem of
‘Closeness.’” Philosophical Studies 128 (2006): 585–617.
Fletcher, George P. “Domination in the Theory of Justification and Excuse.”
University of Pittsburgh Law Review 57 (1996): 553–78.
Flikschuh, Katrin. “Reason, Right, and Revolution: Kant and Locke.” Philosophy
and Public Affairs 36 (2008): 375–404.
Forer, Lois. “Justice by Numbers.” Washington Monthly, April 1992.
Frankena, William K. “The Concept of Social Justice.” Social Justice. Ed. Richard
Brandt. Englewood Cliffs, N.J.: Prentice-Hall, 1962.
Fraser, David. “‘This Is Not Like Any Other Legal Question’: A Brief History of
Nazi Law before U.K. and U.S. Courts.” Connecticut Journal of International
Law 19 (2003): 59–125.
Fried, Charles. “The Artificial Reason of the Law Or: What Lawyers Know.” Texas
Law Review 60 (1981): 35–58.
———. “Impudence.” Supreme Court Review 1992 (1992): 155–94.
———. “Scholars and Judges: Reason and Power.” Harvard Journal of Law and
Public Policy 23 (2000): 807–32.
Friedman, Barry. “The Cycles of Constitutional Theory.” Law and Contemporary
Problems 67 (2004): 149–74.
Fuller, Lon L. The Morality of Law. 2nd rev. ed. New Haven, Conn.: Yale University
Press, 1969.
———. “Positivism and Fidelity to Law: A Reply to Professor Hart.” Harvard Law
Review 71 (1958): 630–72.
Gandhi, M. K. Selected Writings of Mahatma Gandhi. Boston: Beacon, 1951.
Bibliography 325

Gardner, John. “Concerning Permissive Sources and Gaps.” Oxford Journal of Legal
Studies 8 (1988): 457–61.
Garner, Bryan A., ed. Black’s Law Dictionary. 8th ed. St. Paul, Minn.: Thomson/
West, 2004.
Garvey, John H., and Amy V. Coney. “Catholic Judges in Capital Cases.” Marquette
Law Review 81 (1998): 303–50.
Gert, Bernard. Morality: Its Nature and Justification. New York: Oxford University
Press, 1998.
Gettier, Edmund L. “Is Justified True Belief Knowledge?” Analysis 23 (1963): 121–
23.
Geyh, Charles Gardner. “Informal Methods of Judicial Discipline.” University of
Pennsylvania Law Review 142 (1993): 243–331.
Gilbert, Margaret. “Group Membership and Political Obligation.” The Monist 76
(1993): 119–31.
———. Living Together: Rationality, Sociality, and Obligation. Lanham, Md.:
Rowman & Littlefield, 1996.
———. On Social Facts. Princeton, N.J.: Princeton University Press, 1989.
———. “Reconsidering the ‘Actual Contract’ Theory of Political Obligation.”
Ethics 109 (1999): 236–60.
———. Sociality and Responsibility: New Essays in Plural Subject Theory. Lanham,
Md.: Rowman & Littlefield, 2000.
Glover, Jonathan. “It Makes No Difference Whether or Not I Do It.” Applied Ethics.
Ed. Peter Singer. Oxford: Oxford University Press, 1990, 125–44.
Goldman, Alan H. The Moral Foundations of Professional Ethics. Totowa, N.J.:
Rowman & Littlefield, 1980.
———. Practical Rules: When We Need Them and When We Don’t. Cambridge:
Cambridge University Press, 2002.
———. “The Rationality of Complying with Rules: Paradox Resolved.” Ethics 116
(2006): 453–70.
Goodin, Robert E. Utilitarianism as a Public Philosophy. New York: Cambridge
University Press, 1995.
Goodwin, Barbara. Justice by Lottery. Hemel Hempstead, U.K.: Harvester
Wheatsheaf, 1992.
Graber, Mark A. Dred Scott and the Problem of Constitutional Evil. Cambridge:
Cambridge University Press, 2006.
Graves, Rachel. “The Terri Schiavo Case; Schiavo Dies, but Debate Lives; Delay
Insists Judges Must ‘Answer for Their Behavior.’” Houston Chronicle, April 1,
2005, A1.
Gray, Cynthia. “The Line between Legal Error and Judicial Misconduct: Balancing
Judicial Independence and Accountability.” Hofstra Law Review 32 (2004):
1245–80.
Green, Leslie. The Authority of the State. Oxford: Clarendon, 1988.
Greenawalt, Kent. Conflicts of Law and Morality. New York: Oxford University
Press, 1987.
———. “Discretion and Judicial Decision: The Elusive Quest for the Fetters that
Bind Judges.” Columbia Law Review 75 (1975): 359–99.
———. Law and Objectivity. Oxford: Oxford University Press, 1992.
———. “The Perplexing Borders of Justification and Excuse.” Columbia Law
Review 84 (1984): 1897–927.
Greenberg, Mark. “How Facts Make Law.” Legal Theory 10 (2004): 157–98.
326 Bibliography

Grice, Paul. “Logic and Conversation.” Studies in the Way of Words. Cambridge,
Mass.: Harvard University Press, 1989.
Griffen, Wendell L. “The Case for Religious Values in Judicial Decision-Making.”
Marquette Law Review 81 (1998): 513–21.
Grimes, Warren S. “Hundred-Ton-Gun Control: Preserving Impeachment as the
Exclusive Removal Mechanism for Federal Judges.” UCLA Law Review 38
(1991): 1209–55.
Gunther, Gerald. Learned Hand: The Man and the Judge. Cambridge, Mass.:
Harvard University Press, 1998.
Haley, John O. “The Civil, Criminal and Disciplinary Liability of Judges.” American
Journal of Comparative Law 54 (2006): 281–91.
Hall, Jerome. General Principles of Criminal Law. 2nd ed. Indianapolis, Ind.: Bobbs-
Merrill, 1960.
Hamburger, Philip. “Law and Judicial Duty.” George Washington Law Review 72
(2003): 1–41.
Hampton, Jean. “Free-Rider Problems in the Production of Collective Goods.”
Economics and Philosophy 3 (1987): 245–73.
Hand, Learned. The Bill of Rights. Cambridge, Mass.: Harvard University Press,
1958.
Hardimon, Michael O. “Role Obligations.” Journal of Philosophy 91 (1994): 333–63.
Harris, John. Violence and Responsibility. London: Routledge & Kegan Paul, 1980.
Hart, H. L. A. “Are There Any Natural Rights?” Philosophical Review 64 (1955):
175–91.
———. “Commands and Authoritative Legal Reasons.” Essays on Bentham. Oxford:
Oxford University Press, 1982.
———. The Concept of Law. 2nd ed. Oxford: Oxford University Press, 1994. First
published 1961.
———. “Positivism and the Separation of Law and Morals.” Harvard Law Review
71 (1958): 593–629.
Hart, H. L. A., and Tony Honoré. Causation in the Law. Oxford: Clarendon,
1959.
Hasnas, John. “The Myth of the Rule of Law.” Wisconsin Law Review 1995 (1995):
199–233.
Hatch, Orrin, and Sam Brownback. “‘Extreme’ Judicial Activism.” Washington
Times, Feb. 10, 2005, A19.
Hawkins, Keith, ed. The Uses of Discretion. Oxford: Oxford University Press,
1992.
Hellman, Deborah. “The Importance of Appearing Principled.” Arizona Law
Review 37 (1995): 1108–51.
Henderson, Lynne. “Authoritarianism and the Rule of Law.” Indiana Law Journal
66 (1991): 379–456.
Henkin, Louis. “Privacy and Autonomy.” Columbia Law Review 74 (1974): 1410–33.
Heyd, David. Supererogation. Cambridge: Cambridge University Press, 1982.
Higgins, Richard S., and Paul H. Rubin. “Judicial Discretion.” Journal of Legal
Studies 9 (1980): 129–38.
Himma, Kenneth Einar. “Judicial Discretion and the Concept of Law.” Oxford
Journal of Legal Studies 19 (1999): 71–82.
Hodges, Ann C. “Protecting Unionized Employees against Discrimination: The
Fourth Circuit’s Misinterpretation of Supreme Court Precedent.” Employee
Rights and Employment Policy Journal 2 (1998): 123–74.
Bibliography 327

Hoffer, Peter Charles. “Principled Discretion: Concealment, Conscience, and

Chancellors.” Yale Journal of Law and the Humanities 3 (1991): 53–82.
Hohfeld, Wesley Newcomb. Fundamental Legal Conceptions. New Haven, Conn.:
Yale University Press, 1919.
Honoré, Tony. “Must We Obey? Necessity as a Ground of Obligation.” Virginia
Law Review 67 (1981): 39–61.
Hooker, Brad. Ideal Code, Real World: A Rule-Consequentialist Theory of Morality.
Oxford: Oxford University Press, 2000.
Horton, Elizabeth M. “Selective Publication and the Authority of Precedent in the
United States Courts of Appeals.” UCLA Law Review 42 (1995): 1691–778.
Horwitz, Morton J. “The Rule of Law: An Unqualified Human Good?” Yale Law
Journal 86 (1977): 561–66.
———. The Transformation of American Law: 1780–1860. Cambridge, Mass.:
Harvard University Press, 1979.
Hulse, Carl. “Lawmakers Vow to Fight Judges’ Ruling on the Pledge.” New York
Times, June 27, 2002, A20.
Hurd, Heidi M. “Justifiably Punishing the Justified.” Michigan Law Review 90
(1992): 2203–324.
———. Moral Combat. Cambridge: Cambridge University Press, 1999.
Hurley, Paul. “Agent-Centered Restrictions: Clearing the Air of Paradox.” Ethics
108 (1997): 120–46.
Hurley, Susan. Natural Reasons. Oxford: Oxford University Press, 1989.
Husak, Douglas. Overcriminalization: The Limits of the Criminal Law. New York:
Oxford University Press, 2008.
Hutchinson, Allan C. Dwelling on the Threshold: Critical Essays on Modern Legal
Thought. Toronto: Carswell, 1988.
Idleman, Scott C. “A Prudential Theory of Judicial Candor.” Texas Law Review 73
(1995): 1307–417.
———. “The Role of Religious Values in Judicial Decision Making.” Indiana Law
Journal 68 (1993): 433–87.
Igneski, Violetta. “Distance, Determinacy and the Duty to Aid: A Reply to Kamm.”
Law and Philosophy 20 (2001): 605–16.
Ingram, Peter. “Maintaining the Rule of Law.” Philosophical Quarterly 35 (1985):
359–81.
Isaacs, Tracy. “Collective Moral Responsibility and Collective Intention.” Midwest
Studies in Philosophy 30 (2006): 59–73.
Jackson, Frank, and Michael Smith. “Absolutist Ethical Theories and Uncertainty.”
Journal of Philosophy 103 (2006): 267–83.
Jacobson, Arthur J. “Publishing Dissent.” Washington and Lee Law Review 62
(2005): 1607–36.
Johnston, David, and Neil A. Lewis. “Ending Raids of Dispensers of Marijuana for
Patients.” New York Times, March 19, 2009, A20.
Kadish, Mortimer R., and Sanford H. Kadish. Discretion to Disobey: A Study of
Lawful Departures from Legal Rules. Stanford, Calif.: Stanford University Press,
1973.
Kagan, Shelly. The Limits of Morality. Oxford: Oxford University Press, 1989.
Kamm, F. M. Intricate Ethics: Rights, Responsibilities and Permissible Harm. New
York: Oxford University Press, 2007.
———. Morality, Mortality: Rights, Duties, and Status. Vol. 2. New York: Oxford
University Press, 1996.
328 Bibliography

Kaplow, Louis. “Rules versus Standards: An Economic Analysis.” Duke Law Journal
42 (1992): 557–629.
Kaplow, Louis, and Steven Shavell. Fairness versus Welfare. Cambridge, Mass.:
Harvard University Press, 2001.
Karlan, Pamela S. “Two Concepts of Judicial Independence.” Southern California
Law Review 72 (1999): 535–58.
Katz, Leo. “Incommensurable Choices and the Problem of Moral Ignorance.”
University of Pennsylvania Law Review 146 (1998): 1465–85.
Kauper, Paul G. “Penumbras, Peripheries, Emanations, Things Fundamental and
Things Forgotten: The Griswold Case.” Michigan Law Review 64 (1965): 235–58.
Keith, Darrell L. “The Court’s Charge in Texas Medical Malpractice Cases.” Baylor
Law Review 48 (1996): 675–814.
Kelsen, Hans. “On the Theory of Interpretation.” Legal Studies 10 (1990): 127–35.
Kennedy, Duncan. A Critique of Adjudication (Fin De Siècle). Cambridge, Mass.:
Harvard University Press, 1997.
———. “Form and Substance in Private Law Adjudication.” Harvard Law Review
89 (1976): 1685–778.
———. “Freedom and Constraint in Adjudication: A Critical Phenomenology.”
Journal of Legal Education 36 (1986): 518–62.
Klarman, Michael J. “Bush v. Gore through the Lens of Constitutional History.”
California Law Review 89 (2001): 1721–65.
Klosko, George. “Parfit’s Moral Arithmetic and the Obligation to Obey the Law.”
Canadian Journal of Philosophy 20 (1990): 191–214.
———. “Presumptive Benefit, Fairness, and Political Obligation.” The Duty to Obey
the Law: Selected Philosophical Readings. Ed. William A. Edmundson. Lanham,
Md.: Rowman & Littlefield, 1999.
———. The Principle of Fairness and Political Obligation. Lanham, Md.: Rowman &
Littlefield, 1992.
Kmiec, Keenan D. “The Origin and Current Meanings of ‘Judicial Activism.’”
California Law Review 92 (2004): 1441–77.
Kohlberg, Lawrence. Essays on Moral Development: The Philosophy of Moral
Development. Vol. 1. New York: Harper & Row, 1981.
Kornhauser, Lewis A. “Adjudication by a Resource-Constrained Team: Hierarchy
and Precedent in a Judicial System.” Southern California Law Review 68 (1995):
1605–29.
———. “Modeling Collegial Courts I: Path Dependence.” International Review of
Law and Economics 12 (1992): 169–85.
———. “Modeling Collegial Courts II: Legal Doctrine.” Journal of Law, Economics,
and Organization 8 (1992): 441–70.
Korsgaard, Christine M. “Skepticism about Practical Reason.” Journal of Philosophy
83 (1986): 5–25.
Kramer, Larry D. The People Themselves: Popular Constitutionalism and Judicial
Review. Oxford: Oxford University Press, 2004.
Kramer, Matthew. “How Moral Principles Can Enter into the Law.” Legal Theory 6
(2000): 83–108.
Kress, Ken. “Legal Indeterminacy.” California Law Review 77 (1989): 283–337.
———. “Why No Judge Should Be a Dworkinian Coherentist.” Texas Law Review
77 (1999): 1375–427.
Kretzmann, Norman. “Lex Iniusta Non Est Lex: Laws on Trial in Aquinas’ Court
of Conscience.” American Journal of Jurisprudence 33 (1988): 99–122.
Bibliography 329

Kroger, John R. “Supreme Court Equity, 1789–1835, and the History of American
Judging.” Houston Law Review 34 (1998): 1425–86.
Kronman, Anthony. The Lost Lawyer: Failing Ideals of the Legal Profession.
Cambridge, Mass.: Belknap, 1993.
Kuran, Timur, and Cass R. Sunstein. “Availability Cascades and Risk Regulation.”
Stanford Law Review 51 (1999): 683–768.
Kutz, Christopher. Complicity: Ethics and Law for a Collective Age. Cambridge:
Cambridge University Press, 2000.
———. “The Judicial Community.” Philosophical Issues 11 (2001): 442–69.
Lamond, Grant. “The Coerciveness of Law.” Oxford Journal of Legal Studies 20
(2000): 39–62.
Landers, Scott. “Wittgenstein, Realism, and CLS: Undermining Rule Scepticism.”
Law and Philosophy 9 (1990): 177–203.
Landes, William M., and Richard A. Posner. “Legal Precedent: A Theoretical and
Empirical Analysis.” Journal of Law and Economics 19 (1976): 249–307.
Lefkowitz, David. “A Contractualist Defense of Democratic Authority.” Ratio Juris
18 (2005): 346–64.
Leiter, Brian. “Heidegger and the Theory of Adjudication.” Yale Law Journal 106
(1996): 253–82.
———. “Legal Realism and Legal Positivism Reconsidered.” Ethics 111 (2001):
278–301.
Levenbook, Barbara B. “The Meaning of a Precedent.” Legal Theory 6 (2000): 185–
240.
Levinson, Sanford. “The Confrontation of Religious Faith and Civil Religion:
Catholics Becoming Justices.” DePaul Law Review 39 (1990): 1047–81.
Lewis, David. Convention. Cambridge, Mass.: Harvard University Press, 1969.
———. “The Punishment That Leaves Something to Chance.” Philosophy and
Public Affairs 18 (1989): 53–67.
Li, Xiaorong. “License to Coerce: Violence against Women, State Responsibility,
and Legal Failures in China’s Family-Planning Program.” Yale Journal of Law
and Feminism 8 (1996): 145–91.
Lightman, David. “Lawmakers Stand up to Court; Bipartisan Coalition Hopes to
Dilute Impact of Eminent Domain Ruling.” Hartford Courant, July 1, 2005,
A1.
Locke, John. Second Treatise of Government. Ed. C. B. McPherson. Indianapolis:
Hackett, 1980.
Lode, Eric. “Slippery Slope Arguments and Legal Reasoning.” California Law
Review 87 (1999): 1469–543.
Luban, David. Lawyers and Justice: An Ethical Study. Princeton, N.J.: Princeton
University Press, 1988.
Lubet, Steven. “Judicial Discipline and Judicial Independence.” Law and Contem-
porary Problems 61 (1998): 59–74.
Luce, R. Duncan, and Howard Raiffa. Games and Decisions. New York: Dover,
1989.
Lyons, David. “The Correlativity of Rights and Duties “ Noûs 4 (1970): 45–55.
———. “Derivability, Defensibility, and the Justification of Judicial Decisions.”
Moral Aspects of Legal Theory. Cambridge: Cambridge University Press, 1993.
———. Forms and Limits of Utilitarianism. Oxford: Clarendon, 1965.
———. “On Formal Justice.” Moral Aspects of Legal Theory. Cambridge: Cambridge
University Press, 1993.
330 Bibliography

MacAdam, James I. “The Precepts of Justice.” Mind 77 (1968): 360–71.

Macchiarola, Frank J. “Why the Decision in Zelman Makes So Much Sense.” N.Y.U.
Annual Survey of American Law 59 (2003): 459–67.
MacCormick, Neil. Legal Reasoning and Legal Theory. Oxford: Clarendon, 1978.
———. “Voluntary Obligations and Normative Powers I.” Proceedings of the
Aristotelian Society 46 (1972): 59–78.
Macey, Jonathan R. “The Internal and External Costs and Benefits of Stare Decisis.”
Chicago-Kent Law Review 65 (1989): 93–112.
Mackie, J. L. “Clauses and Conditions.” American Philosophical Quarterly 2 (1965):
245–64.
———. Ethics: Inventing Right and Wrong. Harmondsworth, U.K.: Penguin, 1977.
———. “Morality and the Retributive Emotions.” Persons and Values. Oxford:
Clarendon, 1985.
———. “Obligations to Obey the Law.” Virginia Law Review 67 (1981): 143–58.
Manning, John F. “The Absurdity Doctrine.” Harvard Law Review 116 (2003):
2387–486.
———. “Textualism and the Equity of the Statute.” Columbia Law Review 101
(2001): 1–127.
Marmor, Andrei. “The Rule of Law and Its Limits.” Law and Philosophy 23 (2004):
1–43.
———. “Should Like Cases Be Treated Alike?” Legal Theory 11 (2005): 27–38.
Mason, Elinor. “We Make No Promises.” Philosophical Studies 123 (2005): 33–46.
McCarthy, David. “Rights, Explanation, and Risks.” Ethics 107 (1997): 205–25.
McCarthy, Nancy. “Judge Faces Discipline: Commission Charge Unleashes Protest
in Legal Community.” California State Bar Journal, August 1998, 1–3.
McDermott, Daniel. “The Permissibility of Punishment.” Law and Philosophy 20
(2001): 403–32.
McDowell, Gary. Equity and the Constitution: The Supreme Court, Equitable Relief,
and Public Policy. Chicago: University of Chicago Press, 1982.
McIntyre, Alison. “Doing Away with Double Effect.” Ethics 111 (2001): 219–55.
McKeever, Sean, and Michael Ridge. Principled Ethics: Generalism as a Regulative
Ideal. Oxford: Oxford University Press, 2006.
McKerlie, Dennis. “Rights and Risk.” Canadian Journal of Philosophy 16 (1986):
239–51.
McMahan, Jeff. “Self-Defense and the Problem of the Innocent Attacker.” Ethics
104 (1994): 252–90.
McMahon, Christopher. Collective Rationality and Collective Reasoning. Cambridge:
Cambridge University Press, 2001.
Mian, Emran. “The Curious Case of Exclusionary Reasons.” Canadian Journal of
Law and Jurisprudence 15 (2002): 99–124.
Miceli, Thomas J., and Metin M. Cogel. “Reputation and Judicial Decision-
Making.” Journal of Economic Behavior and Organization 23 (1994): 31–51.
Miller, Arthur R. “The Pretrial Rush to Judgment: Are the ‘Litigation Explosion,’
‘Liability Crisis,’ and Efficiency Cliches Eroding Our Day in Court and Jury
Trial Commitments?” New York University Law Review 78 (2003): 982–1134.
Miller, David. On Nationality. Oxford: Oxford University Press, 1995.
Miller, Richard W. “Beneficence, Duty and Distance.” Philosophy and Public Affairs
32 (2004): 357–83.
Miller, William Ian. Bloodtaking and Peacemaking: Feud, Law, and Society in Saga
Iceland. Chicago: University of Chicago Press, 1990.
Bibliography 331

Mintz, Howard. “Disciplinary Case against Judge Raises Legal Uproar.” San Jose
Mercury News July 11, 1998, 1A.
Mitchell, Allison. “Clinton Pressing Judge to Relent.” New York Times, March 22,
1996, A1.
Modak-Truran, Mark C. “Reenchanting the Law: The Religious Dimension of
Judicial Decision Making.” Catholic University Law Review 53 (2004): 709–816.
Moore, Michael. “Law as a Functional Kind.” Natural Law Theory: Contemporary
Essays. Ed. Robert P. George. Oxford: Oxford University Press, 1992.
Moore, Michael S. “Authority, Law, and Razian Reasons.” Southern California Law
Review 62 (1989): 827–96.
———. “The Metaphysics of Causal Intervention.” California Law Review 88
(2000): 827–77.
———. Placing Blame. Oxford: Clarendon, 1997.
Murphy, Liam B. Moral Demands in Nonideal Theory. Oxford: Oxford University
Press, 2000.
Murphy, Mark C. “Natural Law Jurisprudence.” Legal Theory 9 (2003): 241–67.
———. “Natural Law, Consent, and Political Obligation.” Social Philosophy and
Policy 18 (2001): 70–92.
Nagel, Thomas. The View from Nowhere. Oxford: Oxford University Press, 1986.
———. “War and Massacre.” Mortal Questions. Cambridge: Cambridge University
Press, 1979.
Newman, Jon O. “The Judge Baer Controversy.” Judicature 80 (1997): 156–64.
Nichols, Shaun. Sentimental Rules: On the Natural Foundations of Moral Judgment.
Oxford: Oxford University Press, 2004.
Noonan, John T., Jr. Persons and Masks of the Law: Cardozo, Holmes, Jefferson, Holmes,
and Wythe as Makers of the Masks. New York: Farrar, Straus & Giroux, 1976.
———. “The Religion of the Justice: Does It Affect Constitutional Decision
Making?” Tulsa Law Review 42 (2007): 761–70.
Nozick, Robert. Anarchy, State, and Utopia. New York: Basic Books, 1974.
Nussbaum, Martha C. “Emotion in the Language of Judging.” St. John’s Law Review
70 (1996): 23–30.
O’Hara, Erin. “Social Constraint or Implicit Collusion? Toward a Game Theoretic
Analysis of Stare Decisis.” Seton Hall Law Review 24 (1993): 736–78.
Olowofoyeku, Abimbola A. Suing Judges: A Study of Judicial Immunity. Oxford:
Clarendon, 1993.
Orsi, Francesco. “Obligations of Nearness.” Journal of Value Inquiry 42 (2008):
1–21.
Osborn, Rebekah L. “Beliefs on the Bench: Recusal for Religious Reasons and the
Model Code of Judicial Conduct.” Georgetown Journal of Legal Ethics 19 (2006):
895–905.
Otsuka, Michael. “The Paradox of Group Beneficence.” Philosophy and Public
Affairs 20 (1991): 132–49.
Parfit, Derek. Reasons and Persons. Oxford: Clarendon, 1984.
Parker, Tom. “Alabama Justices Surrender to Judicial Activism.” Birmingham (Ala.)
News, January 1, 2005, 4B.
Paulsen, Michael Stokes. “Accusing Justice: Some Variations on the Themes of
Robert M. Cover’s Justice Accused.” Journal of Law and Religion 7 (1989):
33–97.
Peirce, Charles S. Collected Papers of Charles Sanders Peirce. Eds. Charles Hartshorne
and Paul Weiss. Cambridge, Mass.: Harvard University Press, 1932.
332 Bibliography

Penzell, Abigail. “Note: Apology in the Context of Wrongful Conviction: Why the
System Should Say It’s Sorry.” Cardozo Journal of Conflict Resolution 9 (2007):
145–61.
Perry, Stephen R. “Second-Order Reasons, Uncertainty, and Legal Theory.” Southern
California Law Review 62 (1989): 913–94.
Pettit, Philip. The Common Mind: An Essay on Psychology, Society, and Politics.
Oxford: Oxford University Press, 1993.
———. “The Consequentialist Can Recognise Rights.” Philosophical Quarterly 38
(1988): 42–55.
———. “Responsibility Incorporated.” Ethics 117 (2007): 171–201.
———. “Universality without Utilitarianism.” Mind 72 (1987): 74–82.
Plato. Crito.Trans. Hugh Tredennick. Baltimore, Md.: Penguin, 1959.
Posner, Eric. Law and Social Norms. Cambridge, Mass.: Harvard University Press,
2000.
Posner, Richard A. Economic Analysis of Law. 4th ed. Boston: Little, Brown, 1992.
———. The Problematics of Moral and Legal Theory. Cambridge, Mass.: Belknap,
1999.
———. “What Do Judges and Justices Maximize? (the Same Thing Everybody
Else Does).” Supreme Court Economic Review 3 (1993): 1–41.
Postema, Gerald J. Bentham and the Common Law Tradition. Oxford: Oxford Uni-
versity Press, 1986.
———. “Moral Responsibility in Professional Ethics.” New York University Law
Review 55 (1980): 63–89.
Prichard, H. A. “The Obligation to Keep a Promise.” Moral Obligation. Oxford:
Clarendon, 1949.
Prosser, William Lloyd, et al. Prosser and Keeton on the Law of Torts. 5th ed.
St. Paul, Minn.: West, 1984.
Pryor, William H., Jr. “The Religious Faith and Judicial Duty of an American
Catholic Judge.” Yale Law and Policy Review 24 (2006): 347–62.
Quinn, Warren. “Actions, Intentions, and Consequences: The Doctrine of Double
Effect.” Philosophy and Public Affairs 18 (1989): 334–51.
———. “The Right to Threaten and the Right to Punish.” Philosophy and Public
Affairs 14 (1985): 327–73.
Rabin, Edward H. “Symposium: The Revolution in Residential Landlord-Tenant
Law: Causes and Consequences.” Cornell Law Review 69 (1984): 517–84.
Radbruch, Gustav. “Gesetzliches Unrecht Und Übergesetzliches Recht.”
Süddeutsche Juristen-Zeitung 1 (1946): 105–8. Translated as “Statutory Lawless-
ness and Supra-Statutory Law (1946).” Trans. Bonnie Litschewski Paulson and
Stanley L. Paulson. Oxford Journal of Legal Studies 26 (2006): 1–11.
Radin, Margaret Jane. “Reconsidering the Rule of Law.” Boston University Law
Review 69 (1992): 781–819.
Raftery, William E. “The Legislatures, the Ballot Boxes, and the Courts.” Court
Review 43 (2006): 102–7.
Railton, Peter. “Alienation, Consequentialism, and the Demands of Morality.”
Philosophy and Public Affairs 13 (1984): 134–71.
———. “Locke, Stock, and Peril: Natural Property Rights, Pollution, and Risk.”
To Breathe Freely. Ed. Mary Gibson. Totowa, N.J.: Rowman & Littlefield,
1985.
Rawls, John. “The Idea of Public Reason Revisited.” University of Chicago Law
Review 64 (1997): 765–807.
Bibliography 333

———. “Legal Obligation and the Duty of Fair Play.” Law and Philosophy: A
Symposium. Ed. Sidney Hook. New York: New York University Press, 1964.
———. Political Liberalism. New York: Columbia University Press, 1993.
———. A Theory of Justice. Cambridge, Mass.: Harvard University Press, 1971.
Rawls, Phillip. “Parker ‘Attack’ Irks Fellow State Justice.” Huntsville Times
(Alabama), January 16, 2006, 2B.
Raz, Joseph. The Authority of Law. Oxford: Clarendon, 1979.
———. “Law and Value in Adjudication.” The Authority of Law. Oxford: Clarendon,
1979.
———. “Legal Positivism and the Sources of Law.” The Authority of Law. Oxford:
Clarendon, 1979.
———. “Legal Reasons, Sources, and Gaps.” The Authority of Law: Essays in Law
and Morality. Oxford: Clarendon, 1979.
———. The Morality of Freedom. Oxford: Clarendon, 1986.
———. “The Obligation to Obey: Revision and Tradition.” Ethics in the Public
Domain. Rev. ed. Oxford: Clarendon, 1994.
———. Practical Reason and Norms. 1970. 2nd ed. New York: Oxford University
Press, 1990.
———. “Promises in Morality and Law.” Rev. of P. S. Atiyah, Promises, Morals, and
Law. Harvard Law Review 95 (1982): 916–38.
———. “The Rule of Law and Its Virtue.” The Authority of Law. Oxford: Clarendon,
1979.
Regan, Donald H. “Authority and Value: Reflections on Raz’s Morality of Freedom.”
Southern California Law Review 62 (1989): 995–1095.
Reibetanz (now Moreau), Sophia. “A Problem for the Doctrine of Double Effect.”
Proceedings of the Aristotelian Society 98 (1998): 217–23.
Reynolds, William L., and William M. Richman. “The Non-Precedential
Precedent—Limited Publication and No-Citation Rules in the United States
Courts of Appeals.” Columbia Law Review 78 (1978): 1167–208.
Richardson, Henry S. “Review of Alan H. Goldman, Practical Rules.” Notre Dame
Philosophical Reviews, 2002.
———. “Specifying Norms as a Way to Resolve Concrete Ethical Problems.”
Philosophy and Public Affairs 19 (1990): 279–310.
Ridge, Michael. “Reasons for Action: Agent-Neutral vs. Agent-Relative.” Stanford
Encyclopedia of Philosophy, August 11, 2005. Retrieved July 21, 2009, from
http://plato.stanford.edu/entries/reasons-agent/.
Ring, Wilson. “Vt. Judge Criticized for Molester Sentence.” Associated Press Online,
January 10, 2006.
Rizzo, Mario J., and Douglas Glen Whitman. “The Camel’s Nose Is under the Tent:
Rules, Theories, and Slippery Slopes.” UCLA Law Review 51 (2003): 539–92.
Roberts, Thomas E. “Facial Takings Claims under Agins-Nectow: A Procedural
Loose End.” Hawaii Law Review 24 (2002): 623–55.
Robinson, Paul H. “The A.L.I.’s Proposed Distributive Principle of ‘Limiting
Retributivism’: Does It Mean in Practice Anything Other than Pure Desert?”
Buffalo Criminal Law Review 7 (2003): 3–15.
Ross, W.D. The Right and the Good. Oxford: Clarendon, 1930.
Roth, Abraham Sesshu. “Shared Agency and Contralateral Commitments.”
Philosophical Review 113 (2004): 359–410.
Rovane, Carol. The Bounds of Agency. Princeton, N.J.: Princeton University Press,
1998.
334 Bibliography

Sachs, Stephen E. “Saving Toby: Extortion, Blackmail, and the Right to Destroy.”
Yale Law and Policy Review 24 (2006): 251–61.
Sagoff, Mark. “Controlling Global Climate: The Debate over Pollution Trading.”
Philosophical Dimensions of Public Policy. Ed. Verna V. Gehring. New Brunswick,
N.J.: Transaction, 2002.
Salmon, Marylynn. Women and the Law of Property in Early America. Chapel Hill:
University of North Carolina Press, 1986.
Sankar, Sambhav N. “Disciplining the Professional Judge.” California Law Review
88 (2000): 1233–80.
Saphire, Richard B. “Religion and Recusal.” Marquette Law Review 81 (1998):
351–63.
Sartorius, Rolf. Individual Conduct and Social Norms. Encino, Calif.: Dickenson,
1975.
———. “Political Authority and Political Obligation.” Virginia Law Review 67
(1981): 3–17.
Scalia, Antonin. A Matter of Interpretation: Federal Courts and the Law. Princeton,
N.J.: Princeton University Press, 1997.
———. “Originalism: The Lesser Evil.” University of Cincinnati Law Review 57
(1989): 849–65.
———. “The Rule of Law as a Law of Rules.” University of Chicago Law Review 56
(1989): 1175–88.
Scanlon, T. M. What We Owe to Each Other. Cambridge, Mass.: Belknap, 1998.
Schapiro, Tamar. “Compliance, Complicity, and the Nature of Nonideal Conditions.”
Journal of Philosophy 100 (2003): 329–55.
Schauer, Frederick. “Easy Cases.” Southern California Law Review 58 (1985): 399–
440.
———. “Incentives, Reputation, and the Inglorious Determinants of Judicial
Behavior.” University of Cincinnati Law Review 68 (1999): 615–36.
———. Playing by the Rules: A Philosophical Examination of Rule-Based Decision-
Making in Law and in Life. Oxford: Oxford University Press, 1991.
———. “Precedent.” Stanford Law Review 29 (1987): 571–605.
———. “Slippery Slopes.” Harvard Law Review 99 (1985): 361–83.
Scheffler, Samuel. The Rejection of Consequentialism. 1982. Revised ed. Oxford:
Oxford University Press, 1994.
Schmidtz, David. “Islands in a Sea of Obligation: Limits of the Duty to Rescue.”
Law and Philosophy 19 (2000): 683–705.
Searle, John R. “Collective Intentions and Intentional Actions.” Intentions in
Communication. Eds. Philip R. Cohen, Jerry Morgan, and Martha E. Pollack.
Cambridge, Mass.: MIT Press, 1990.
Seelye, Katharine Q. “A Get-Tough Message at California’s Death Row.” New York
Times, March 24, 1996, 29.
Shaman, Jeffrey M. “Judicial Ethics.” Georgetown Journal of Legal Ethics 2 (1988):
1–20.
Shapiro, David. “In Defense of Judicial Candor.” Harvard Law Review 100 (1995):
731–50.
Shapiro, Ian, ed. The Rule of Law: Nomos XXXVI. New York: New York University
Press, 1994.
Shapiro, Martin. “Judges as Liars.” Harvard Journal of Law and Public Policy 17
(1994): 155–56.
Shapiro, Scott J. “Judicial Can’t.” Noûs 35, Supp. 1 (2001): 530–57.
Bibliography 335

———. “Law, Morality, and the Guidance of Conduct.” Legal Theory 6 (2000):
127–70.
———. “Law, Plans, and Practical Reason.” Legal Theory 8 (2002): 387–441.
———. “On Hart’s Way Out.” Hart’s Postscript. Ed. Jules Coleman. Oxford: Oxford
University Press, 2000.
Shepard, Sam. Fool for Love. New York: Dramatists Play Service, 1984.
Sher, George. Approximate Justice: Studies in Non-Ideal Theory. Lanham, Md.:
Rowman & Littlefield, 1997.
Sherrer, Hans. “The Complicity of Judges in the Generation of Wrongful Convic-
tions.” Northern Kentucky Law Review 30 (2003): 539–83.
Shklar, Judith N. “Political Theory and the Rule of Law.” The Rule of Law: Ideal or
Ideology. Eds. Allan C. Hutchinson and Patrick Monahan. Toronto: Carswell, 1987.
Sidgwick, Henry. The Methods of Ethics. 7th ed. Indianapolis: Hackett, 1981. First
published 1874.
Siegel, Neil S. “The Virtue of Judicial Statesmanship.” Texas Law Review 86 (2008):
959–1032.
———. “Why President Bush Should Not Take the 5th; Judges Who Ignore Law
Are Possible Court Candidates.” Houston Chronicle, June 17, 2005, B11.
Simmons, A. John. “Locke and the Right to Punish.” Philosophy and Public Affairs
20 (1991): 311–49.
———. Moral Principles and Political Obligations. Princeton, N.J.: Princeton
University Press, 1979.
Simon, Dan. “The Double-Consciousness of Judging: The Problematic Legacy of
Cardozo.” Oregon Law Review 79 (2000): 1033–80.
Simon, William H. The Practice of Justice: A Theory of Lawyers’ Ethics. Cambridge,
Mass.: Harvard University Press, 1998.
———. “Should Lawyers Obey the Law?” William & Mary Law Review 38 (1996):
217–53.
Singer, Peter. “Famine, Affluence, and Morality.” Philosophy and Public Affairs 1
(1972): 229–43
Smith, M. B. E. “Do Appellate Courts Regularly Cheat?” Criminal Justice Ethics 16
(1997): 11–20.
———. “Is There a Prima Facie Obligation to Obey the Law?” Yale Law Journal 82
(1973): 950–76.
———. “May Judges Ever Nullify the Law?” Notre Dame Law Review 74 (1999):
1657–71.
Smith, Matthew Noah. “The Law as a Social Practice: Are Shared Activities at the
Foundations of Law?” Legal Theory 12 (2006): 265–92.
Smith, Steven D. “Why Should Courts Obey the Law?” Georgetown Law Journal
77 (1988): 113–64.
Solum, Lawrence B. “Equity and the Rule of Law.” The Rule of Law. Ed. Ian Shapiro.
New York: New York University Press, 1994.
———. “Faith and Justice.” DePaul Law Review 39 (1990): 1083–106.
———. “Indeterminacy.” A Companion to Philosophy of Law and Legal Theory. Ed.
Dennis Patterson. Oxford: Blackwell, 1996.
———. “On the Indeterminacy Crisis: Critiquing Critical Dogma.” University of
Chicago Law Review 54 (1987): 462–503.
Songer, Donald R., Martha Humphries Ginn, and Tammy A. Sarver. “Do Judges
Follow the Law When There Is No Fear of Reversal?” Justice System Journal 24
(2003): 137–61.
336 Bibliography

Soper, Philip. The Ethics of Deference: Learning from Law’s Morals. Cambridge:
Cambridge University Press, 2002.
———. A Theory of Law. Cambridge, Mass.: Harvard University Press, 1984.
Sowell, Thomas. “Real Judicial Crisis Is Judges Who Ignore the Law.” The Post and
Courier (Charleston, S.C.), January 14, 1998, A11.
Stark, Cynthia A. “Decision Procedures, Standards of Rightness and Impartiality.”
Noûs 31 (1997): 478–95.
Stavropoulos, Nicos. Objectivity in Law. Oxford: Clarendon, 1996.
Stern, Gerald. “Is Judicial Discipline in New York State a Threat to Judicial
Independence?” Pace Law Review 7 (1987): 291–388.
Strauss, David A. “The Common Law Genius of the Warren Court.” William &
Mary Law Review 49 (2007): 845–79.
———. “Little Rock and the Legacy of Brown.” Saint Louis Law Journal 52 (2008):
1065–86.
Suleiman, Ramzi. “Provision of Step-Level Public Goods under Uncertainty.”
Rationality and Society 9 (1997): 163–87.
Sullivan, Kathleen M. “The Justices of Rules and Standards.” Harvard Law Review
106 (1992): 22–123.
Sulmasy, Daniel P. “What Is an Oath and Why Should a Physician Swear One?”
Theoretical Medicine and Bioethics 20 (1999): 329–46.
Summers, Robert S. “The Principles of the Rule of Law.” Notre Dame Law Review
74 (1999): 1691–712.
Sunstein, Cass R. “Moral Heuristics.” Behavioral and Brain Sciences 28 (2005):
531–73.
———. “On Academic Fads and Fashions.” Michigan Law Review 99 (2001):
1251.
———. One Case at a Time: Judicial Minimalism on the Supreme Court. Cambridge,
Mass.: Harvard University Press, 1999.
———. “The Supreme Court 1995 Term: Foreword: Leaving Things Undecided.”
Harvard Law Review 110 (1996): 6–101.
———. “What Did Lawrence Hold? Of Autonomy, Desuetude, Sexuality, and
Marriage.” Supreme Court Review (2003): 27–74.
———. Why Societies Need Dissent. Cambridge, Mass.: Harvard University Press,
2003.
Sunstein, Cass R., and Richard A. Epstein, eds. The Vote: Bush, Gore, and the
Supreme Court. Chicago: University of Chicago Press, 2001.
Sunstein, Cass R., and Adrian Vermeule. “Deterring Murder: A Reply.” Stanford
Law Review 58 (2006): 847–57.
“Symposium: Bush v. Gore.” University of Chicago Law Review 68 (2001): 613–
791.
Sypnowich, Christine. “Utopia and the Rule of Law.” Recrafting the Rule of Law.
Ed. David Dyzenhaus. Oxford: Hart, 1999.
Talley, Eric. “Precedential Cascades: An Appraisal.” Southern California Law Review
73 (1999): 87–137.
Tamanaha, Brian Z. Law as a Means to an End: Threat to the Rule of Law. Law in
Context. New York: Cambridge University Press, 2006.
———. On the Rule of Law: History, Politics, Theory. Cambridge: Cambridge
University Press, 2004.
Temkin, Larry S. “Rethinking the Good, Moral Ideals and the Nature of Practical
Reasons.” Reading Parfit. Ed. Jonathan Dancy. Oxford: Blackwell, 1997.
Bibliography 337

Ten, C. L. “Positive Retributivism.” Social Philosophy and Policy 7 (1990): 194–

208.
Thayer, James B. “The Origin and Scope of the American Doctrine of Constitu-
tional Law.” Harvard Law Review 7 (1893): 129–56.
Thomson, Judith Jarvis. Rights, Restitution, and Risk. Cambridge, Mass.: Harvard
University Press, 1986.
———. “Self-Defense.” Philosophy and Public Affairs 20 (1991): 283–310.
Tuomela, Raimo. The Importance of Us. Stanford, Calif.: Stanford University Press,
1995.
Tuomela, Raimo, and Kaarlo Miller. “We-Intentions.” Philosophical Studies 53
(1988): 115–37.
Tushnet, Mark. Taking the Constitution Away from the Courts. Princeton, N.J.:
Princeton University Press, 1999.
Tyler, Tom R. Why People Obey the Law. New Haven, Conn.: Yale University Press,
1990.
Unger, Roberto Mangabeira. Law in Modern Society. New York: Free Press, 1976.
———. What Should Legal Analysis Become? London: Verso, 1996.
Velleman, J. David. “How to Share an Intention.” Philosophy and Phenomenological
Research 57 (1997): 29–50.
———. The Possibility of Practical Reason. Oxford: Oxford University Press,
2000.
Vermeule, Adrian. Judging under Uncertainty: An Institutional Theory of Legal
Interpretation. Cambridge, Mass.: Harvard University Press, 2006.
Vila, Marisa Iglesias. Facing Judicial Discretion: Legal Knowledge and Right Answers
Revisited. Boston: Kluwer, 2001.
Volcansek, Mary L. Judicial Impeachment: None Called for Justice. Urbana:
University of Illinois Press, 1993.
Volokh, Eugene. “The Mechanisms of the Slippery Slope.” Harvard Law Review
116 (2003): 1026–137.
Wald, Patricia M. “The Rhetoric of Results and the Results of Rhetoric: Judicial
Writings.” University of Chicago Law Review 62 (1995): 1371–419.
———. “Violence under the Law: A Judge’s Perspective.” Law’s Violence. Eds.
Austin Sarat and Thomas R. Kearns. Ann Arbor: University of Michigan Press,
1992.
Waldron, Jeremy. “The Core of the Case against Judicial Review.” Yale Law Journal
115 (2006): 1346–406.
———. “Kant’s Legal Positivism.” Harvard Law Review 109 (1996): 1535–66.
———. Law and Disagreement. Oxford: Oxford University Press, 1999.
———. “The Rule of Law in Contemporary Liberal Theory.” Ratio Juris 2 (1989):
79–96.
———. “Who Is My Neighbor? Humanity and Proximity.” Monist 86 (2003):
333–54.
Walen, Alec. “Reasonable Illegal Force: Justice and Legitimacy in a Pluralistic,
Liberal Society.” Ethics 111 (2001): 344–73.
Walker, A. D. M. “Political Obligation and the Argument from Gratitude.”
Philosophy and Public Affairs 17 (1988): 191–211.
Walton, Douglas. Slippery Slope Arguments. Oxford: Clarendon, 1992.
Waluchow, W. J. “Authority and the Practical Difference Thesis.” Legal Theory
6 (2000): 45–81.
———. Inclusive Legal Positivism. Oxford: Clarendon, 1994.
338 Bibliography

———. “Indeterminacy.” Canadian Journal of Law and Jurisprudence 9 (1996):

397–409.
Wasserstrom, Richard A. The Judicial Decision: Toward a Theory of Legal Justifica-
tion. Stanford, Calif.: Stanford University Press, 1961.
———. “Lawyers as Professionals: Some Moral Issues.” Human Rights 5 (1975):
2–15.
———. “The Obligation to Obey the Law.” UCLA Law Review 10 (1963): 780–
807.
Wechsler, Herbert. “Toward Neutral Principles of Constitutional Law.” Harvard
Law Review 73 (1959): 1–35.
Weick, Karl E. Sensemaking in Organizations. Thousand Oaks, Calif.: Sage, 1995.
Weinstein, Jack B. “Every Day Is a Good Day for a Judge to Lay Down His
Professional Life for Justice.” Fordham Urban Law Journal 32 (2004): 131–70.
Wellman, Christopher Heath. “Liberalism, Samaritanism, and Political Obliga-
tion.” Philosophy and Public Affairs 25 (1996): 211–37.
Wenar, Leif. “The Nature of Rights.” Philosophy and Public Affairs 33 (2005): 223–52.
West, Robin. “The Lawless Adjudicator.” Cardozo Law Review 26 (2005): 2253–61.
Westen, Peter. Speaking of Equality: An Analysis of the Rhetorical Force of “Equality”
in Moral and Legal Discourse. Princeton, N.J.: Princeton University Press, 1990.
White, David. “Parker Says He’s Willing to Defy High Court.” Birmingham (Ala.)
News, May 26, 2006, 1C.
Williams, Bernard. Ethics and the Limits of Philosophy. Cambridge, Mass.: Harvard
University Press, 1985.
———. Moral Luck. Cambridge: Cambridge University Press, 1981.
Winston, Ken. “On Treating Like Cases Alike.” California Law Review 62 (1974):
1–39.
Wolf, Arthur P., and William H. Durham, eds. Inbreeding, Incest, and the Incest
Taboo: The State of Knowledge at the Turn of the Century. Stanford, Calif.:
Stanford University Press, 2005.
Wolf, Susan. “Morality and Partiality.” Philosophical Perspectives 6 (1992): 243–59.
Wolff, Robert Paul. In Defense of Anarchism. New York: Harper & Row, 1970.
Woodard, Calvin. “Joseph Story and American Equity.” Washington and Lee Law
Review 45 (1988): 623–44.
Yeazell, Stephen C. “Good Judging and Good Judgment.” Court Review 35 (1998):
8–10.
Zapf, Christian, and Eben Moglen. “Linguistic Indeterminacy and the Rule of Law:
On the Perils of Misunderstanding Wittgenstein.” Georgetown Law Journal 84
(1996): 485–520.
Zimmerman, Michael J.“Intervening Agents and Moral Responsibility.” Philosophical
Quarterly 35 (1985): 347–58.
Zipursky, Benjamin C. “The Model of Social Facts.” Hart’s Postscript: Essays on the
Postscript to the Concept of Law. Ed. Jules Coleman. New York: Oxford University
Press, 2001.
Index

Page numbers in boldface indicate where a term is introduced or defined.

abortion on precedent, 11n45, 280n29

as controversial legal question, 18, on promises, 146n10
80, 87 on pure rules, 115
John T. Noonan’s position on, on renegade officials, 47n55
250–51 on rule-sensitive particularism,
Roman Catholic position on, 109, 124n7
175 on serious rules, 112
Ronald Dworkin’s position on, 89 on settlement function of law, 118n
as subject of checkerboard on threshold deontology, 204n,
legislation, 262–63 210n11
Abrahamson, Leslie W., 59n21 Alexy, Robert, 129n26
absolute principles, 204 Alfini, James J., 62n38, 62n40
absurdity doctrine, 41 Altman, Andrew, 46n53, 84n37,
Ackerman, Bruce, 6n20, 81n23 84n38, 84n39
activism, judicial, 16–17, 50–52 Altman, Scott, 297n13
adaptation, 185–86, 229–31 anarchism, philosophical, 15, 32, 168
additive principles, 203, 208 Anderson, Lisa R., 235n5
adherence (to legal standards), 39 antimiscegenation, 169
affirmative action, 89, 240, 250 apartheid, 9, 169, 239
agency (legal concept), 28 appeals (to higher court), 3, 5, 63, 70,
agent neutrality, 14–15, 182, 203, 310 71, 178
agent-relative principles and, Applbaum, Arthur Isak, 4n7,
208–11 20n8, 97
imperceptible effects and, 221 appraisal reasons. See reasons:
agent relativity, 14, 182, 204, 208, 310 appraisal
aggregative principles, 209, 212 Aquinas, Thomas, 247n26
Alexander, Larry Aristotle, 16, 83n33, 255
on coordination and efficiency, Ashcroft v. Free Speech Coalition, 80n18
120n35 Atiyah, P. S., 144n6
on criminal law, 224n17 attenuation (of reasons), 21
on defiance of rules, 39n15 Augustine, 87
on exclusionary reasons, 113n11 Austin, J. L., 33n55
on formalism, 292–94 Austin, John, 98n17
on gaps, 46n48, 82n31, 83n33, authority
134–35 asymmetry of, 134n45
on indeterminacy, 85n43 judicial, 30–34
on legal principles, 44n44, 89n65 legal, 103–4
on particularism, 94n7 moral, principle of, 32
on posited rules, 43n39 autonomy, 121–22, 137–40

339
340 Index

backward-L (shape of curve), 235 Brown, Darryl K., 42n32

Baer, Harold, Jr., 7 Brown v. Board of Education, 81, 99
Baier, Kurt, 146n8 Brownback, Sam, 6n18
Bales, R. E., 22n16 Budd, Malcolm, 290n42
Balkin, Jack M., 6n20, 81n23, 85n41 Burton, Stephen J.
bandwagon effects, 235 on economic commons analogy,
Banerjee, Abhijit, 235n5 241n19
Barak, Aharon, 46n47 on indeterminacy, 46n53, 84n39,
Barnes v. Glen Theatre, Inc., 245n23 85n42
Baron, Marcia, 23n20 on judicial duties, 57n8
Bassett, Debra Lyn, 37n6, 59n24 on practical reason, 19n1
battered women, 80 on use of extralegal considerations,
Bayles, Michael, 255n5 8n34
Benn, S. I., 255n5 Bush v. Gore, 6, 17n61
Bennett, Jonathan, 105n40 Bushnell, Eleanore, 68n86
Bennis v. Michigan, 79n11 Butler, Paul, 273n12
Bentham, Jeremy, 127
Berger, Raoul, 80n21, 95n10 Calabresi, Guido, 41n24
Bethlehem Steel Co. v. New York State California Commission on Judicial
Labor Relations Board, 41n19 Performance, 65
Bikhchandani, Sushil, 235n5 Caminker, Evan H., 50n63, 127n12
birth control. See contraception candor, judicial, 270–76, 286
Bix, Brian, 84n39 cap-and-trade. See emissions trading
Black, Hugo, 95n10 capital punishment
Blackmun, Harry, 82 as applied to juveniles, 12
Blackstone, William, 86 as controversial legal question, 79,
blame 250
for adhering, 268–69 Justice Blackmun’s position on, 82
for deciding randomly, 266 refusing to enforce, 88
for deviating, 93, 249–52 Roman Catholic position on, 109,
when warranted, 21–22, 218, 224 175
Bloom, Frederic M., 12n47 Cardozo, Benjamin N., 13n52, 57
Bobbitt, Philip, 275n16 Carhart, Gonzales v., 17n61, 80n16
Bork, Robert H., 80n21, 95n10 Carter, Stephen L., 109n51
Bowers v. Hardwick, 277n26 cascades, 235
Boy Scouts of America v. Dale, 17n61 cases
Bradley v. Fisher, 61n33 clear, 8–10
Bradley v. Richmond School Board, easy, 39n14, 108
128n23 gap, 46n48, 133, 134–35, 248–49,
Brand-Ballard, Jeffrey, 136n52, 181n2, 280–82
211n15 hard, 39n14, 90–91
Bratman, Michael, 192–93, 195–97, high-level, 246–48
207n7 impermissible-result, 77–78
Brennan, Samantha, 211n14 legally regulated, 77–78
Brennan, William, 251n31 low-level, 246–48
Brewer, Scott, 290n43 negative-closure, 82
bribery, 36, 54, 116 negative-gap, 82–83, 240, 248,
Bright, Stephen B., 6n18 303
Brink, David O., 22n16, 99n25, 272n7 inevitability of, 86, 188
Broadman, Howard R., 65 John Rawls on, 165
Index 341

optimal-but-not-required-result, Cohen, Mark A., 37n7, 72n104

77–78, 114–15, 144 coherence, 231–32
optimal-result, 77 coin flip, as basis for decision, 52n79,
agent-relative principles and, 260n11, 266, 280
202–7 Coleman, Jules
comparative justice and, 257, 260, on conventionality thesis, 298n18
269, 271 on definition of “positivism,” 99n25,
identifying, 133 99n26
judicial reasons in, 96, 113–14 as inclusive positivist, 44n41, 89n63
mimesis and, 187–201 passim, on indeterminacy, 84n39, 85n42
212 on separability thesis, 98n17
optimal adherence rules and, on social fact thesis, 98n18, 197
220–31 passim collective agency, 194
permissive rule and, 163 collective reasons. See reasons:
permissible-result, 140, 144, 154, collective
177 combined restriction, 229, 233
positive-closure, 82 common morality, 23, 231–32
positive-gap, 82–83 comparative harm functions, 208–10
required-result, 77–78, 114–15, 144 comparative justice, 246, 254–62, 311
suboptimal-but-permissible-result, as argument for restrictive rule, 122
77–78, 170 conditional legal, 258
suboptimal-result, 77 legal, 257, 261
assumptions underlying existence moral, 258–61
of, 83–88 as related to equity, 303
classified, 82–83 complicity, 225–26
suboptimal-rule, 165, 240, 246–49, concept/conception distinction,
303 100n30
Casey, Planned Parenthood v., 80n15 Coney, Amy V., 109n51
Cashman, Edward, 68 conflicts of interest, 54, 116
Catholicism, Roman, 10, 109–10, conform/comply distinction, 159
175–77, 251 congruence condition (of formal
causal limitation, 221 legality), 116–17
causal moderate rule, 293 Conkle, Daniel O., 109n51
checkerboard legislation, 262–64 Connecticut National Bank v. Germain,
Chickasaw Nation v. United States, 51n72
51n72 Connor, John, 69
child molestation, 79. See also sex with conscientious objection, 15
minors consequentialism
Christopher, Russell L., 204n2 act, 147–48, 173, 221
Cicero, 86 causal limitation and, 221
civil disobedience, 15 indirect, 23, 135–36
claim-rights (Hohfeldian), 24n25 prospective arguments and, 182n4
clean hands, 226–27 rule, 135–36
cliffs, 235–36, 287 sophisticated, 15
Code of Conduct of United States status of agent-relative principles
Judges, 36n5, 54n86, 56n1, within, 210
58–61, 62n37, 63 as theory of punishment, 173–74
Cogel, Metin M., 37n7, 72n104 conservatives, 16–18, 80, 82, 275,
cognitive dissonance, 94–95 305–6
cognitive laziness, 73 constitutional theory, 17
342 Index

content independence, 113–14, 118, defection, 190, 220

158–60, 231 DeLay, Tom, 6n18
contraception, 94–95, 272, 274, 282, deontology, 23, 204, 213
306 DePaul, Michael R., 232n29
contractualism, 23, 136–37, 210, 216 deportation, 79
Controlled Substances Act, 3–4, 10 Dershowitz, Alan M., 6n20
conventionalism, strict, 98 desuetude, 41
conventionality thesis, 298 determinacy, 83–85, 112. See also
conversational implicature, 307 indeterminacy
Cook, Walter Wheeler, 302n20 deviation, 5, 39
cooperation, 190–91, 216–17, 220 contrasted with judicial activism,
with evil, 109–10, 175–77 50–52
coordination express, 53
as function of shared intention, 192 from the law, simpliciter, 45
impossibility of, 234, 265 legally unauthorized, 48
possibility of, 238 surreptitious, 213n3, 271, 273, 276,
as reason for rules, 118–20, 126 286
as requiring predictability, 130 tacit, 53
Cooter, Robert D., 71n101 token-selective, 246, 253–54, 258,
Copp, David, 190n28 261–64, 286, 291, 303
Corvino, John, 277n23 type-selective, 244–46, 281–82, 286
Coughlin, John W., 69 deviation credits, 254, 260–61, 269,
Coulter, Ann, 35 274, 295–98
counterfeiting, 76–77 deviation density, 198–99, 228,
courts, collegial, 35 235–37, 304
Cover, Robert M., 5, 9, 27n37, 28 mimetic failure and, 213–14, 235
Crisp, Roger, 297n14 deviation density threshold, 212–14,
critical legal studies, 84–85, 313 310
Cuba, constitution of, 169 criticism of judiciary and, 288
Cullity, Garrett, 190n27 deviation rate below or above,
cumulative harm. See harm, cumulative 217–20
equity and, 304
Dagger, Richard, 76n7 Griswold v. Connecticut and, 273–74
Dalton, Clare, 84n38 guidance rules and, 234–41
Damasio, Antonio R., 37n10 precedent and, 281
D’Amato, Anthony, 84n40 rebuttable presumption and, 243–44
Dan-Cohen, Meir, 123n4, 159n5, deviation rate, 198, 228–29
272n6 optimal average, 216–19, 229, 233,
Dancy, Jonathan, 21n10, 93n6, 113, 237
297n14 optimal collective, 214
Daniels, Norman, 232n29 disagreement, moral, 75, 118, 262.
Darwall, Stephen L., 14n54, 210n13 See also pluralism, reasonable
Daschle, Tom, 6 discipline, judicial, 52n75, 61–69
De Sousa, Ronald, 37n10 informal, 71
death penalty. See capital punishment discretion, 50, 69, 90
Death with Dignity Act (Oregon), 259 abuse of, 63, 260n11
decision patterns, 212 contrasted with deviation, 42–43,
decision rules, 123–24 46–48
default rule, 265, 267, 279–80, 311 in criminal sentencing, 286
unreflective conformity to, 289–91 delegated, 47
Index 343

deviational, 42–43, 90n67 Dyzenhaus, David, 9, 90–91, 97–98,

objective, 46, 47 128n24
optimal results and, 76, 87, 104,
153 Easterbrook, Frank, 263n21, 292n3
disqualification. See recusal Edlin, Douglas E.
distinguishing (cases), 12, 41–42, on epistemic threshold, 250n29
48–49, 274–78, 283 on judicial role, 141n58
divorce, 109–10, 175 on obsolete laws, 41n24
Dole, Robert, 7 on unjust laws, 43n37, 43n38, 87–88
double effect, doctrine of, 105–6, 205 Edmundson, William A., 15n56,
Douglas, William O., 272–73, 306 103n39, 158n1, 160n6
Drahos, Peter, 84n39 Edwards, Harry T., 3n4
Dred Scott v. Sandford, 7 efficiency, 119–20, 131–32
Dresden, bombing of, 225, 227 Eggleston, Ben, 271n5, 272n7
Dubber, Markus Dirk, 28n37 Ellis, Anthony, 210n11
duties Ely, John Hart, 81n24
of fidelity (promise-keeping), emissions trading, 241–42
144–46 emotion, 37n10, 295
of judicial office, 36, 56–57, 58–61 enablers (of reasons), 21, 113, 154,
natural, 23 296–97
of nonmaleficence, 23, 26, 308 Endicott, Timothy A. O., 116, 256n6
samaritan, 24, 147, 167 Enoch, David, 211n16
duty to obey the law, 32, 103, 157–60 entrenchment (of reasons), 115, 126
consent argument for, 161 environmental policy, 250
fair play argument for, 161–64 epistemic threshold. See threshold,
gratitude argument for, 164 epistemic
natural duty argument for, 165–67 Epstein, Richard A., 6n20, 7
samaritan argument for, 167 equal protection, 6, 17, 81, 128, 275
duty to rescue. See samaritanism equilibrium, 235, 286–89
Duxbury, Neil, 46n52, 267n35 equity, 16, 302–4
Dworkin, Ronald error argument, 120, 132–35
on Bush v. Gore, 6n20 Eskridge, William N., Jr., 41n24
on checkerboard laws, 262–64 esoteric rules, 271–72, 293
on comparative justice, 256n6 establishment clause, 6
on concept/conception distinction, Estreicher, Samuel, 5n11, 6n17,
100n30 41n25, 129n28
on controversial legal questions, 89 euthanasia, 89. See also suicide:
on duty to obey, 158n2 assisted
on force of law, 85–86 evidence, exclusion of, 79
on hard cases, 39n14, 91 evidentiary standards, 250–52
on integrity, 128, 311 evil, cooperation with, 109–10, 175
on judicial candor, 273n12 exclusionary reasons. See reasons:
on judicial discretion, 46n47 exclusionary
on legal pragmatism, 115n17 excuses, 33n55
on locality, 40n18 expectations, epistemically reasonable,
on nonenforcement, 8–10, 89n64, 151–52
89n66, 92, 127–28, 239n13
on principles, 40n17, 44–45, 47 fair share argument, 215–20
on strict conventionalism, 98n23 fairness, 122. See also comparative
on wealth redistribution, 127–28 justice
344 Index

Fallon, Richard H., Jr., 100n29 Ginn, Martha Humphries, 71n100

Farley, John J., III, 302n20 Ginsburg, Ruth Bader, 251
Farnsworth, Ward, 6n20 Glassroth v. Moore, 80n19
Farrell, Daniel M., 25n30 Glover, Jonathan, 222n12
favorers (of reasons), 21, 297 Goldman, Alan H., 15n57, 310
Feinberg, Joel, 26n32, 207n7, 255n5 comparative justice argument
fellatio, 79 (Kantian constraint), 254–55, 258
Ferzan, Kimberly Kessler, 224n17 defense of restrictive rule, 183
final moderate rule, 243, 254, 267, epistemic argument, 234, 238
279–80 on error, 134
First Amendment, 6, 80, 245 on judges, 184, 239
first moderate rule, 238–39 on moderate rules, 184
first priority rule, 243–47, 267 on optimal decision pattern, 212,
Fletcher, George P., 27n36 234, 236
Flikschuh, Katrin, 24n28 on predictability, 130n35,
force, judicial use of, 27–30, 95–96, 131n36
237, 268 on prisoner’s dilemma, 190–92
Forer, Lois, 176–77 on professional ethics, 4n7
formal equality, 255, 259–61, 264 on prudential reasoning, 238
formal justice. See comparative justice on randomizing, 265
formal legality, 100–102, 116–18 on rationality of rule compliance,
formalism, legal, 15, 85–86, 183, 215
292–94, 311 on serious rules, 112n4, 292–93
fornication, 247, 276–77 on suboptimal results, 120n35,
Frankena, William K., 255n5 120n36
Fraser, David, 169n34 on systemic effects, 182–85
free riding, 149–50, 215–19, 228 Goodin, Robert E., 211n16
free speech, 80, 245 Goodwin, Barbara, 267n35
Fried, Charles, 60n28, 273n11, Graber, Mark A., 9n38
290n43 Gray, Cynthia, 63n50, 66n75, 68n84
Friedman, Barry, 17 Green, Leslie, 159
Fugitive Slave Act, 10, 88–89 Greenawalt, Kent
Fuller, Lon L., 9, 52n74, 101n31 on indeterminacy, 84n39
Furman v. Georgia, 82 on institutions of amelioration,
268n37
Gandhi, Mahatma, 176–77 on judicial discretion, 90n70
gaps (in the law), 46 on judicial nullification, 6n17, 127,
Gardner, John, 46n50 128, 129
Garvey, John H., 109n51, 110n56 on justification versus excuse, 33n55
gay men. See sexual orientation on political obligation, 159n5
Gee, Thomas, 176–77 Greenberg, Mark, 44n41
genocide, 9, 93 Grice, Paul, 307n29
Gert, Bernard, 144n6 Griffen, Wendell L., 109n51
Gettier, Edmund L., 83n32 Grimes, Warren S., 68n86, 68n87
Geyh, Charles Gardner, 68n88, 71n102 Griswold v. Connecticut, 94–95,
Gilbert, Margaret 272–74, 282, 284
on collective agency, 192, 194n41, group adherence argument, 191–92
194n42, 196n44 Group O, 187, 239–41, 310
on political obligation, 158n2, average deviation rate of, 228–29
161n10 maintaining equilibrium, 235–37
Index 345

optimal adherence rules for, 212–17, Hohfeld, Wesley Newcomb, 24n25

233 Holder, Eric H., Jr., 3n1
reasons to adhere, 191–92, 202–11 Holt, Charles A., 235n5
group obligation, 191 homosexuality. See sexual orientation
group restriction, 213 Honoré, Tony, 160n7, 170n35, 207n7
guidance reasons. See reasons: Hooker, Brad, 135–36
guidance Horton, Elizabeth M., 284n36
gun control, 80, 185 Horwitz, Morton J., 100n28, 257n7
Gunther, Gerald, 60n28 Hurd, Heidi M., 83n33, 146n8
Hurley, Paul, 210n13
habituated view (of official use of Hurley, Susan, 23n19
force), 31–34 Husak, Douglas, 174n40
Haley, John O., 68n89 Hutchinson, Allan C., 100n28
Hamburger, Philip, 57n8 Hutto v. Davis, 34n59
Hampton, Jean, 190n27, 235n4 hypersensitivity, 190
Hand, Learned, 81n24, 81n28 hyposensitivity, 198
Hardimon, Michael O., 4n7 hypothetical cases
harm, cumulative, 183 Adams/Bonn/Carlo, State v., 259–60
Harris, John, 267n35 bandit and villagers, 222, 227
Hart, H. L. A. bank and borrower, 185–86
on causation, 207n7 Bruce (squirrel kicker), 146, 153
on comparative justice, 256n6 Chester (pedophile), 172–74
on content independence, 113 churchgoing tax, 101–2
on conventionality thesis, 298 Cordelia and Goneril (sisters),
on formalism, 85–86 299–302
on judicial discretion, 8n35, 46n47 Dan (supermarket), 224, 227
on political obligation, 158n2 Fred and George (shoplifters),
on rules, 98n20 253–54
on sanctions, 61 lazy assassin, 149
on secondary rules, 56n2 Leo (sexual harasser), 151
on unjust laws, 9 Marine sniper, 22
Hasnas, John, 100n28 Ned (island), 218–19
Hatch, Orrin, 6n18 Rachel’s grandmother’s terrorist
hate speech, 245, 290 boyfriend, 152
Heiple, James, 68–69 rerouting tax payments to charity
Heller, District of Columbia v., 17n61, (Alan H. Goldman), 191, 215,
80n20 217, 220, 225, 234, 265–66
Hellman, Deborah, 271n3 Russell, Mrs. (medical marijuana), 3
Henchey v. City of Chicago, 90n69 Susan, J. (nude dancing/hate
Henderson, Lynn, 28n37 speech), 245–46
Henkin, Louis, 95n10 Yasmin and Rafael (landlord/
Henningsen v. Bloomfield Motors, 44 tenant), 78–79, 87, 153–54,
herd behavior, 235 183, 200–201, 229, 296–97
heroin, 247
Heyd, David, 23n18 Idelman, Scott C., 109n51, 271n3
Higgins, Richard S., 71n100, 72n105 Igneski, Violetta, 206n6
Himma, Kenneth Einar, 44n41, 46n47, impeachment, 36, 62, 68–70, 312
298n18 of federal judges, 68
Hirshleifer, David, 235n5 of state judges, 68–70
Hoffer, Peter Charles, 302n20 imperceptible effects, 221–22
346 Index

incentives, judicial, 36, 54–55, 70–73, Kadish, Mortimer R. and Sanford H.

92, 287 (co-authors), 42, 47, 87, 130
incest, 276–79 Kagan, Shelly, 21n9, 105n40
inclusive positivism. See positivism, Kamm, F. M., 105n40, 107, 206
legal: inclusive Kant, Immanuel, 24n28, 254
independence, judicial, 59 Kantian constraint, 254
indeterminacy, 46, 84–85, 313. See Kantianism, 210
also determinacy Kaplow, Louis, 101n33, 211n16
indirect consequentialism. See Karlan, Pamela S., 53n81
consequentialism, indirect Katz, Leo, 204n2
individual adherence argument, Kauper, Paul G., 95n10
188–89 Kelo v. City of New London, 7
individual policy, 233, 237 Kelsen, Hans, 46n51
individual restrictions Kennedy, Anthony, 278
first, 216 Kennedy, Duncan, 84n40, 101n33,
second, 219 123n5
third, 229 Klarman, Michael J., 6n20
influence, judicial, 37, 162 Kline, J. Anthony, 53, 64n64, 65,
Ingram, Peter, 256n6 65n73, 140n57
injustices (mandated/nonmandated), Klosko, George, 158n2, 162, 190n27
168–70 Kmiec, Keenan, 50–52
integrity, 262–64, 311 Kohlberg, Lawrence, 31n49
intend/forsee distinction, 203, 205 Kornhauser, Lewis A., 11n44, 72n106
intensifying (of reasons), 21 Korsgaard, Christine M., 19n4
intentional states, 192 Kramer, Larry D., 41n22
intentions, 104–7, 205 Kramer, Matthew, 44n41, 89n63,
shared, 192–98 298n18
intransitivity, 199–201 Kress, Ken, 44n44, 84n37, 84n38,
INUS condition, judicial oath-taking 84n39, 89n65
as, 155–56 Kretzmann, Norman, 129n27
invitee/licensee distinction, 48–49 Kroger, John R., 302n20
Isaacs, Tracy, 194n42 Kronman, Anthony, 290n43
Kuran, Timur, 235n5
Jackson, Frank, 223n14 Kutz, Christopher, 190, 197, 225
Jackson v. Lykes Bros. S.S. Co., No. 575,
60n27 Lamond, Grant, 30n47
Jacobson, Arthur J., 284n36 Landers, Scott, 84n39
Japanese-Americans, internment of, Landes, William M., 37n8
169 landlord and tenant, law of, 26n33
judges (anterior/posterior), 186 law, simpliciter, 11, 40
judicial activism. See activism, judicial law, theory of, 44
judicial authority. See authority: lawmakers, 111n2, 270
judicial lawmaking, judicial, 141
Judicial Conference of the United Lawrence v. Texas, 275–79, 284
States, 59 lawsuits against judges, 62
judicial misconduct. See misconduct, layered pyramid, 246
judicial Lefkowitz, David, 151n21
judicial nullification, 5 legal content, theory of, 44, 85–86
judicial review (of legislation), 41 legal error, 3, 5, 7, 62–68, 276
jury nullification, 6n17, 42 egregious, 64, 67–68
Index 347

legal positivism. See positivism, legal on predictability, 119n33

legal pragmatism. See pragmatism, on threats against judges, 268n38
legal
legalistic justification, doctrine of, MacAdam, James I., 255n5
129–30 MacCormick, Neil, 57n9, 71, 125–26,
legitimacy, 121–22, 137–40 150n18, 255n5
Leiter, Brian, 38n12, 84n39, 85n42, Macey, Jonathan R., 37n9
290n43 Mackie, J. L., 155n26, 160n7, 160n8,
lesbians. See sexual orientation 236n9
Levenbook, Barbara B., 280n29 mala in se offenses. See offenses: mala
Levinson, Sanford, 109n51 in se
Lewis, David, 193n35, 196n44, Manning, John F., 41n25, 302n20
267n35 Marbury v. Madison, 41n21
lex iniusta non est lex, 87, 129 marijuana, 3–4, 34n59, 80
lexical priority, 251, 267 Marmor, Andrei, 100n29, 257n8
Li, Xiaorong, 169n31 Marshall, Thurgood, 35
liability, judicial, 62 Mason, Elinor, 147n15
liberals, 17–18, 80, 275, 305–6 McCarthy, David, 223n14
local applicability, 40 McCarthy, Nancy, 65n71
Locke, John, 24n28, 25, 25n30 McCulloch v. Maryland, 41n21
Lockyer v. Andrade, 79n10 McDermott, Daniel, 25n30
Lode, Eric, 199n55 McDowell, Gary, 302n20
Lopez, United States v., 17n61 McHugh, Kenneth R., 69
Luban, David, 4n7, 16n58, 20n8 McIntyre, Alison, 105n40
Lubet, Steven McKeever, Sean, 93n6
on bias, 66n76 McKerlie, Dennis, 223n14
on Circuit Judicial Councils, 60n26 McMahan, Jeff, 24n27
on decision by coin flip, 266n31 McMahon, Christopher, 190, 216
on decisional conduct, 61–62 means (to an end), 106–8, 206
on disciplinary complaints, 63n49 mediated/unmediated distinction, 204,
on egregious error, 67 207
on Judge Heiple case, 69n91 mental states, judicial, 52–54, 268
on Judge Kline case, 65–66 Mian, Emran, 113n11
on patterns of error, 66n74 Miceli, Thomas J., 37n7, 72n104
on willful refusal to follow law, Miller, Arthur R., 54n87
66n78 Miller, David, 158n2
Luce, R. Duncan, 264n26 Miller, Kaarlo, 193n37
Lyons, David Miller, Richard W., 206n6
on correlativity of rights and duties, Miller, William Ian, 26n32
22n18 mimesis, 186–88, 213, 229–31
on density, 199n54 mimetic failure, 187
on deviation, 129–30 effect of judicial candor on, 271
on duty to obey the law, 159n5 empirical assumptions about, 212
on formal justice, 256n6, 257n8 excessive, 264
on gap cases, 83n33 excessive rule revision as form of,
on hard cases, 39n14 281
on judicial commitment, 142n1 as reason to adhere, 188, 202
on justification of coercion, 30 vulnerability to, 198
on moral predicament of judges, minimalism, judicial, 283–85
108 minimum-purchase policy, 242
348 Index

minorities, discrete and insular, 264 of federal judges, 142

mirages, legal, 290 as INUS condition, 155–56
misconduct, judicial, 36, 62 judicial, 10, 60–62, 161, 226
misdemeanors, 253–54 special weight of, 154–55
mixed strategies, 264 of state judges, 142n4
Modak-Turan, Mark C., 109n51 obedience, 123
Model Code of Judicial Conduct, 36n5, objective reasons. See reasons:
54n86, 56n4, 58–61, 62–64 objective
Model Penal Code, 224n15 obligation, political, 14–15, 160–61
criminal solicitation, 28n41 offenses
self-defense, 26n34 mala in se, 31n51, 76, 140, 171–72
Moglen, Eben, 84n39 mala prohibita, 76, 171–72
Montesquieu, 101n31 O’Hara, Erin, 72n104
Moore, Michael S., 43n40, 86–87, Olowofoyeku, Abimola A., 62n39
113n11, 204n2, 207n7 open texture (of rules), 46
Morrison, United States v., 17n61 opinions, unpublished, 285
Morrow v. Hood Communications, Inc., optimal results, 75
53 optimizing collective agents, 194–97,
Mosley, United States v., 47n55 195
Murphy, Liam B., 181n1 Oregon Waste System, Inc. v.
Murphy, Mark C., 9n36, 86n52, Department of Environmental
158n2, 161n10 Quality, 3n2
original position (John Rawls), 166
Nagel, Thomas, 14n55, 105n41, 204n2 Orsi, Francesco, 206n6
naive view (of official use of force), Osborn, Rebekah L., 109n51
31–34 Otsuka, Michael, 190n27
natural executive right. See rights:
natural executive paired privileges (Hohfeldian), 24n26
natural lawyers, 8–9, 15, 86, 247 Panglossians, 86–88, 272, 311–12
Nazism, 9, 169, 239 Pareto principle, 209–10, 266
Neary v. Regents of the University of Parfit, Derek, 14n55, 222n13, 271n5,
California, 53 286n37
neutrality, political, 305–7 Parker, Stephen, 84n39
New Jersey, Supreme Court of, 69 Parker, Tom, 12, 53–54
Newdow v. U.S. Congress, 6 Partial Birth Abortion Ban Act, 80n16
Newman, Jon O., 7n29 particularism, 115n17, 115n18, 183
Nichols, Shaun, 37n10 causal rule-sensitive, 293
no-minimum policy, 241–42 legal, 93–94
nonideal theory, 136, 181, 217, 272 moral, 93–94, 183
Noonan, John T., Jr., 28n37, 250–51 rule-sensitive, 124, 134n46, 240,
novus actus interveniens, 207–8 292–93
Nozick, Robert, 25n28, 25n30, 162, Paulsen, Michael Stokes, 86n49,
204n2 127n12
nude dancing, 245, 290 Paulson, Bonnie Litschewski, 129n26
nullum crimen, nulla poena, sine lege, Paulson, Stanley L., 129n26
171–74 Payne v. Tennessee, 41n20
Nussbaum, Martha C., 37n10 Peirce, C. S., 246n25
Peller, Gary, 84n38
oath arguments, 143, 145, 147 Pelosi, Nancy, 35
oaths, 142–43 Penzell, Abigail, 33n56
Index 349

Perry, Stephen R., 22n15, 44n41, principle of permissible harm

113n11, 250n29 (F. M. Kamm), 107–8, 206
Peters, R. S., 255n5 principles, legal, 44, 46, 47, 86, 89–90
Pettit, Philip, 14n55, 23n20, 194n41, priority rules
194n42, 210n13 first, 243–47, 267
Plato, 102n35, 103 second, 247–48, 253, 264–65, 267
pleasure (of adhering to law), 72, 265 prisoner’s dilemma, 189
Pledge of Allegiance, 6 moral-moral, 190–91, 265
pluralism, reasonable, 75n1, 79, 249, pro tanto reasons. See reasons: pro tanto
305 promises
politicians, 6, 69, 93, 289 to act immorally, 146–54
pollution, 241–42 forming shared intentions, 196
pornography, 80 in judicial oath, 61, 105, 142–56
positivism, legal, 98–99 promissory obligation, theories of,
exclusive, 85, 298, 301–2, 311 147–54
inclusive, 43, 85, 89, 298 act-consequentialist, 147–48
presumptive, 134n46, 240, 292–93 deflationary, 147–48
as theory of legal validity, 97–98 free-rider, 149–50
Posner, Eric, 270n2 nonpractice, reliance-based, 150–54
Posner, Richard A. rule-consequentialist, 148–49
on desire to be cited, 72 prostitution, 250, 276–77
on deviation, 5n10 proximity principle, 206–7
on judicial incentives, 37n7, 70 Pryor, William H., Jr., 109–10, 175–77
as legal pragmatist, 304–5 pseudorules, 111–12
opinion in Khan v. State Oil Co., public goods, 191, 215, 220, 225
93n5 public officials, 165
on pleasure of judging, 72, 167, 265 demands of law upon, 170
on precedent, 37n8 reasons of, 20
on reputation, 71n101 rule of recognition and, 125
on reversal, 71n100 use of force by, 27–34, 96–97
Postema, Gerald J., 16n58, 83n33, public schools, funding of, 79
127n15, 239n13, 303n21 publicity condition, 272, 280, 293
practical difference thesis, 297–302 punishment, 25–26, 102–3, 171–74
practical reasons. See reasons: practical
pragmatism, legal, 115n17, 304–5 Quantum Leap (television series),
precedent 156n27
disregarding, 245 Quinn, Warren, 25n30, 105n41
horizontal, 11–12, 48–49, 305
judicial activism and, 50–51 Rabin, Edward H., 26n33
mimesis and, 186 race discrimination, 169, 262, 264
standard, 281 Radbruch, Gustav, 9, 129n26
vertical, 11–12, 48–49, 245, 305 Radbruch formula, 129, 239–40
predictability, 118–19, 130–31, 183 Radin, Margaret Jane, 287n39
preemption (of state law), 3, 41–42 Raich, Gonzales v., 3n3, 80n17
preponderance of the evidence, 250 Raiffa, Howard, 264n26
presumptive positivism. See Railton, Peter, 21n14, 223n14, 271n5
positivism, legal: presumptive random decisions, 37
Prichard, H. A., 149n16 as blameworthy, 233
Prima Paint Corp. v. Flood & Conklin as permitted by default rule,
Mfg. Co., 60n27 264–67, 279, 286, 289–91
350 Index

rationality, of judges, 37n10, 281–82 preferential, 19–20

R.A.V. v. City of St. Paul, 245n22 prima facie, 21n9
Rawls, John private, 19–20, 144–45, 166–67,
on concept/conception distinction, 177
100n30 to deviate, 54–55, 115
on fact of reasonable pluralism, 75 pro tanto, 21
on moral significance of prudential, 19–20, 125–26, 271, 288
consequences, 182n4 residual, 154
on natural duty, 164–66, 170 role, 20
on nonideal theory, 181n1 subjective, 21–22
on political obligation, 158n2, theoretical, 19
160n7, 162, 170 rebuttable presumption, 239–43, 292
on promising, 149n16 recourse roles, 42–43
on publicity condition, 272n7 recusal
Raz, Joseph to avoid difficult cases, 36–37
on authority, 124–25, 132–35 for conflict of interest, 5, 54–57
on conform/comply distinction, failure as basis for discipline, 62n42
159 as option, 92
on definition of “positivism,” 99n26 reasons against, 113, 177–78
on duty to obey the law, 103n37, for religious reasons, 29n43, 110,
159–60 174–78
on exclusionary reasons, 21n13, Regan, Donald H., 113n11
113–15 regulation (of decisions by legal
as exclusive positivist, 85n46 standards), 39–40, 77, 92,
on gaps in the law, 46n48, 46n49 299–300
on legal duties of judges, 57n7 Reibetanz (now Moreau), Sophia,
on norm-applying institutions, 105n40
28n38 reliance, 118–19, 123, 126, 130,
on promising, 146n8 150–54
on rule of law, 102 religion, free exercise of, 80
on sources thesis, 98n19 religious beliefs (as basis for
realism, legal, 38 deviation), 10, 35, 88, 108–10
reasonable doubt, 42n32, 250–52 religious persecution, 9
reasonable suspicion, 7 remote applicability, 40
reasons remote/local distinction, 204–6
all-things-considered, 21 removal (from office), 36, 68–70, 271,
appraisal, 22, 214 312
collective, 191–98 rent control, 229–30
discounting, 299–302 repugnant conclusions, 222
exclusionary, 21, 126, 134n46, 144, reputation
156 as incentive to adhere, 71, 106, 189,
Joseph Raz on, 113–15, 117 265, 268–69, 288–89
guidance, 22 as incentive to perform judicial
impartial, 20, 115, 119, 144 duties, 37
legal, 20, 23 resignation, judicial, 57n8, 89n64,
moral, 20, 126–41 92–93, 144, 174–78
natural, 23, 78, 104, 182, 226 calls for, 7, 68
objective, 21, 185, 215, 221, 223 resources, decisional, 131–32
of partiality, 20, 119 respect (for lawmakers), 137–40
practical, 19, 197 restitution, 24–26, 31–32, 103–4
Index 351

restrictive rule, 114–15, 231–32, crossing threshold as destroying,

309–12 236, 243, 289, 313
Alan Goldman on, 183–85, 236, formal legality as conception of,
292 100–102, 116–17
as codified, 117 judicial candor and, 273
constraint model and, 294–97 as reason to adhere, 11, 105–6, 108,
formalism and, 292–94 188, 206, 302
Griswold v. Connecticut and, 273–76 as reducing unpredictability, 280
as guidance rule, 237–40 restrictive rule and, 287
legal pragmatism and, 305 selective optimization as upholding,
promulgating, 119–22 293
psychological burdens of, 285–91 traditional conception of, 179, 313
reasons to obey, 126–41, 145–78, 268 rule worship, 135n49
as social norm, 287 rules
result-oriented judging, 50–52 adherence
retributivism, 25, 78, 171–74 codified, 58–61
reversal Group O and, 203
of horizontal precedent, 49–50, promulgation of, 123–24
70–71 as serious rules, 111–12
as incentive to adhere, 70–72, 288 adjudication, 56–57, 136–37,
of lower court decision. 143–45, 156–58, 162–64
See appeals appraisal, 123–24, 219, 233–34, 237
of vertical precedent, 49 closure, 46
Revesz, Richard L., 5n11, 129n28 conduct, 57n11, 118, 123, 132, 158,
Reynolds, William L., 284n36 272n6
rhetoric, judicial, 138, 311 decision. See decision rules
Richardson, Henry S., 49n59, 291n44 guidance, 123–24, 219, 233–52,
Richman, William M., 284n36 267, 311
Ridge, Michael, 14n55, 93n6 impure, 115
Riggs v. Palmer, 44 legal, 40, 40
rights determinacy of, 84
constitutional, 51, 94, 276 deviation from, 42–45, 130
as correlated with duties, 22 judicial discretion and, 90
of justice, 23–27, 30, 32–33, 308 practical difference thesis and,
legal, 26, 30, 78, 150, 153, 273 297–302
natural, 23–25, 32–33, 308 revising, 280–81
natural executive, 25, 31 as serious rules, 292–94
samaritan, 23, 24, 26, 146–47, 308 mandatory, 57, 85, 113–14, 159
risk, 23, 210, 223–28, 234–36, 273 moderate, 115
Rizzo, Mario J., 199–200 nonpermissive, 115, 309–10
role morality, 4 agent-relative principles and,
roles, natural, 97 205–10
Romer v. Evans, 274–79, 284 contractualism and, 136–37
Roper v. Simmons, 12n49 judicial oath and, 155–56
Ross, W. D., 21n9 systemic effects and, 181–92,
Roth, Abraham Sesshu, 194n42 230
Rovane, Carol, 192, 194n41 permissive, 115
Rubin, Paul H., 71n100, 72n105 contrasted with selective
rule guidance, 295–301 optimization, 292–96
rule of law, 4, 15, 43, 81 as guidance rule, 237
352 Index

rules (continued) on exclusionary reasons, 113n11,

in hyposensitive systems, 198 114–15
judicial duty to obey, 163–68 on gap cases, 46n49
judicial oath and, 144–47 on judicial incentives, 37n7
reasons to obey, 119–26, 132–37 on legal standards, 44n43
pure, 115 on locality, 40n18
restrictive. See restrictive rule on mandatory rules, 57n10
serious, 111–12, 124, 131n36, on negative-gap cases, 83n33
115n15, 240, 292–94 on particularism, 94n7
of thumb, 124, 131 on precedent, 280n29
rule-sensitive particularism. on pressure to deviate, 179
See particularism: rule-sensitive on presumptive positivism, 240n17,
Rumble, Wilfred E., 98n17 292
Russell, Mrs. (fictional marijuana user), on rules, 15n57, 48, 124
10 on rule-sensitive particularism, 124,
293
Sachs, Stephen E., 146n13 on rule-worship, 135n49
Sagoff, Mark, 242n20 on slippery slopes, 199n55
Salmon, Marylynn, 169n32 on unwritten rules, 270n1
samaritanism, 167. See also rights: Scheffler, Samuel, 14n55, 105n40
samaritan Schmidtz, David, 206n6
San Antonio v. Rodriguez, 79n9 Searle, John, 192
sanctions, for deviating, 58, 61–62, 67, segregation, racial, 99, 169
70, 268, 287. See also discipline, selective optimization, 267–75,
judicial 279–307 passim, 311–13
Sankar, Sambhav N., 53n81, 57n9, internalizing, 285–86
61n32, 64n64, 71n102 self help (legal concept), 24, 26–27
Santa Fe Independent School District v. self-defense, 24, 27, 31, 175
Doe, 80n14 self-effacing theories, 286
Saphire, Richard B., 109n51 Sensenbrenner, F. James, Jr., 7
Sarat, Austin, 28n37 sentencing, mandatory minima, 47,
Sartorius, Rolf, 83n33, 159n5 79, 230
Sarver, Tammy A., 71n100 separability thesis, 97
Scalia, Antonin separation of powers, 10, 140–41
on creation of new rights, 51n71 settlement function (of law), 118,
on flogging statute, 128–29 293
on formalism and textualism, 292n3 sex with minors, 172–74. See also child
opinions in sexual orientation cases, molestation
275–77, 279, 284 sexual orientation, 230, 274–79, 282
on wishful thinking by judges, 86 Shaman, Jeffrey, 61–64, 66
Scanlon, T. M., 23n20, 136, 143n5, Shapiro, David, 271n3
150–54, 216 Shapiro, Martin, 273n12
Schapiro, Tamar, 181n1 Shapiro, Scott J.
Schauer, Frederick adapting Michael Bratman’s work,
on arguments for rules, 119–21, 197
133n40 on conservatives, 82
on asymmetry of authority, 134n45 as exclusive positivist, 85n46
on definition of “positivism,” 99n25 on indeterminacy, 84n36
on easy cases, 39n14 on practical difference thesis,
on error, 132n38 297–302
Index 353

on psychology of deviation, 71n103 Smith, Michael, 223n14

on rule following, 294–97 Smith, Steven D., 4n9, 6n17, 127
shared intentions. See intentions, shared social fact thesis, 97, 197
shared values, community of, 138 social norms, 214, 270, 276, 287–88
Shavell, Steven, 211n16 sodomy, 109, 275–79, 282, 284
Sher, George, 181n1 Solum, Lawrence B., 84n35, 84n37,
Sherrer, Hans, 28n37 84n39, 109n51, 303–4
Sherwin, Emily Songer, Donald R., 71n100
on coordination and efficiency, Soper, Philip, 126, 138
120n35 sources thesis, 97
on defiance of rules, 39n15 Sowell, Thomas, 6n18
on formalism, 292–94 stability, 120, 183
on gaps, 83 n33, 134–35 stare decisis, 127n15, 213, 280–81,
on indeterminacy, 85 n43 287. See also precedent
on legal principles, 89 n65 Stark, Cynthia A., 22n16
on particularism, 94 n7 state of nature, 23–24, 26, 31–33, 167,
on posited rules, 43n39 308
on promises, 146 n10 State Oil Co. v. Khan, 49n62
on pure rules, 115 States v. Schooner Peggy, 60n27
on renegade officials, 47n55 statutory construction, 51, 111
on rule-sensitive particularism, 124n7 Stavropoulos, Nicos, 44n41
on serious rules, 112 step goods, 235n4
on settlement function of law, 118n Stern, Gerald, 52n75, 63–64, 66n77
on unwritten rules, 270n1 Stewart, Potter, 60n27, 95n10, 272n8
Shklar, Judith N., 100n28, 100n29, Strauss, David A., 80n22, 81n27
101n31 Stump v. Sparkman, 62n38
side-effect/aspect distinction, 107–8, subjective reasons. See reasons:
204, 206 subjective
Sidgwick, Henry, 271n5 suboptimal results, 75
Siegel, Neil S., 18n62 examples of, 79–80
Simmons, A. John, 25n30, 32n52, subplans, 195n43
103n37, 159, 162 suicide, 92n2
Simon, Dan, 13n52 assisted, 109, 250, 259. See also
Simon, William H., 16n58, 81n27 euthanasia
Singer, Joseph, 84n38 Suleiman, Ramzi, 235n4
Singer, Peter, 206n6 Sullivan, Kathleen M., 101n33
single privileges (Hohfeldian), 24n26 Sulmasy, Daniel P., 143n5, 155n24
slavery summary judgment, 54n87
rulings on, 27–28, 88–89, 303 Summers, Robert S., 43n38, 101n31,
as unjust, 9–10, 82–83, 93, 169, 127
239–40 Sunstein, Cass R.
slippery-slope arguments, 199–201 on agent-relative principles, 211n16
Slote, Michael A., 23n20 on availability cascades, 235n5,
Smith, M. B. E. 236n6
on cheating by appellate courts, on Bush v. Gore, 6n20
5n10, 83n34 on critical legal studies, 85n41
on duty to obey the law, 103n37, 159 on discrimination, 278n27
on judicial nullification, 4n9, 5n17, judicial minimalism of, 283–85
127n12 on moral heuristics, 210n12
Smith, Matthew Noah, 197 Sypnowich, Christine, 100n28
354 Index

systemic effects, 181–85 Radbruch formula on, 239

agent-relative principles and, 203–8 unjust war, 169, 175
Alan H. Goldman on, 187–93 utilitarianism, 236
complicity and, 225–26
as imperceptible, 220–21 Van Alstyne, William, 256n6
of rule revision, 281 Velleman, J. David, 19n4, 192–94, 196
Vermeule, Adrian, 133n42, 211n16
Talley, Eric, 236n6, 297n13 Vila, Marisa Iglesias, 46n47
Tamanaha, Brian Z. violence (of law), 28
on Brown v. Board of Education, 81, 99 virtue ethics, 23, 210
on indeterminacy, 85, 111n2 Volcansek, Mary L., 68n86
on rule of law, 100n28, 116–17, Volokh, Eugene, 199n55
287n38
Temkin, Larry S., 222n13 Wald, Patricia M., 12n52, 28n37
Ten Commandments, 80 Waldron, Jeremy
Tennessee Senate, 69 on deviation, 127n18
terror bomber (in double effect on duty to aid, 206n6
debates), 106 on natural rights, 25n28
terrorists, 22 opposition to judicial review, 41n22
Thayer, James B., 50n68 on rule of law, 102n34
Thomas, Clarence, 35 Walen, Alec, 115n19
Thomson, Judith Jarvis, 24n27, 27n36, Walker, A. D. M., 158n2, 164n16
31n48, 223n14 Walton, Douglas, 199n55, 200n57
threats, blocking of, 29–30, 95 Waluchow, W. J., 44n41, 84n39,
threshold. See deviation density 89n63, 298n18
threshold Warren, Earl, 17, 80–81
threshold, epistemic, 88, 250n29 Wasserstrom, Richard A., 16n58,
threshold principles, 204, 208–11 78n8, 159
treating like cases alike, 239, 246, Wechsler, Herbert, 81n24, 81n28
255–60. See also comparative Weick, Karl E., 73n110
justice Weinstein, Jack B., 5n17, 60n28
triggers, 223–24, 226–28, 235 Welch, Ivo, 235n5
Tuomela, Raimo, 192–94 Wellman, Christopher Heath, 158n2,
Tushnet, Mark, 41n22, 84n38 167n29
Tyler, Tom R., 31n50 Wenar, Leif, 24n25, 24n26, 30n45
type-selective deviation. See deviation: West, Robin, 127
type-selective Westen, Peter, 256n6
type/token distinction, 246n25 Whitman, Douglas Glen, 199–200
Williams, Bernard, 33n56, 272n7
ultra vires, official actions taken, 30, Winston, Ken, 257n8
168–69 Wolf, Susan, 20n6
uncertainty, 234 Wolff, Robert Paul, 159, 159n4
undermining principle, 34, 182, 308–9 women, injustices against, 169
arguments for, 96–110 Woodard, Calvin, 302n20
Unger, Roberto Mangabeira, 85, 100n28
unjust laws Yeazell, Stephen C., 53n81
Augustine on, 87
deviating from, 88 Zapf, Christian, 84n39
duty to obey, 9, 160, 168–70, 247 Zimmerman, Michael J., 207n7
John Rawls on, 164–66 Zipursky, Benjamin C., 98n18