JUSTICE, EQUAL OPPORTUNITY, AND THE FAMILY

JUSTICE,
EQUAL OPPORTUNITY,
AND THE FAMILY

JAMES S. FISHKIN

YALE UNIVERSITY PRESS

NEW HAVEN AND LONDON
In order to keep this title in print, this edition was produced using digital
printing technology in a relatively short print run. This would not have hccn
attainable using traditional printing methods. Although the reproduction of this
copy may not appear the same as in the original edition. the text remains the
same and all materials and methods used still conform to the highest hook-
making standards.

Pubhshed with assistance from the foundation

established in memory of Ph1hp Hamilton McMlilan
of the Class of 1894, Yale College.

Copyright© 1983 by Yale Umversity.

All rights reserved.
Tins book may not be reproduced, in whole
or in part, m any form !beyond that
copying permitted by Sections 107 and 108
of the U.S. Copyright Law and except by
reviewers for the pubbc press), without
written permission from the publishers.

Designed by Sally Hams

and set in Trump Medieval type by
Coghill Composition Company.
Prmted m the Umted States of America by Lightning Source

Llbrary of Congress Catalogmg in Publication Data

F1shkin, James S.
Justice, equal opponumty, and the fam1ly.

Includes b1bhographical references and mdex.

l. Equality. 2. Justice. 3. Family.
4. Liberalism. I. T1tle.
JC578.F57 1983 320'.01'1 82-10939
ISBN 0- 300- 02865 - 2
ISBN 0-300-03249-8 ISBN-13: 978-0-30003-2-l9-9
CONTENTS

Acknowledgments vii
1. INTRODUCTION I
2. EQUAL OPPORTUNITY AND THE FAMILY 11
2.1 Distributive Justice: Three Problems 11
2.2 The Principle of Merit 19
2.3 Equality of Life Chances 30
2.4 The Autonomy of the Family 35
3. THE TRILEMMA OF EQUAL OPPORTUNITY 44
3.1 Ideal Theory 44
3.2 Background Inequalities 47
3.3 The First Option 51
3.4 The Second Option 55
3.5 The Third Option 64
3.6 Strategies of Intervention 67
3. 7 Preferential Treatment and Compensation for
the Past 82
4. FAIRNESS, GROUP COMPENSATION, AND
EQUALITY: RESPONSES TO THE TRILEMMA 106
4.1 Fairness and Merit 107
4.2 Competition and Lotteries 110
4.3 Group Competition and Compensation 113
4.4 Unequal Positions 131
4.5 Policy Implications 146
5. OPTIONS FOR LIBERAL THEORY 152
5.1 Contemporary Liberalism: The General Issue 152
5.2 Living with Inconclusiveness 169
Index 195

v
ACKNOWLEDGMENTS

My work on this book was generously supported by a research

grant from the National Endowment for the Humanities. The
manuscript was completed during a fellowship at the Woodrow
Wilson International Center for Scholars in Washington, D.C.,
in the spring and summer of 1981. This support is gratefully
acknowledged. Several colleagues and friends gave me crucial
help at various stages. I would especially like to thank Bruce
Ackerman, Brian Barry, Robert Dahl, Richard Flathman, Wil-
liam Galston, Douglas Rae, and Bernard Williams. I would also
like to thank Richard R. Nelson and Charles E. Lindblom,
present and past directors of Yale's Institution for Social and
Policy Studies. They have created an environment in which
conversations connecting political theory and public policy
flourish spontaneously and this book is one result.

vii
1. INTRODUCTION

E
qual opportunity is the central
doctrine in modern liberalism
for legitimating the distribution of goods in society. Rather
than being concerned with equality of outcomes, liberalism, in
both theory and practice, has been concerned with the ration·
ing of opportunities for people to become unequal. Inequality
has found a justification in the basic idea of a fair competition
among individuals for unequal positions in society. This idea,
sometimes dismissed as weakly reformist or even conserva·
tive, is, in fact, startlingly radical in its implications. If taken
seriously, it would require systematic intrusions into the fam-
ily and a vast reform in the way of life we commonly take for
granted.
The family is the one crucial source of inequality in modern
society that has gone largely unexamined in the theory of
distributive justice. Inequalities between the races, between
the sexes, between states, between generations, inequalities
that result from market forces, from genetic differences, from
the political process-all of these inequalities have become
familiar issues in recent moral and political theory. 1 But the
1. Each of these issues has produced a vast literature. Representative exam·
pies include Marshall Cohen, Thomas Nagel, and Thomas Scanlon, eds.,
Equality and Preferentzal 'freatment IPnnceton: Pnnceton University Press,
1977) on racial and sexual equality; Charles R. Beitz, Political Theory and
lnternatwnal Relations !Princeton: Princeton University Press, 1974) and
Michael Walzer, Just and Unjust Wars !New York: Basic Books, 1977) on
justice between states; R.I. S1kora and Bnan Barry, eds., Obligations to Future
GeneratiOns !Philadelphia: Temple Umverslty Press, 1978) on justice between
generations; Charles E. Lindblom, Politics and Markets !New York: Bas1c
2 l:'IITRODUCTION

role of the family in generating inequalities-while the subject

of extensive empirical study2_has not been systematically
incorporated into normative debate. No defensible and coher-
ent ideal of dtstribution can afford so serious an omission.
Perhaps the family has been largely invisible in liberal the-
ory because of the latter's individualistic presuppositions. Lib-
eralism presents us wtth a picture of isolated, atomic individ-
uals who seem to spring from nowhere in order to experience
utility or claim rights or enjoy liberties and then vanish. My
argument is that any systematic attempt to apply common
liberal assumptions to the normal process by which these
individuals are created and nurtured must break down. A
pattern of new and difficult choices emerges. This pattern will
requue that we modify our commitment to one or another in a
series of crucial liberal assumptions, all of which had previ-
ously seemed fully realizable.
In many ways, this argument is a new variation on an old
theme: the conflicts between liberty and equality. Both princi-
ples have maintained an important place in the liberal tradi-
tion-a tradition that can, in fact, be viewed as a continuing
dialogue about the relative place of the two princtples. 1
Books, 1977) on the 1mpact of market forces; N. f. Block and Gerald Dworkm,
eds., The IQ Controversy jNew York: Pantheon, 19761 on genetic differences;
and James S. F1shkin, ed., SymposiUm on the Theory and Pracllce of Repre~en­
tatwn, a specJalJssue of Eth1cs 91, no. 3IApnl19R1I, on pohtlcal equahty_
2. See, for example, James S. Coleman et al., Equality of Educatwnal
Opportumty, 2 vols., U.S, Dept. of Health, Education, and Welfare, Office of
Education, IWashmgton, D.C.: Government Prmting Office, 19661, Chns-
topher Jencks et al., Inequality INew York: Bas1c Books, 19721 and Christopher
Jencks et al., Who Gets Aheadl The Determmants of Economic Succes.~ m
America !New York: Basic Books, 1979); Fredenck Mosteller and Damcl P.
Moynihan, eds, On Equality of Educatwnal Opportumty INew York: Vmtage
Books, 1972); R1chard H. DeLone, Small Futures. Children. InecJIWllly. and
the L1m1ts of Liberal Reform !New York: Harcourt Brace Jovanovich, 1977);
and Samuel Bowles and Herbert Gmt1s, Schoolmg in Capitalist Amenca INew
York: Bas1c Books, 1976).
3. Two recent mterpretauons of the hberal trad1t1on m terms of the tensmn
between hberty and equahty can be found m J. Roland Pennock, DemocratiC
Theory IPnnceton: Princeton Umverslty Press, 1979) and Amy Gutmann,
L1beral Equality ICambndge. Cambndge Umversity Press, 191!01.
 INTRODUCTION 3

Libertarians have formulated this conflict most dramati-

cally. The interferences with liberty that would be required in
order to maintain equality of outcomes have been a recurring
nightmare for them. 4 Liberals, on the other hand, have usually
managed to defuse the problem by aspiring to realize only a
less demanding form of equality. Instead of equality of out-
comes, they have focused on equality of opportunities. By
prescribing only some appropriate equalization of opportuni-
ties to become unequal, liberals have managed to fend off the
charge that their egalitarian aspirations would require continu-
ous government interference with liberty in order to maintain,
over time, a particular distributional structure (such as equal-
ity of outcomes). 5
I will argue, however, that this apparent success in avoiding
conflicts with liberty is only illusory. If taken seriously, the
liberal strategy of attempting to ration fairly opportunities for
the achievement of unequal positions would require system-
atic intrusions into the family. Only then could the mainte-
nance of background inequalities be rendered compatible with
equal opportunities for the development of talents and other
4. See, for example, F. A. Hayek, Law, Legislatwn and Liberty, Vol. 2, The
Mirage of Social {ustice (London, Routledge and Kegan Paul, 1976); and Robert
Nozick, Anarchy, State, and Utopza (New York: Bas1c Books, 1974) especially
part 2. I w1ll return to this issue m section 4.4 below.
5. Two particularly Important attempts in recent hberal theory to prescnbe
equal opportumties for the achievement of unequal positions can be found m
John Rawls, A Theory of {ustice (Cambndge: Harvard Umverslty Press, 1971),
particularly chapter 2 and Bruce Ackerman, Socwl {ustice m the L1beral State
(New Haven: Yale University Press, 1980), particularly part 2. Two Important
works probmg the hm1ts of this strategy are Bernard Williams's "The Idea of
Equahty" in Peter Laslett and W. G. Runciman, eds., Philosophy, Politics and
Society, Second Senes (Oxford: Basil Blackwell, 1962), pp. 110-31 and Douglas
Rae et al., Equalities (Cambridge: Harvard University Press, 1981), particularly
chapter 4. For claims about the central role of equal opportunity m the
American political tradition see J. R. Pole, The PursUit of Equality m Ameri-
can H1story (Berkeley: Umvers1ty of California Press, 1978); and David M
Potter, People of Plenty (Chicago· University of Chicago Press, 1954), particu-
larly chapter 5. For a critical discussion of the legitlmatmg function for
mequahties served by mentocratlc notions of equal opportunity see Bowles
and Gintls, Schooling, part 2.
4 INTRODUCTION

qualifications. The conflict with liberty, in other words, can-

not be avoided. Rather, it reappears in a particularly excruciat-
ing form, directed at the family.
Once the role of the family is taken into account, the
apparently moderate aspiration of equal opportunity produces
conflicts with the private sphere of liberty-with autonomous
family relations-that are nothing short of intractable. Ele-
ments that are essential to the liberal doctrine of equal oppor-
tunity come into irreconcilable conflict with the private core
of the notion of liberty, the portion that touches most of our
lives most directly.
These conflicts can be formulated in terms of three assump-
tions-two central liberal assumptions about equal opportu-
nity, on the one hand, and our common moral assumptions
about the family, on the other. The first liberal assumption
might be called the principle of merit. According to this
assumption, there should be widespread procedural fairness in
the evaluation of qualifications for positions. No discrimina-
tion should be permitted on the basis of race, sex, class, ethnic
origin, or other irrelevant characteristics. While there are
many interesting controversies about how qualifications for
positions ought to be defined, the principle of merit in some
form is a basic and familiar element in the liberal credo. 6
The second assumption might be called equality of life
chances. According to this notion, I should not be able to enter
a hospital ward of healthy newborn babies and, on the basis of
class, race, sex, or other arbitrary native characteristics, predict
the eventual positions in society of those children. Of course,
there are many different ways of evaluating their eventual
positions. By whatever plausible criterion these evaluations
are made, however, it should be clear that in this society, I can
confidently make such predictions. 7

6. For a detailed discusswn of ment see section 2.2 below.

7. See, for example, Jencks, Who Gets Ahead! pp. 81-83 and DeLane, Small
Futures. pp. 3-19. The independent cumulative effects of meritocratic sorting
 INTRODUCTION 5

These two assumptions can both be defended in terms of the

basic liberal approach to equal opportunity, namely, the notion
that there should be fair competition among individuals for
unequal positions in society. Henceforth, I will refer to this
basic notion as the fair competition assumption. The two
principles just defined can be viewed as explications, respec-
tively, of what might be meant by "competition" in this
context and of what might be meant, in any ultimately defensi-
ble sense, by "fairness."8 On the one hand, the principle of
merit is merely the claim that the competition should be in
terms of qualifications relevant to job performance in the
positions to be filled. The principle of equal life chances, on the
other hand, can be viewed as the central condition that would
render the competition fair. If one can predict where people
will end up in the competition merely by knowing their race or
sex or family background, then the conditions under which
their talents and motivations have developed must be grossly
unequal. It is unfair that some persons are given every conceiv-
able advantage while others never really have a chance, in the
first place, to develop their talents. The principle of equal life
chances, when combined with the principle of merit, would
require equal developmental conditions for talent develop-
ment.
When these two assumptions about equal opportunity are
combined with a third assumption, the autonomy of the fam-
ily-permitting parents to substantially influence the develop-
ment of their children-a pattern of difficult choices emerges.
This pattern takes the form of a "trilemma," a kind of dilemma
with three corners. I will argue that commitment to any two of
these assumptions rules out the third. Attempting to maintain
all three assumptions-the principle of merit, equality of life
on the distnbuuon of life chances are mvesugated by Raymond Boudon m
Education, Opportumty and Socwl Inequality (New York: John Wtley & Sons,
Inc., 19741.
8. There are, of course, some notions of procedural fairness in the very
notion of a compeution. But wtthout equahty of life chances, the competitiOn
lacks background fauness. See sect10n 2.3 below.
6 INTRODUCTION

chances, and the autonomy of the family-would be like

attempting to hold up a three-cornered stool when only two
legs are available. No matter which two corners one chooses to
hold up, lack of the third is enough to undermine the whole
structure.
I will defend this claim for ideal theory. It would be less
surprising if one or more of our central assumptions had to be
sacrificed under unfavorable conditions where extreme scar-
city, lack of compliance, or a historical legacy of injustice made
any particular principle difficult to implement. 9 My argument,
rather, is that under the best conditions that might realistically
be imagined for a large-scale industrial society, this trilemma
arises. In this sense, the basic liberal approach to equal oppor-
tunity does not amount to a coherent ideal once complications
involving the family are systematically taken into account.
Thus far, the argument may be considered a kind of thought
experiment testing the viability of the basic liberal approach to
equal opportunity-the assignment of persons to unequal posi-
tions according to a fair competition-when that approach is
combined with our common assumptions about the autonomy
of the family. Either systematic intrusions into the family
would be required to equalize developmental conditions de-
spite unequal outcomes or the whole liberal focus on equal
opportunities-as distinct from equal outcomes-would have
to be abandoned.
This difficulty with equal opportunity can also be viewed as
part of a more general problem confronting liberal theory. It
has been characteristic of liberalism, in both theory and prac-
tice, to accept background inequalities provided that certain
process-related equalities are maintained. 10 Assignment to po-
sitions in a fair system of equal opportunity is only one such
process equality. Political equality, or equal consideration of
one's preferences by the political system, is another. Equality

9. Sec sccuon .U below for a d1scussion of 1dcal theory.

10. See sectiOn 5.1 below for further d1scuss10n of these process equahties.
 INTRODUCTION 7

before the law, or equal consideration of one's claims by the

judicial system, is a third central process equality. A fourth,
which has had serious advocates, is equal consideration of
one's health care needs by the medical care system. In each of
these cases the central liberal aspiration has been to institu-
tionalize practices that would grant everyone's claims equal
consideration-whether in the political sphere, or before the
judicial system or the medical care system, or as in the case of
equal opportunity, in the system of job assignment. In each of
these cases, the attempt to institutionalize equal consideration
in a selected sphere of life (for example, politics, the courts, or
the system of job assignment) cannot be realized while both
background inequalities and certain key liberties are main-
tained. As we shall see, the background inequalities spill over
onto the process, undermining its claim to equality. And
attempts either to insulate the process or to eliminate the
inequalities directly, would exact, in turn, a substantial cost in
liberty.
Hence, I will argue that the trilemma of equal opportunity to
be explored at length here offers a theoretical case study of a
more general difficulty facing liberal theory. No commonly
recognizable version of liberalism today could do without a
doctrine of equal opportunity any more than it could do
without a doctrine of political equality or of equality before the
law. Yet I will argue that these familiar process equalities are
all subject to a general version of the trilemma. In each case,
without systematic attempts at insulation, background in-
equalities, both social and economic, undermine the claim to
equal consideration in any meaningful sense. It is not only true
that those from the higher strata have greater life chances and
more than equal opportunities; they also have greater than
equal influence on the political process and greater than equal
consideration from the health care and legal systems. In each
of these cases, something closer to full equality in the process
could be systematically restored-but at a cost in liberty. Just
as the liberty of families to influence the development of their
8 INTRODUCTION

children would have to be restricted to fully realize equal

opportunity, other liberties would have to be restricted for
each of the other process equalities. For example, the Supreme
Court ruled in the controversial case of Buckley v. Valeo that
the First Amendment protects expenditures advocating the
election or defeat of political candidates. Such expenditures
could only be restricted, the court reasoned, at a cost to liberty
of expression and association. 11 Yet such restrictions would be
required for any systematic barrier preventing unequal eco-
nomic resources from being translated into unequal political
influence. Similarly, the liberty of individuals to choose, and
pay for, their own lawyers, doctors, and related services would
be at stake in any effort to fully equalize consideration by the
judicial and health care systems of the claims of individuals,
regardless of class or income throughout the society.
In each of these cases, the strategy of insulating background
inequalities from the process in question could only come at
some cost in liberty. Failing to insulate, however, would seri-
ously undermine the claim to equality, for it would be to
accept that the less advantaged have less than equal opportuni-
ties, or less than equal political influence, or less than equal
consideration of their claims or interests from the legal or
medical care systems. To accept such implications under ideal
theory, under the best conditions that might realistically be
imagined for modern industrial societies, would be to admit
that liberalism can only offer a seriously tarnished and corn-
promised ideal for public policy. 12
I will thus argue that the stark conflicts involving equal
opportunity to be explored here represent a particularly dis-
turbing version of a more general difficulty. In broad outlines,

11. Buckley v. Valeo, 424 U.S. 1 11976). See also First National Bank v.
Bellotti, 435 U.S. 765 11978). For a general discussion of the 1ssues raised by
these cases see David Adamany, "PAC's and the Democratic Fmancing of
Pohucs," Arizona Law Review 22, no. 2 (1980): 569-602. See also the dJscus-
SlOn below m section 5.1.
12. See sections 5.1 and 5.2 below.
 INTRODUCTION 9

it is that those process equalities that are at the core of modern

liberalism can be maintained only at a substantial cost in
liberty. This is unavoidable because the process equalities can
be maintained only if either (i) they are systematically insu-
lated from background inequalities, or (ii) those background
inequalities are, themselves, eliminated. Either strategy, insu-
lation or elimination, would require a substantial cost in
liberty. Systematic conflicts between liberty and these merely
process-related equalities are thus unavoidable. Defensible ver-
sions of liberal theory cannot casually affirm fundamentally
incompatible commitments and, at the same time, aspire to
offer us a unified and coherent ideal for public policy.
It might be argued that in the pages that follow I am taking
certain principles too seriously. I make an issue, too quickly,
out of the fact that they cannot be fully realized in a manner
compatible with other key principles. I would reply, however,
that the argument takes place under the conditions of ideal
theory, that is, under the most favorable conditions that can be
realistically imagined for a modern industrial society. We learn
something central about the character of our most fundamen-
tal principles if they stand in irreconcilable conflict under such
conditions. We learn that they do not define a unified and
coherent ideal that we can aspire to implement. Rather, they
add up only to a collection of conflicting considerations, each
one of which would take public policy in a quite different
direction if it were given further emphasis.
Perhaps no general solution at the level of first principles
needs to be found for this pattern of hard choices. A less
ambitious form of liberal theory, one committed only to root
conflicting principles to be traded-off in particular cases, will
be sketched at the end. While less satisfying than some new
and exceptionless first principle, it may be a more honest
response to the true difficulties of the trilemmas to be explored
here. In broad outline, this book will move from the specific
issue of equal opportunity, to the general version of the tri-
lemma applied to other process equalities, and to certain
10 INTRODUCTION

recurring sources of inconclusiveness m theories that aspire to

offer any more systematic solution.
In the end, I will argue that if we can learn to expect less, we
may find a great deal to be satisfied with in a limited liberal-
ism, one that frankly accepts these recurring conflicts, even
under ideal conditions. Instead of offering a single coherent
vision of the just society, to be gradually approached, such a
limited liberalism will focus on identifying the complex vari-
ety of factors to be weighed against each other in particular
cases. It would consist of ideals without an ideal, of conflict-
ing principles without a unified vision in clear focus. Whether
others can arrive at defensible versions that are more system-
atic will depend, I believe, on how they face up to the hard
choices to be explored below.
2. EQUAL OPPORTUNITY
AND THE FAMILY

2.1 Distributive fustice:

Three Problems D istributive justice may be
thought of as a conjunction
of three problems: (a) the problem of value, (b) the problem of
structure, and (c) the problem of assignment. The first is the
issue of what should be considered in questions of distribution
(for example, utility, income, status, primary goods); the sec-
ond is the issue of how it ought to be distributed (for example,
to increase equality, to raise the minimum, or to raise the
total); the third is the issue of to whom it ought to be distrib-
uted (that is, how opportunities for valuable positions ought to
be rationed).
My focus in this book will be on the problem of assignment.
My claim will be that even if the extremely controversial
issues arising from any known proposals directed at the first
two problems were resolved, the third problem, of assignment,
would be intractable within the framework of common liberal
assumptions-once the role of the family were systematically
taken into account.
There are well-known difficulties applying to the first two
problems. While I will not focus on them here, they are worth
identifying briefly in order to place our discussion in context.
By the problem of value, I mean the issue of what it is that is
valuable to individuals whose distribution is being assessed. Is
it income and wealth? Is it utility? Does it include rights and
liberties (as in the case of John Rawls's theory of "primary
goods") ?I Theories of value defined in terms of the satisfaction
1. John Rawls, Theory of fustice, especially pp. 90--95 and 395-99. For a
general discussion of solutions to the problem of value see my Tyranny and

11
12 EQUAL OPPORTU~ITY AND THE FAMILY

of actual wants or preferences must face the well-known

conundrums arising from efforts to make interpersonal com-
parisons of subjective satisfaction. 2 On the other hand, theories
of value not defined in terms of such actual wants or prefer-
ences face an alternative challenge-justifying paternalistic
inferences, in other words, claims that a person is better or
worse off although his own wants and preferences would
support a conflicting assessment."
If we somehow surmounted these difficulties with a work-
able theory of value, a workable theory of what was being
assessed for distribution, we would arrive immediately at a
second basic issue: what is the best structure of distribution for
that value? Should more equal distributions be preferred? Or
distributions with a higher total? Or perhaps those with a
higher minimum?
Each of the principles just mentioned is purely structural;
given any two situations one can determine which is better (or
rank them as equally good) based entirely on the information
available from a listing of payoffs to positions. By payoffs I
mean shares of value as specified by some answer to the first
problem discussed above. By positions I mean either individ-
uals listed in the order of their shares of value or numerically
equal groupings of individuals (n-tiles, each consisting of 1/n of
the population) listed in order of their shares of value.
Purely structural principles require no information apart

LegJtimacy: A Critique of Political Theones (Baltimore: Johns Hopkms, 19791,

chapter 3.
2. See Lionel Robbins's famous article "Interpersonal Compansons of Util-
Ity: A Comment," Economic fournal 48, no. 192(19381: 635-41 and I. M.D.
Little, A Critique of Welfare Economics 2nd ed. (Oxford: Oxford University
Press, 19S 71. For an assertion that at least some rough interpersonal compari·
sons of mtensity are possible, see Little, p. S3 and A. K. Sen, Collective Choice
and Social Welfare (San Francisco: Holden Day, 19701 pp. 79-102.
3. Rawlsian primary goods offer an example. For a critique of Rawls's
account of pnmary goods sec my "Justice and Rationality: Some ObJections to
the Central Argument in Rawls's Theory" Amencan Political Science Review
69, no. 2 (June 19791: 615-29. For a general claim about the paternahsuc
burden facmg such thconcs, sec my Tyranny and LegJtimacy, chapter 3.
 EQUAL OPPORTUNITY AND THE FAMILY 13

from such a listing. We can determine, for example, which

situation is more equal, which has a higher total, or which has
a higher minimum merely by reference to this kind of listing.
Elsewhere, I have argued that such purely structural principles
must be, by themselves, inadequate as ethical criteria for social
choice. As we will see in a moment, they take no account of
assignment-how persons may or may not be moved around
from one position to another in the structure. Through this
insensitivity to assignment, structural principles may legiti-
mate severe deprivations that are entirely avoidable. 4
Of course, some theorists !most notably Nozick) have denied
any role whatsoever for structural considerations. But such
theorists are committed, as a result, to legitimating some
outcomes regardless of structural considerations, regardless of
how minimal the bottom share becomes, or how unequal the
disparity in shares is, or how much the aggregate public wel-
fare declines. 5 Most theorists would not accept such implica-
tions.
The structural considerations just referred to-equality, ag-
gregate utility, and Rawlsian maximin justice !requiring max-
imization of the minimum share)6 are the structural principles
most prominent in contemporary theory. Each of them, how-
ever, is vulnerable to obvious objections.
Equality, as Rawls points out forcefully, can make everyone
worse-off, including those at the bottom. Consider this simpli-
fied choice !where the numbers represent some index of pri-
mary goods or utilities to each representative position in
society under situations A and B, respectively).
4. This argument IS developed at length m Tyranny and Legitimacy. chapter
10. See the examples in sectiOn 2.3 below.
5. Robert Nozick, Anarchy. State, and Utopw. For further cnt1c1sms of
Nozick along these hnes sec, H. L.A. Hart, "Between Utihty and Rights," m
The Idea of Freedom: Essays in Honor of Isaiah Berlin, ed. Alan Ryan (Oxford:
Oxford University Press, 1979) pp. 77-98 and my Tyranny and Legitimacy.
chapter 9.
6. By Rawls1an maximm justice I mean his general conceptiOn that requires
equal distributions unless unequal distributions of primary goods will be to
the advantage of the least fortunate. (See Rawls, Theory of fustice. p. 303.)
14 EQUAL OPPORTUNITY AND THE FAMILY

Situation A Situation B
3 4
3 6
3 8
3 9
Everyone, including the person who is worst-off, is better off in
situation B. The leveling approach to equality can be ruinous
to all.
Utilitarianism, on the other hand, captures the interests of
all-but in a manner entirely insensitive to differing distribu-
tions of the same aggregate quantity of utility. To take two
simplified alternatives, consider the choice between situations
A and B below.
Situation A Situation B
100 0
100 0
100 0
100 400
There is no moral question, for a utilitarian, in the choice
between A and B. They must be ranked precisely equal because
they have precisely the same aggregate utility (and, of course,
the same average utility since population is constant). B must
be as good as A even though B involves severe sacrifices for
three out of four persons while everyone is well-off under A.
And, of course, if the total under B were 401 (instead of 400),
then a utilitarian would be required to prefer B to A.
The utilitarian insensitivity to the special moral claim of the
lowest stratum is remedied by maximin. Rawls's general con-
ception of justice (that the minimum share of primary goods
should be maximized) is the remaining structural solution
which has been particularly prominent in recent discussions.
But how special is the moral claim of the lowest stratum?
Maximin is entirely insensitive to all harms and benefits to
strata other than the minimum. Rawls himself questions
 EQUAL OPPORTUNITY AND THE FAMILY 15

whether the justice or injustice of billions of dollars to higher

strata should turn on whether the lowest stratum 1s improved
by a penny or so. 7 He hypothesizes that these possibilities are
unlikely to arise. Yet there is no adequate empirical basis for
ruling out such possibilities. If the "bucket" by which we
redistribute to the minimum is sufficiently "leaky,"8 we may
be faced with choices like the following:
Situation A Situation B
3.99999 4
15 4
20 4
50 4
Maximin will place any improvement, however slight for the
bottom stratum, above all interests, however great, of all other
strata.
These difficulties are instances of a more general dilemma.
The principles we commonly apply to the problem of structure
fall into two categories. One kind of principle is aggregative-it
is indifferent between all possible distributions of the same
aggregate quantity of goods or welfare. Classical and average
utilitarianism are preeminent examples. A second major cate-
gory consists in principles that are distributive. A distributive
principle prescribes a state of affairs if, and only if, it improves
the (relative or absolutel standing of (one or more ofl the less
fortunate strata. 9 Hence maximin is a distributive principle

7. See Rawls, page 157: "Yet 1t seems extraordmary that the JUStlce of
increasmg the expectations of the better placed by a b1lhon dollars, say, should
turn on whether the prospects of the least favored mcrease or decrease by a
penny." H1s reply on the next page: "The poss1b1hties which the objection
envisages cannot anse m real cases; the feasible set 1s so restncted that they
are excluded."
8 For the "leaky bucket" analogy, see Arthur M. Okun, Equallty and
Efficiency. The Blg Trade-Off (Washmgton, D.C.: The Brookmgs Institution,
19751, pp. 90-120.
9. I borrow this termmology from an 11lummatmg discussion in Brian Barry,
Polltlcal Argument (London: Routledge and Kegan Paul, 19651, p 43. I have
16 EQUAL OPPORTUNITY AND THE FAMILY

because it prescribes choices based entirely on whether they

improve the standing of the bottom stratum. Equality, simi-
larly, is a distributive principle because it prescribes situations
based entirely on whether they improve the relative standing
of less fortunate strata (which might be defined as those below
the median).
The dilemma facing any solution to the problem of structure
is that any aggregative principle (such as utilitarianism) is
necessarily insensitive to distributive objections that must, at
some point, if they are great enough, become overwhelming;
and any distributive principle (such as maximin or equality) is
necessarily insensitive to aggregative objections that must
also, at some point, if they are great enough, become decisive.
While there have been some efforts to develop hybrid princi-
ples incorporating features from both aggregative and distribu-
tive principles, adequate versions of such approaches remain to
be developed. 10 Hence the problem of structure has proved no
more susceptible to easy solution than has the problem of
value.
My focus here, however, will be on the third basic issue in
theories of distributive justice, the problem of assignment:
who gets what position in the structure!
As noted above, many familiar principles are formulated so
that they do not have to take account of this issue at all.
Consider the three columns below:

modtfied the defimuons, however, in order to employ them here. Barry defines
distributive principles more generally to include any cnteria that prescnbe
how a given quantity ts to be distnbuted. My more restrictive defimtion is
meant to limit 1t to those pnnciples whtch are espectally compelling m their
conflict with aggregative constderations. A distnbutive princtple m Barry's
broader sense might, for example, requue that all mcreases be gtven to the top
stratum. Such a princtple would not play a role m the fundamental dilemma
outlined here.
10. For one particularly bold effort, see Douglas Rae, "A Princtple of Simple
Justice," in Peter Laslett and James Fishkin, eds., Philosophy. Politics and
Society, Fifth Series jNew Haven: Yale Univemty Press, 1979), pp. 134-54. As
Rae was the first to pomt out, his hybrid principle-by permuting the alterna-
tion of mdependent aggregative and dtstnbuuve clauses-violates transiuvtty.
 EQUAL OPPORTUNITY AND THE FAMILY 17

Positions Payoffs Persons

P. s. I•
Pz Sz 12
p3 s3 13
p4 s4 14
Ps Ss Is
By positions I mean groupings of persons according to their
payoffs or shares of whatever value !income, primary goods,
utility) is under discussion. So in this case, those in the top
fifth, position P1, receive payoffs of S11 those in the second fifth,
P2, receive payoffs of S2, and so on. Of course, positions may be
divided however finely one might wish into n-tiles and one
may, thus, specify the payoffs for each 1/n of the population.
Utilitarianism, equality, and maximin Ithe maximize the min-
imum share principle) are all examples of purely structural
criteria in that they need reference only to these two columns,
positions and payoffs, to evaluate a situation or compare it to
another. From a mere listing of payoffs to positions, in other
words, one can determine the aggregate or average utility, the
extent of equality, or the size of the minimum share. From this
information alone, there is no way of knowing how persons are
assigned to positions. The information offered us by the first
two columns abstracts from the life history of individuals over
time. It does not tell us how persons acquire and maintain
positions in the structure.
Consider, for example, two alternative societies that are
structurally identical. By this I mean that the payoffs to posi-
tions as specified by our first two columns are exactly the
same. One of these societies offers equal and widespread oppor-
tunities to all strata, ethnic groups, and races in the society.
The other society, however, practices systematic apartheid
throughout its system of assignment so that it is divided
rigidly on racial grounds into distinct sub-societies within the
same geographical territory. Despite such a division, if payoffs
to positions are calculated for the entire apartheid society, the
18 EQUAL OPPORTUNITY AND THE FAMILY

statistics when aggregated into a single set of figures are

identical to our first society. The overall structure of aggregate
welfare, of inequality, and of well-being at the bottom, and so
forth, are all identical in the two societies as represented by
aggregate statistics. Yet the process of assigning people to
positions is starkly different. In the apartheid society racial
classifications play a decisive role whereas in our first society
they do not. Blacks are differentially assigned to the lower
positions (and perhaps some segregated middle positions)
while whites arc differently assigned to the upper positions
(and perhaps some segregated middle positions). By contrast,
race plays no role in the assignment process in the other,
structurally identical society.
This kind of example reveals, I believe, the inadequacy of
any purely structural principle of justice. For any purely struc-
tural principle, defined so that it can determine choices be-
tween any two situations based merely on an account of
payoffs to positions, would have to he indifferent between
these two societies. Within such a framework, there is no room
even for stating the moral issue most of us would raise about
the apartheid society compared to its structurally identical
alternative.
Or consider a less blatant example. Suppose that in one of
the two societies, members of the bottom stratum, 15 in the
column labeled "persons," have no realistic chance of ever
aspiring to the higher positions. Over time, if we took a series
of structural snapshots of the distribution, groups I1 through 14
would change considerably in their assignment according to
some process of equal opportunity. Group I,, however, would
always maintain its position at the bottom. And this great
inequality in life chances, we might assume, perpetuates itself
from one generation to the next. If this system were compared
to one that was structurally identical while endowing all strata
with roughly equal life chances, most of us, presumably, would
see strong reasons for preferring the latter arrangement. Yet
this preference cannot be formulated, or defended, within the
 EQUAL OPPORTUNITY AND THE FAMILY 19

framework of purely structural principles, that is, the frame-

work provided by answers to only our first two problems
above.
The central point is that purely structural principles have no
way of accounting for effects on persons that are independent
of effects on positions. Yet, it is persons, not ranked positions,
who would experience the effects of distributive justice or
injustice. It is their life histories that are at stake, and criteria
that ignore such effects must be inadequate. They must be
supplemented by criteria for assignment to positions-criteria
for how persons may, or may not, justifiably be moved around
from one position to another.
However the first two problems-value and structure-are
resolved, my general point will be that the third one, assign-
ment, poses a special challenge for liberal theory. I have noted
some of the distinctive theoretical issues applying to the
problems of value and structure. These issues, while open and
contested, have generated well-known controversies. Yet the
true difficulty of the third issue, assignment, has never been
squarely faced because liberal theorists have, for some mystify-
ing reason, been blind to its systematic connections with the
family. Hence, while realizing that the first two issues are far
from settled, I will focus on certain distinctive claims about
the true difficulties of the third-difficulties that would apply
even if the first two problems were settled beyond question.

2.2 Th_e Prmc1ple The issue that I dubbed the

of Ment problem of assignment in the
last section has been formulated within liberal theory as the
issue of equal opportunity. In other words, people should be
assigned positions according to the basic competitive assump-
tion stated earlier: there should be a fair competition among
individuals for unequal positions in society.
It is worth noting that some formal or narrow constructions
of equal opportunity would require no more than an impartial
20 EQUAL OPPORTUNITY AND THE FAMILY

assessment of talents and other qualifications relevant to the

positions to be filled. I say "no more than" because this
requirement is open to powerful objections unless it is supple-
mented by criteria specifying an appropriate chance to develop
the desired talents and/or other qualifications in the first place.
There is, for this reason, a strong impulsion within liberal
theory to undertake a two-fold commitment: a principle of
merit (or impartial competition of talents and qualifications,
as developed) and a principle of equal life chances specifying
roughly equal expectations for everyone regardless of the con-
ditions into which they are born. Obviously, assignment by
merit and equal life chances would be simultaneously
achieved only when the causal conditions for talent develop-
ment were substantially equalized across all sectors of society.
I will call the demanding position that fully embraces both
principles the strong doctrine of equal opportunity.
Rawls offers a prime recent example of a liberal theorist who
embraces both components of this strong doctrine of equal
opportunity. He wishes, however, to maintain this commit-
ment and, at the same time, to maintain the family in much its
present form.u By contrast, I will argue that if these two
component notions of equal opportunity are combined with
our customary notions about the family, a trilemma results.
Fulfillment of any two of these assumptions can realistically
be expected to preclude the third. Liberal theorists must either
modify their commitment to one component or the other of
the strong doctrine of equal opportunity, or they must accept
the radical implications of their position for the family. Unless
they face squarely this pattern of conflicting assumptions, they
will have failed to offer a systematic and coherent ideal of
social justice. 12
11. See my d1scuss1on in section 5.1 below.
12. I am assuming that each of these commitments 1s a "strong" one. Note
that these prmc1ples are not formulated so as merely to apply ceteris pan bus or
prima fac1e. They are formulated wnhout exceptions or qualifications so that
any sacnfices m merit, equal hfe chances, or family autonomy must be
cons1dered violations of the stated pnnc1ple, even in cases where one of these
 EQUAL OPPORTUNITY AND THE FAMILY 21

Let us turn now to the first half of this strong doctrine of

equal opportunity. The principle of merit is a common feature
of liberal approaches to equal opportunity. Strikingly, it was
asserted by both sides in the DeFunis and Bakke controversies
over preferential treatment in university admissions. 13 At least
in public debate, there has been an "apparent unanimity re-
garding hiring by competence" (one formulation of this princi-
ple).14 This general kind of principle has achieved wide sup-
port, both on grounds of fairness and of efficiency. 15 Even those
theorists who explicitly criticize it have usually adopted the
strategy exemplified by Rawls-they incorporate it into their
own proposal as a necessary part of an adequate account of
equal opportunity.t6
pnnciples conflicts w1th another. Whether we can arnve at one or more
pnnc1ples for this problem that can be defensibly apphed Without exceptions
or qualifications, or whether we should construe these principles as suscepti·
ble to an "mtuitionistic" process of "trading off" or balancmg IS an issue we
shall return to m section 5.2. The strong and weak terminology 1s standard. It
can be found, for example, m David Lyons, The Forms and Limits of Utili tan·
anism !Oxford: Oxford University Press, 1965), pp. 19-23. I have proposed
some further distinctions in The Limits of Obligation jNew Haven: Yale
University Press, 1982), section 6.
13. Ment IS endorsed not only by opponents of preferential treatment but
also by many of Its supporters. See, for example, Brief of the N.A.A.C.P. Legal
Defense and Education Fund as Amicus Cunae, DeFums v. Odegaard, 416 U.S.
31211974), pp. 4-7; Brief of the National Counc1l of Jewish Women and Others
as Amici Curiae, DeFums, pp. 66-67; Bnef of the City of Seattle as Am1cus
Curiae, DeFums, pp. 18-20. For a more general discussion, see Allan P. Sindler,
Bakke, DeFunis, and Mmonty Admisswns INew York: Longman, 1978),
chapters 7 and 8.
14. Alan Goldman, Justice and Reverse Discnmination IPnnceton: Pnnce·
ton Umvemty Press, 1979), p. 22.
15. Both arguments can be found m Rawls, Theory of Justice, pp. 66--86, and
in Goldman, Justice and Reverse Discrimination, pp. 24-34. For an argument
that efficiency considerations should have priority and that this prionty
restncts the apphcauon of the ment pnnc1ple to agglomeratiOns of JOb assign-
ments that are max1mally productive l"pos1tions" m our definition thus bemg
interpreted broadly), see Norman Damels, "Ment and Mentocracy," Philoso-
phy and Public Affairs 7, no. 3jSpnng 1978): 206--23.
16. See Rawls, Theory of Justice, p. 73; Wilhams, "The Idea of Equality,"
pp. 120-31; and John H. Schaar, "Equahty of Opportumty, and Beyond" in
J. Roland Pennock and John W. Chapman, eds., Nomos IX Equality jNew York:
Atherton Press, 1967), pp. 228-50.
22 EQUAL OPPORTUNITY AND THE FAMILY

We can begin with this definition:

THE PRINCIPLE OF MERIT There should be widespread proce-
dural fairness in the evaluation of qualifications for posi-
tions.
The notion of "procedural fairness," built into the principle
of merit, should be distinguished from the deeper issues of
"background fairness" in the structure of conditions under
which the competition takes place. The latter notion will lead
us, as noted already, to equal life chances. Brian Barry usefully
illuminates the distinction between these two senses of fair-
ness:
Procedural fairness rules out one boxer having a piece of
lead inside his gloves, but background fairness would also
rule out any undue disparity in the weight of the boxers;
similarly background fairness would rule out sailing boats
or cars of different sizes being raced against one another
unless suitably handicapped. In a court case the fact that
one side's counsel showed far greater skill would be
grounds for complaint under the rubric of background
fairness but not procedural fairness. 17
A fair competition, I believe, has certain aspects of proce-
dural fairness captured by the principle of merit. Yet without
conditions of background fairness as well-laying down
ground rules for the acquisition of the characteristics people
are to bring to the competition-procedural fairness would be
insufficient. Some of the competitors might never have a
chance to develop the required skills and qualifications. When
that is the case, there is a serious deficiency in background
fairness.
The details of the above definition of merit deserve com-
ment. "Positions" can be interpreted as individual job assign-
ments which carry with them payoffs in any of the value

17. Barry, Polwcal Argument, pp. 98-99.

 EQUAL OPPORTUNITY AND THE FAMILY 23

schemes discussed in the last section. Or, positions can be

interpreted more broadly as agglomerations of individual job
assignments that would be ranked similarly in any such
scheme of value. By "widespread," I mean something ap-
proaching universality throughout the society. "Procedural
fairness" is more complicated; for the moment, however, we
may identify the general notion as those processes that ap-
proach the model of an impartial competition. Persons are to
be selected for certain characteristics regarded as relevant. For
the moment, let us call them "qualifications." Other irrelevant
characteristics are, so far as possible (within reasonable limits
of cost, time, and effort), to be insulated from affecting the
choice. Of course, this does not mean that every job in society
has to be subjected to formal competition. Informal processes
may strive for, and achieve, procedural fairness in this sense as
fully as more formal methods (that is, by choosing those most
qualified who would have been selected in a formal competi-
tion). The familiar notion of a formal and impartial competi-
tion provides a model against which other more informal
processes can be assessed.
By "qualifications," I mean criteria that are job-related in
that they fairly can be interpreted as indicators of competence
or motivation for an individual's performance in a given posi-
tion. Education, job history, fairly administered test results, or
other tokens of ability or effort might all be included.
I assume, furthermore, that criteria for qualifications can be
developed for each position, which permit rankings of appli-
cants for that position. In other words, the analogue of a fair
competition, determined for the most part by the skill and
effort of applicants, holds for each position.
Some aspects of what might be meant by a "fair assessment"
of qualifications require little discussion. A fair assessment is,
of course, both unbiased and objective. The latter notion might
be interpreted to mean that fair assessments must maintain
some reasonable degree of inter-subjective reliability. But there
is, in addition, another aspect of the fair assessment of qualifi-
24 E Q ll A L 0 P P 0 R T U :'1 I T Y A N 0 T H E FA M I L Y

cations that deserves more comment. I assume that a fair

assessment of an individual's qualifications must rest, cru-
cially, on his own past or present actual performance of rele-
vant tasks, for example, exams, previous employment, or other
relevant experience. Therefore, a determination of qualifica-
tions should not rest simply on statistical inferences (derived
from the behavior of others) about how one might expect a
given person to perform. Such a loose determination of qualifi-
cations could be stretched to include clearly irrelevant factors
such as race or sex if it turned out to be the case that mere
membership in a given race or sex statistically predicted per-
formance in a given job. Because of systematic inequalities in
developmental conditions, this might certainly turn out to be
the case under some empirical conditions. It is important,
however, that the principle of merit should not be extended so
as to codify such inequalities between groups. Not only would
such an extension serve to perpetuate the injustice of the
unequal developmental conditions (that is, those that support
the statistical inference), but it would also constitute an inter-
pretation of qualifications that is manifestly unfair. An indi-
vidual in a given group who is subjected to "statistical discrim-
ination,"1R in this way, would never have had a chance to
compete, to prove his own competence. He would be judged
entirely on the basis of the performance of other persons-
persons who happen to share some arbitrary characteristic
with him (or her). For this reason, I will assume that a fair
assessment of an individual's qualifications must be based,
crucially, on that individual's past or present actual perform-
ance of relevant tasks.
This interpretation is supported not only by considerations
of procedural fairness but also by reasonable interpretations of
our basic notion of fair competition. If individuals were as-
signed to positions without being given the chance to actually

18. I take this term from Lester Thurow, Generatmg Inequality (New York:
Basic Books, 1975), pp. 170-81.
 E Q U A L 0 P P 0 R T U N IT Y A N D T H E FA M I L Y 25

perform relevant tasks, in an important sense they would have

been denied the chance to compete actively for positions.
Having others draw statistical inferences about how they ex-
pect you to perform in a competition is not the same as
participating in a competition and performing the relevant
tasks yourself. For this reason, individuals must be classified
according to their actual performance.
This notion of qualifications should generally rule out deci-
sions conforming to the principle of merit that are based on
native characteristics such as race, sex, or ethnic origin. It
might leave room, however, for a few isolated cases where such
a characteristic has a strong contingent association with suc-
cessful performance in a given job-because of the idiosyn-
cratic character of the tasks involved. In some familiar cases,
this contingent association is likely to be so strong that a
native characteristic approaches the status of a necessary con-
dition for holding a job. Being female might be considered
necessary for modeling certain clothes in the fashion industry.
Being male might be necessary for playing tackle in the Na-
tional Football League. Or consider a looser association.
Should being black be considered necessary for the job of
teaching Afro-American studies? While many would hold that
it should not be considered a qualification, in itself, it is likely
to be strongly associated with other characteristics-such as
sympathy and experience with the culture to be taught-and
most successful applicants for such a position might, justifi-
ably, be black. Nevertheless, so long as race is not, itself, made
a decisive qualification, it would be theoretically possible for
other applicants to compete successfully.
In cases where the contingent association is so strong that
membership in a given race, sex, or religion amounts to a
necessary condition for successfully performing a given job las
defined in that society) and as long as such cases are idiosyn-
cratic and isolated, they would not bear on the major issue.
The principle of merit, in other words, need not apply to every
single position if its application is widespread throughout the
26 E Q U A L 0 P P 0 R T U :'II I T Y A N D T II E FA 1\\ I L Y

society. A few isolated cases of nonmeritocratic assignment

would not be worrisome, particularly If those excluded from
competing for a given position have an abundance of compara-
bly attractive alternatives they can compete for.
On the other hand, when tasks are commonly defined so that
arbitrary native characteristics are often relevant to the per-
formance of work roles, as so defined, then a challenge to the
principle of merit of a quite different kind arises. Suppose that
in a given society all the higher positions are presently occu-
pied by white males and that the ability to work effectively,
without disruption, with the other occupants of high positions
is built into the definition of the tasks required for those
positions. In such a society, it would not be difficult to imagine
widely accepted ideological tenets differentiating race and sex
whereby females and nonwhites introduced into the higher
positions would be a disrupting influence (at least for a time).
Could such considerations justifiably enter into the definition
of the work roles for which people are to compete, according to
the principle of merit?
The difficulty is that if qualifications were stretched to
include such considerations, certain native characteristics
would, de facto, turn out to be decisive in the process of
assignment. The definition of "merit" would have been dis-
torted in such a way that certain groups would never really
have had a chance to compete for the positions in question.
This example points out that an adequate principle of merit
must do more than require a fair assessment of the extent to
which an individual has acquired certain qualifications. It
must also require a fair determination of what are to count as
qualifications in the first place.
Recall the definition of qualifications specified above: "crite-
ria that are job-related in that they fairly can be interpreted as
indicators of competence or motivation for an individual's
performance in a given position." Should a fair interpretation
of an individual's competence and motivation to perform in a
given position include reference to his or her race, sex, or
 EQUAL 0 P P 0 R TUN IT Y A N D THE FA M I L Y 27

ethnic origin? Familiar interpretations of the principle of merit

would, in fact, hold all such factors to be irrelevant. Some
views apply this notion quite strictly so that they must be
completely irrelevant to each particular position falling under
the principle of merit. Other views would accept "separate but
equal" competitions under certain circumstances. For exam-
ple, while males might not be eligible to model female clothes
in the fashion industry, they would be eligible to model male
clothes. If the opportunities were truly comparable, the looser
interpretations of the principle of merit (those permitting dual
competitions) would find such arrangements admissible. Of
course, the "separate but equal" doctrine has a notorious
history but that is, in large part, because it was employed in
contexts where separate clearly was not-and perhaps could
never have been-equal. 19 We need not decide this question
here. The argument I will develop applies equally to the strict
meritocratic doctrine for each single position and to those
interpretations that would permit some separate but equal
competitions-provided that the parallel opportunities were,
in fact, equal.
I leave open the question, in other words, whether separate
but truly equal competitions are ever admissible for sexual,
racial, or ethnic divisions in the population. I leave this ques-
tion open because my argument is meant to apply to all of the
defensible varieties of the principle of merit; different but
plausible versions of the doctrine handle this question in
different ways. My hope is that the argument developed here
will apply to all of them.
The basic point, however, is that native characteristics
should not have any direct role in determining the payoffs any
individual can expect from the process of meritocratic assign-
ment. Either such factors should be completely irrelevant or
their effect should be ameliorated through parallel but equal
19. See R1chard Kluger, Simple fustlce: The H1story of Brown v. Board of
Educatwn and Black Amenca's Struggle for Equal1ty !New York: Vintage
Books, 1977).
28 EQUAL OPPORTUNITY AND THE FAMILY

competitions. For the moment I will set aside this looser

interpretation, permitting the possibility of some parallel com-
petitions, and concentrate on the straightforward case of a
unified process of meritocratic assignment regardless of all
native characteristics.
There are two aspects of fairness that must be emphasized.
The first concerns the assessment of qualifications. The sec-
ond concerns the determination of standards for that assess-
ment, the determination, in other words, of what are to count
as qualifications for a given position. Except where their effect
has been ameliorated by equal parallel competitions, native
characteristics should not have an explicit role in either pro-
cess. By a native characteristic, I mean any factor knowable at
birth that could be employed to differentiate adult persons of at
least normal health and endowment. 20 A person's race, sex,
ethnic origin, and family background would all count as native
characteristics. Even the purely genetic component of IQ, if it
were possible to isolate such a thing, would count as a native
characteristic. A ban on considering the purely genetic compo-
nent of IQ would not rule out considering fully developed IQ
which is, of course, influenced by environment as well as by
heredityY
The basic idea is that an individual has no responsibility
20 A similar conception of native charactenstics 1s employed by Goldman,
Justice and Reverse D1scnminatwn. p. 102, for example. However, one of h1s
arguments agamst cons1denng them in ass1gnmg people to positiOns mterprcts
them as "unalterable charactenstics unrelated to performance" jp. 34). Even
though native charactenstics can be ascribed to an mdJvJdual at buth, they are
not necessarily unalterable, as cases of sex change Illustrate dramatically. I
have restncted the discussiOn to persons of normal health and endowment
because there are spec1al issues applymg to the handicapped-Issues of suffi-
Cient Importance to JUStify lengthy treatment that would distract from our
central argument here. It would hardly be novel to cla1m that there are d1fficult
controversies applymg to equal opportumty for the handicapped. My argument
here IS duected at estabhshmg the difficulty of equal opportumty for the
supposedly easy case of normal persons who have no s1gmficant physical or
mental handicaps. For discussion of justice mvolvmg the handicapped see
Goldman, Justice and Reverse Discnmmatwn. pp. 180-182 and Susan Rose-
Ackerman, "Mental Retardation and Soc1ety," forthcommg m Ethics.
21. For the current confused state of research on the genetic component of
 E Q U A L 0 P P 0 R T U N I T Y A N D T H E FA M I L Y 29

whatsoever for any of his or her purely native characteristics.

Race, sex, family background, or ethnic origin can all be simply
ascribed to an individual at birth. It is, therefore, unfair to
consider them explicitly in the process of meritocratic assign-
ment, either in the assessment of qualifications or in the
determination of what factors are to count as qualifications.
They are not factors subject to competitive efforts at acquisi-
tion; therefore, it would be unfair to let them determine the
competition among talents, as developed. Of course, such
factors will usually be ruled out, in any case, since it would be
unusual for them to support the justificatory burden of being
job-related at least for most of the work roles that are familiar
to us. Were work roles to be defined, however, so that such
factors became relevant, as in the example of the hierarchical
white male system, the above assumption about fairness
would prevent them from being considered.
What do I mean by this ban preventing native characteristics
from being "considered explicitly" in the process of merito-
cratic assignment? The qualifications should be defined, and
the competition should be run, in such a way that it should
always be possible for two persons who differ markedly in any
such native characteristics to be judged equally qualified. More
precisely, the qualifications should be defined so that a black
could change places with a white, a female could change places
with a male, or a Jew could change places with a gentile, for
example, and no noticeable effect on the assessment of qualifi-
cations need arise from the exchange. 22 The evaluation of merit
should be blind to such considerations.
Of course, some have advocated sacrificing or abridging the
IQ see Block and Dworkin, The IQ Controversy. espec1ally part 2, and H. J.
Eysenck and Leon Kamm, The Intelhgence Controversy !New York: john
Wiley & Sons, Inc., 1981).
22. For a more general d1scuss10n of th1s kmd of exchange requuement, see
my L1m1ts of Obhgatwn, secuon 5. For a s1m1lar application to notwns of
merit, see Goldman, p. 53. For an attack on th1s admittedly liberal requue-
ment-on the grounds that It 1gnores differences between the sexes-see
Elizabeth H. Wolgast, Equality and the R1ghts of Women !Ithaca: Cornell
Umvers1ty Press, 1980), chapter 1
JO E Q l 1 A L 0 P P 0 R T l' :'II I T Y A N f) T H E FA M I L Y

principle of merit as compensation for past injustice to particu-

lar groups. As long as the conflict with merit considerations is
explicitly acknowledged, nothing I have said thus far deter-
mines, one way or another, whether such practices would be
justifiable. I will return to this question later.H For the moment,
it is worth noting that my argument about the necessity for
sacrificing the principle of merit under certain conditions will
take place, by contrast, within the ground rules of ideal theory.
My argument is that the principle of merit conflicts irrecon-
cilably with other core liberal assumptions under the best
conditions that might realistically be imagined for a modern
industrial society. While liberal theorists might be resigned to
sacrifices in the principle of merit as compensation for past
injustice, there is a much more disturbing challenge to liberal
theories of justice, I believe, in the requirement for such
sacrifices under ideal conditions. As one corner of the tri-
lemma to be explored later, the principle of merit is, indeed,
placed in such jeopardy.

2.3 Equality of while the principle of merit

Life Chances has usually been accepted
as a necessary part of an adequate account of equal opportunity
jat least under favorable conditions), 24 it is vulnerable to some
obvious objections when applied by itself. There are cases
when the principle of merit could be fully implemented and we
would all be fairly sure that equal opportunity, in any mean-
ingful sense, had not been achieved.
Consider this example which I have adopted from Bernard
Williams. 2' Imagine a society dominated by a warrior class-a
society where the top positions are all occupied by skillful
warriors who have perpetuated their positions from one gener-
ation to the next. Suppose, however, that advocates of equal

2.~. Sec secuons 3.7 and 4.3 below.

24. See Goldman, fustJce and Reverse Discnmmation, chapter 2. See also,
n. 5, chapter I above.
25. W!lhams, "The Idea of Equahty," p. 126.
 EQUAL OPPORTUNITY AND THE FAMILY 31

opportunity institute a reform. They decide that admission to

the warrior class should, from that point on, be determined by
a suitable competition.
As it happens, however, the present warriors' children over-
whelmingly dominate the competition so that the reform
makes no difference in the outcome. We might imagine, for
example, that children of the warrior class are all well
nourished while their competitors are so undernourished that
they fail in the competition for lack of strength. Any reason-
able witness to the spectacle of 300 pound wrestlers vanquish-
ing 90 pound weaklings would conclude that an inadequate
and merely procedural kind of equal opportunity had been
realized by the reform. Nevertheless, it should be noted that
the winners of the competition would, undoubtedly, be better
warriors. Choosing them would be fully compatible with the
principle of merit; our dissatisfaction with the reform has to do
with the unequal causal conditions under which merit-
namely the characteristics measured in the competition-has
developed. In an important sense, children of the lower strata
never really would have had a suitable chance to win the
competition because they never really would have had a suit-
able chance to develop the valued characteristics.
Suppose, however, that children of the wa"rrior class all win,
not because they are better fed, but because their parents have
helped them develop the appropriate skills. They are, indeed,
better warriors, but because of parental influence rather than
because of superior nourishment. If one could predict the
outcome of the competition, just as reliably, by reference to
family background, then the institution of a procedurally fair
competition still would appear inadequate as an account of
equal opportunity. It would merely institutionalize grossly
unequal life chances. The basic intuition is that it seems unfair
that we should be able to predict eventual positions in a
society merely by knowing the strata into which children are
born. If that is the case, then we may say that the initial
distribution of life chances is unacceptably unequal. In this
case, it is unequal because of identifiable and remediable
32 EQUAL 0 P P 0 R T U NIT Y AN D T H E FA M IL Y

causal conditions which differentially affect the development

of qualifications in the first place.
It is worth adding that we might object to such grossly
unequal life chances whether or not they came about through
differential opportunities for talent development. If the top
stratum were not composed of more skillful warriors, hut
constituted, instead, a hereditary aristocracy that perpetuated
its control over important positions in the society-without
any special claim to skill or expertise-then, surely, the result
would be at least equally objectionable. Differential life
chances could, in this case, be predicted just as easily on the
basis of the stratum one was born into. The fact that the top
stratum was mediocre in its talents would only provide further
grounds for criticizing the system.
Let us define the root notion operating in these objections:
EQUALITY OF LIFE CHANCES The prospects of children for
eventual positions in society should not vary in any
systematic and significant manner with their arbitrary
native characteristics.
By "prospects," I mean the probability of reaching given
positions, as that probability might be impartially and objec-
tively assessed. By "positions," I mean here any scheme of
ranking social outcomes or job assignments. Any such scheme
can be thought of as an answer to the questions I identified in
section 2.1 under the problem of value (in theories of distribu-
tive justice). Positions might, in other words, be ranked in
terms of income, utility, social esteem, Rawlsian primary
goods, or other such theories.
What do I mean by "arbitrary" native characteristics? A
native characteristic will be considered arbitrary unless it
predicts the development of qualifications to a high degree
among children who have been subjected to equal developmen-
tal conditions.26 Hence if IQ or even height turned out to be
26. Of course, there are many d1menswns along wh1ch equal developmental
conditions might be measured. Some versions of equal envuonmental cond1·
uons w1ll reveal dramatic differences while others w1ll not. J. M. Thoday cites
 EQUAL OPPORTUNITY AND THE FAMILY 33

major qualifications of job relevance in a given society, know-

ing the genetic component of such factors at birth might
permit one to predict the differential development of qualifica-
tions-even among persons whose environmental conditions
had been fully equalized. But unless a native characteristic
withstands this justificatory burden of predicting the develop-
ment of qualifications under equal conditions, it will be con-
sidered morally irrelevant to the development of qualifications
and, hence, arbitrary. Thus race, sex, ethnic origin, family
background, and other such familiar dimensions of discrimina-
tion will be considered arbitrary here. Under equal develop-
mental conditions, I will assume that knowledge of these
factors would not permit us to reliably predict qualifications of
individuals for desirable positions in the society. Inequalities
in the qualifications achieved by these groups, I will assume,
can be traced to differential developmental conditionsY

some interestmg expenments mvolvmg plants. When two distmct genotypes

of goldenrod are both subjected to the same h1gh hght mtens1ty, theu develop-
ment 1s mdistingu1shable. But when they are both subjected to the same low
hght mtens1ty dramatic d1fferences ensue. Clearly, the chmce between differ-
ent vers10ns of equal developmental conditions may be cruc1al. See J. M.
Thoday, "Limitations to Genetic Companson of Populations," m Block and
Dworkm, The IQ Controversy, pp. 131-45.
27. Relaxing th1s assumption, and treatmg these mequahties between
groups !defined along racial, ethmc, or class lines) as genetic !rather than
envuonmental) 1s of course, another theoretical poss1b1hty. The controversies
recently generated by Jensen and Herrnstem arose from theu bnngmg mto
question the assumption made here Hensen along rac1al hnes, Herrnstem along
class divisions in a meritocracy). I w1ll not enter into th1s controversial and
unsettled empuical 1ssue here. Rather, my strategy w11l be to employ the
optim1st1c and qumtessentially liberal assumption that there are no system-
atic genetic differences between groups !defined along rac1al, ethmc, or class
d1V1s10ns) that would bear on mentocratic assignment. Based on this, my
argument m chapter three can be read as a claim that even 1{ this optimiStic
empirical assumption IS employed, the tnlemma of equal opportunity IS
unavoidable. Clearly, the confhct between ment and equal hfe chances would
only be exacerbated if the Jensen or Herrnstem theses were correct. For then,
even With equalization of developmental conditiOns laccomphshed through
sacnfice of fam1ly autonomy), equal hfe chances could not be achieved with
the pnnc1ple of ment in place. Membership m the allegedly genetically
supenor or infenor group would pred1ct d1fferential development of compe-
tence even under equal developmental cond1t10ns. Hence, it would pred1ct
34 E Q U A L 0 P P 0 R T U N I T Y A N D T H E FA M I L Y

Hence, by the principle of equal life chances, tf I can predict the

outcomes achieved by an individual merely by knowing his or
her race, sex, ethnic origin, or family background, then equal-
ity of life chances has not been realized.
It may seem odd that some native characteristics, nonarbi-
trary ones, are strictly insulated from the process of merito-
cratic assignment but are not, similarly, insulated from the
distribution of equal life chances. For example, the purely
genetic component of IQ, were it possible to measure it with
confidence, could not, by itself, constitute a qualification rele-
vant to the principle of merit. IQ, as fully developed, on the
other hand, would be admissible, as would any other measures
of cognitive skills that were job-related. We have already ex-
plored a variety of reasons for this conclusion. One reason
arises directly from the fairness requirement, introduced ear-
lier: A fair assessment of an individual's qualifications must
rest, crucially, on his own past or present actual performance
of relevant tasks, for example, exams, previous employment,
or other relevant experience. Obviously, any native character-
istic (that must, by definition, be ascertainable at birth) could
not count directly as a qualification in this sense, since it
would not depend on an individual's actual performance of any
relevant tasks. The future ability to perform such a task,
however, would be a different matter.
Why, on the other hand, should the principle of equal life
chances not be formulated, similarly, so as to insulate the
distribution of life chances from all such native characteristics
(and not merely the arbitrary ones)? We might envision a
possible formulation of the principle that would have this
effect-namely, that persons should have equal life chances
differential life chances (because of diffenng competence) With the pnnciple of
ment m place. This kmd of argument would requue not only (a) the controver·
sial geneuc differences between groups and (b) the high heritability of the
relevant factors, but also (c) mentocrauc assignment in the soc1ety according
to precisely those factors that conform to (a) and (b). Even 1f the controversJal
claims alleged for IQ were established (a and b), the substantial role of many
other cognitive and personality factors m any plaus1ble system of meritocrauc
 EQUAL OPPORTUNITY AND THE FAMILY 35

regardless of whether they are born more or less talented (or

with greater or lesser genetic components of the relevant
talents). I have not formulated the principle in this way be-
cause the version of the principle that has had greatest appeal
has been more concerned with the equalization of life chances
across differing environments than across differing genetic
endowments. The idea has been to equalize chances for natural
talents to flourish rather than to equalize chances regardless of
natural talents.
I mention this kind of more demanding principle to note that
the argument of the next chapter applies equally to it. Our
definition of equal life chances could easily be adapted to yield
such a requirement. Such a principle would simply regard all
native characteristics as arbitrary-including those clearly rel-
evant to the development of talents under equal conditions.
This principle would be far more demanding in that it would
equalize life chances not only for all the cases treated by our
proposal (across sectors of society, ethnic groups, sexual and
racial divisions), but also across differing natural endowments
of the relevant talents. Given the difficulties we will encounter
with the weaker version of the principle in the required manip-
ulation of developmental conditions, I note the stronger one
only in passing as a more demanding version that produces the
same trilemma but that has other implications as well.

2.4 The Autonomy

of the Family T he third assumption is de-
fined as follows:
AUTONOMY OF THE FAMILY Consensual relations within a
given family governing the development of its children
ass1gnment would be enough to throw the issue open again. See Block and
Dworkin, The IQ Controversy; Arthur R. Jensen, Educability and Group
Differences !London: Methuen, 1973); R. J. Herrnstem, IQ m the Memocracy
!Boston: Little Brown, 1973). For a samphng of the storm of protest ra1sed by
Jensen's thes1s, see Env1ronment, Heredity, and Intelllgence, compiled from
the Harvard Educational Review, reprint senes, 2ICambndge: Harvard Educa·
tional Review, 19691.
16 EQUAL OPPORTll:'liJTY A:'IID THE FAMILY

should not be coercively interfered with except to ensure

for the children the essential prerequisites for adult par-
ticipation in the society.
By "essential prerequisites," I mean the physical and psycho-
logical health of the child and his or her knowledge of those
social conventions necessary for participation in adult society.
Literacy, the routines of citizenship, and other familiar ele-
ments of secondary education would count among the essen-
tial prerequisites. Hence, parents who failed to ensure such
prerequisites for their children could justifiably be subject to
state interference. 2R
By a "family," I mean here "a community composed of a
child and one or more adults in a close affective and physical
relation which is expected to endure at least through child-
hood."29 Such a family may he more extended than the familiar
nuclear one and it need not, for purposes of this broad defini-
tion, include two parents of opposite sex. I assume, however,
that natural parents have the initial option of creating families
!subject to this principle) if they do so in a consensual manner,
that is, one not subject to sustained and intense disagreement.
Hence this definition leaves open questions about the appro-
priate role of the state when there is sufficient disagreement
!non-consensual relations) between natural parents over who
should be included in a given family. In other words, this
principle is not intended to resolve questions of child custody
when parents disagree, nor is it intended to resolve questions
of child placement when neither of the parents wants the

28. I leave open the quesuon of cntcna for JUStified mtervenuon m such
cases falhng outside the area of famtly autonomy protected by the pnnc1ple
stated here. The ever present possththty of the chtld suffenng more harm than
good from government mterventmns that dtsrupt the conununy of family
relations should be emphastzcd. Sec Joseph Goldstcm, Anna Freud, and Albert
f. Solnit, Beyond the Best lntere.~ts of the Child (New York: Free Press, 19731
and Before the Be.~t Interests of the Ch1ld (New York: Free Press, 1979).
29. John E. Coons and Stephen D. Sugarman, Education by ChOJce: The
Case for Fam1ly Control (Berkeley: Umvcrsny of Cahforma Press, 1978), p. 53.
 EQUAL OPPORTUNITY AND THE FAMILY 37

child. 30 Similarly, this definition is not intended to resolve

questions about "children's rights" when consensual relations
between adults and children break down to the point, for
example, that a child wishes to be placed in another family or
in an institution.·ll
Rather, this definition is directed at cases where the state
might wish to interfere in a family that preserves consensual
relations-where there is not sustained and intense disagree-
ment about its basic terms of cooperation. This principle
would give parents considerable latitude in influencing the
development of their children, in a manner preserving consen-
sus within the family, as long as provision of the essential
prerequisites for adult participation in the society were not
endangered:l 2
Furthermore, by limiting the autonomy principle to matters
bearing on child development, I mean to define the principle
narrowly so as to avoid the most obvious and controversial
areas of conflict with the principle of merit. Most importantly,
family autonomy is not meant to govern job assignments in
adult society. It should not be construed to protect nepotism,
the buying of positions, or large-scale inheritance. Such prac-
tices would raise obvious conflicts with any attempts to fully
institutionalize merit. My argument will require only a narrow
and comparatively noncontroversial construction of family

30. See Goldstem et a!., Beyond the Best Interests, part 2.

31. See Goldstem eta!., Before the Best Interests, chapter 7; Beatnce Gross
and Ronald Gross, eds., The Children's Rights Movement (New York: Anchor
Press/Doubleday, 1977); and Patncta Vardm and Ilene N. Brody, eds., Chil-
dren's R1ghts. Contemporary Perspectives (New York: Teacher's College Press,
1979).
32. For a basically similar position, mamtaming that the autonomy of mtact
families should not be interfered with, except when extreme harm to the ch1ld
IS in questiOn, see Goldstem et a!., Before the Best Interests, chapters I and 2.
See also Laurence D. Houlgate, The Ch1ld and the State: A Normative Theory
of fuvemle R1ghts (Baltimore· Johns Hopkms, 1980). Houlgate's arguments
protecting paternalism Withm the family, along with his arguments for posi-
tive and negative claim nghts protectmg children from various forms of harm
and abuse, are compatible with the pnnc1ple developed here.
38 E Q U A L 0 P P 0 R T U N I T Y A N D T H E FA M I L Y

autonomy. As defined here, the principle protects only the core

of parent-child relations, namely, consensual efforts to influ-
ence child development.
What do I mean by "coercive interference" in this definition
of the autonomy principle? A person (or group), X, attempts to
coercively interfere with the actions of a person (or group), Y, if
X attempts to direct Y to choose one alternative course of
action, A, rather than other courses, B, C, D, and so on, by
either (i) threatening Y with disadvantages he would not other-
wise encounter if he fails to choose A, or (ii) by systematically
suppressing knowledge or discussion of the alternatives to
action A (or the favorable aspects of those alternatives) with
the intention of directing Y to choose A.·13
Hence, if the state threatens nonconforming families with
imprisonment or loss of custody of their children, that would
clearly constitute coercive interference. Such coercive interfer-
ence would, as noted, be justified in some cases in order to
ensure a child's education or essential aspects of his physical or
psychological health. Or, if the government were to manipu-
late behavior by suppressing discussion and knowledge of all
alternatives but certain approved paths, that would also consti-
tute coercive interference.
It is worth noting that this principle is a relatively weak one.
The protection it offers from coercive interference docs not
apply whenever consensual relations in a family are lacking or
whenever the essential prerequisites for adult participation are
in serious question for a given child. It would protect the
family from interference only when these common grounds for
33. For an account of coerciOn generally similar to the first clause m thts
definitiOn, see Robert Noztck, "Coercion," m Peter Laslett, W. G. Runciman,
and Quentm Skmner, Philosophy, Poht1cs and Society, Fourth Series (Oxford:
Basil Blackwell, 1972), pp. 101-35. Unlike Noztck, however, I would consist-
ently treat omissiOns that would be mtenuonally and causally connected to
someone bemg made worse-off as "threats." Nozick ("Coercwn," p. !IS) leaves
the questiOn open, as m the case of P (m a rowboat) refusmg to save Q (in the
water) from drownmg. The second clause of thts definitiOn has an independent
basis; It rehes on the notion that bramwashmg and mdoctrination violate
liberty coerctvely just as effectively as mtght force or the threat of force.
 E Q U A L 0 P P 0 R T U N I T Y A N D T H E FA M I LY 39

justified interference are both lacking. It thus attempts to

single out the least controversial portion of the sphere of
action within which families might claim autonomy from
coercive interference.
Not only is this principle appealing in its own right but it
can also be supported within a broader framework of negative
liberty. While this extension is not, strictly speaking, essential
for the main argument below, it does help identify the stakes at
issue were the autonomy of the family to be intruded upon.
According to some important notions in the liberal tradition,
protection for the family is embedded within a broader concep-
tion of negative liberty. While I will employ the precise as-
sumption defined above, liberals who are committed to the
broader framework of negative liberty will find a crucial por-
tion of that framework at risk in the argument below. The
autonomy of the family assumption defined above identifies
that portion of negative liberty which touches most of our
lives most directly.
It has long been a central concern of the liberal tradition that
there be a significant sphere where a person, in Berlin's words
"should be left to do or be what he is able to do or be, without
interference by other persons."3 4 Much liberal thinking on the
definition of this sphere has been influenced by Mill's "harm
principle"-his "one very simple principle" in On Liberty.
That principle is that the sole end for which mankind are
warranted, individually or collectively, in interfering with
the liberty of action of any of their number is self protec-
tion. That the only purpose for which power can be right-
fully exercised over any member of a civilized community,
against his will, is to prevent harm to others. 3s

34. Isaiah Berhn, Four Essays on Liberty (Oxford: Oxford University Press,
1969), pp. 121-22.
35. John Stuart Mill, On Liberty, (New York: Bobbs-Mernll, 1956), p. 13. I
take the term "harm pnnciple" from Joel Feinberg's influential discusswn in
Social Philosophy, (Englewood Chffs, N.J.: Prenuce-Hall, 1973), chapters 2
and 3.
40 EQUAL OPPORTUNITY AND THE FAMILY

The sphere of what Mill calls "self-regarding" action should

thus be immune from interference precisely because, by defini-
tion, it does not harm others. Regardless of whether Mill's
principle is taken in the strong sense he proposes, namely, as
stating the necessary (and perhaps sufficient) conditions for
justified interferences with liberty, it can usefully be employed
to identify an essential part of the justified sphere of liberty. 36
Mill's principle offers a powerful and appealing position: acts
that do not harm others should be immune from coercive
interference. As Mill describes this self-regarding area of life:
there is a sphere of action in which society, as distin-
guished from the individual, has, if any, only an indirect
interest: comprehending all that portion of a person's life
and conduct which affects only himself or, if it also affects
others, only with their free voluntary, and undeceived
consent and participation.37
Yet this position, as formulated, applies only to those "in the
maturity of their faculties." Its application to children is left
unclear. Mill's comments about children later in the essay
suggest how the general framework might be extended. First,
with respect to education:
It still remains unrecognized, that to bring a child into
existence without a fair prospect of being able, not only to
provide food for its body, but instruction and training for
its mind, is a moral crime, both against the unfortunate
offspring and against society; and that if the parent does
not fulfill this obligation, the State ought to see it fulfilled,
at the charge, as far as possible, of the parent. 38
36. For cnt1c1sm that M1ll's pnnc1ple may cover only a tiny sphere of action,
depending on how the causation of harm IS construed, see R1chard Taylor,
Freedom, Anarchy and the Law (Englewood Chffs, N.j.: Prentice-Hall, 1973),
chapter 9. For a more sympathetic mterpretauon, see Femberg, Social Philoso-
phy, chapters 2 and 3.
37. Mtll, On Liberty, pp. 15-16
38. Ib1d., p. 128. The caveat hmttmg the pnnctple to persons "m the
matunty of their faculties" appears on p. 13.
 EQUAL OPPORTUNITY AND THE FAMILY 41

Hence, the state may require that children be educated up to a

certain standard and this coercive interference, when neces-
sary, is justified in order to prevent severe harm both to the
child and to others in society at large.
A similar extension of the harm principle can be applied to
the initial act of creating a new human being:
The fact itself, of causing the existence of a new human
being, is one of the most responsible actions in the range of
human action. To undertake this responsibility-to be-
stow a life which may be either a curse or a blessing-
unless the being on whom it will be bestowed will have at
least the ordinary chances of a desirable existence, is a
crime against that being. 39
Under certain conditions of overpopulation or scarcity, pro-
duction of a new human being may be coercively regulated in
order to prevent harm to the one conceived (and to others).
While there are many interesting questions about the concep-
tualization of harm to possible persons, 40 this interference
might be interpreted as resting on a claim that there are some
conditions under which it would have been better for a person
not to have been born at all. Under such deprivating condi-
tions, one is, in effect, harming a person by bringing him into
existence to lead such a life. 41
For both these issues-education and procreation-the self-
regarding framework of the harm principle can be extended to
include regulation by the state of parental actions when suffi-
cient harm to the child is in question. Here children are
39. Ib1d., p. 132.
40. See Stkora and Barry, OblJgations to Future GeneratiOns; Derek Parfit,
"On Domg the Best for our Children," in Mtchael Bayles, ed., Ethics and
Populatwn (Cambridge: Schenkman, 1976), pp. 100--18; and Jeff McMahan,
"Problems of Population Theory," Eth1cs 92, no. 1 (October 1981): 96-127.
41. See McMahan, "Problems of Population Theory," part 7, and Jan Narve-
son, "Future People and Us," m Stkora and Barry, eds., OblJgatwns to Future
Generations, pp. 38-60, espectally pp. 47-48. Recall the YJdd1sh JOke cited by
Noz1ck (Anarchy, p. 337): "Ltfe IS so ternble It would be better never to have
been conceived. Yes but who IS so fortunate? Not one m a thousand."
42 E Q U A L 0 P P 0 R T U N I T Y A N D T H E FA M JL Y

included among the others, in the phrase "prevent harm to

others," as the motivation for justified government interfer-
ence.
So children can be incorporated into the protected self-
regarding sphere among consenting adults provided that due
care is taken to ensure that they are not significantly harmed
as a result. It is worth noting that this protected self-regarding
sphere is much broader than the area of negative liberty in-
volved in our argument. For as defined, it would include public
consensual acts that affect others as in the phrase "an affecting
sight."42 Displays of pornography around Times Square might
be protected within the sphere defined so broadly. While a good
case can be made for the protection of such behavior, 43 it adds
further controversy to the argument and need not be defended
here. So let us single out a less controversial portion of the self-
regarding sphere of liberty as follows:
THE PRIVATE SPHERE m LIRERTY" So long as no one is severely
harmed, intimate consensual relations should be im-
mune from coercive mterference.
By "intimate" relations, I mean those personal interactions
that are both private and affective. By "private," I mean interac-
tions that require the option of privacy in that their character
would be changed if they were coercively subjected to public
scrutiny and/or interference. The option of privacy, in other
words, is part of the freedom of interaction that would be
protected by this principle.
Relations between adults and children in a given family are
among the intimate consensual relations protected by this
private sphere of liberty. The latter is broader, of course, in that
intimate relations in addition to family relations would be
protected by it as well. 44 It would not be an exaggeration,

42. Sec Bnan Barry, Pollllcal Arl{ument. pp. 63-64, 71-72.

4.~. David A. J, Rtchards, The Moral Cnt1cism of Law IEncmo, Calif.:
Dickenson Publishmg Co., 1977), chapter.~.
44. Sec, for example, the classtc debate between Hart and Devhn. H. L. A.
 E Q U A L 0 P P 0 R T U N I T Y A N D T H E FA M I L Y 43

however, to conclude that the portion of the private sphere of

liberty that is least controversial and of greatest importance to
most of us is, without doubt, the part singled out by the
autonomy of the family. Hence, it is this portion that I will
employ in my argument below. Were this portion of the private
sphere of liberty to be sacrificed, the realm of negative liberty
remaining would be mutilated virtually beyond recognition.
The argument in the following chapter could be formulated
either in terms of the autonomy of the family or in terms of the
broader claim defined by the private sphere of liberty. And the
violations of the autonomy of the family to be investigated
there must also be violations of the private sphere of liberty. 45
Because the narrower principle is less controversial, however,
and because it also identifies the issues at stake more precisely,
I will employ it. It brings into sharper focus the conflicts
between conventional morality and systematic liberal theory
and, hence, will lead us more easily into the methodological
issues to be raised later in the book in chapter 5.

Hart, Law, L1berty and Morality (Oxford: Oxford UmvelSlty Press, 1963), and
Patnck Devhn, The Enforcement of Morals (Oxford: Oxford University Press,
1965).
45. In other words, the consensual relations between adults and ch1ldren
protected by the autonomy of the family assumption are among the mumate
consensual relations protected by the pnvate sphere of hberty. Lack of the
"essential prerequ1s1tes" m the former prmciple can be treated as the kind of
"severe harm" menuoned in the latter. Prov1ded that we restnct the discussion
to harms mvolvmg ch1ldren, the acts protected by fam1ly autonomy would all
be protected by the broader pnnc1ple of the pnvate sphere of hberty.
3. THE TRILEMMA OF
EQUAL OPPORTUNITY

T fined
he three principles just de-
combine to form a tri-
lemma-realization of any two can be expected to preclude
realization of the third. The resulting options are each disturb-
ing in that each involves a systematic and severe sacrifice in
one of the three principles. Before proceeding to develop this
argument, some ground rules for the discussion should be
established.

3·1 Ideal Theory L

et us embark upon the some-
what utopian task of imagin-
ing the best possible conditions that might realistically obtain
in a modern industrial society, both in terms of objective
economic and social circumstances and in terms of human
cooperation. Following Rawls, let us say that there is only
"moderate scarcity" in this society and that there is "strict
compliance" with principles of justice. In this dual sense,
Rawls's characterization of the problem of "ideal theory" pro-
vides a useful starting point. 1 Later, we will have reason to

1. I take the term "1deal theory" from Rawls, Theory of fu.~tice. p. 24S
"Stnct comphance" refers to support for institutional chmces land then
corresponding pnnc1ples); 1t does not carry w1th It umversal compliance w1th
every act of mdiVidual chmce land behav10r) that instituuons m1ght prescnbe
for theu Citizens. The latter, more amb1Uous assumpuon would render the
argument clearly utopian. Suppose, for example, that we were s1mply to
assume that every md!v!dual worked as hard as he docs now, despite a
perfectly equahzmg tax system, because pohcymakers would hke the norm of
social duty contnbuuons to become widely accepted m the society. To mtcr·
pret stnct compliance so broadly as to yield umversal conformity to all the
hypothesized mdiv1dual chmces would be to beg all the mterestmg 1ssues

44
 THE T R I L E M M A 0 F E QUAL 0 P P 0 R TUN IT Y 45

discuss critically his particular proposals for our central prob-

lem of equal opportunity.
By "moderate scarcity" Rawls means:
Natural and other resources are not so abundant that
schemes of cooperation become superfluous, nor are con-
ditions so harsh that fruitful ventures must inevitably
break down. While mutually advantageous arrangements
are feasible, the benefits they yield fall short of the de-
mands men put forward. 2
In other words, the question of justice arises under these
admittedly favorable conditions because conflicting demands
cannot be fully satisfied; yet this degree of scarcity is far from
extreme. This might be specified further as some assumption
that conditions well beyond subsistence are possible for every-
one in the society.
This condition of moderate scarcity still requires some fur-
ther specification to be useful for our purposes. I will interpret
it as imposing what might be called a realistic budget con-
straint. Resources are not unlimited; the resources devoted to
any particular worthy goal must compete with other legiti-
mate demands. In the next few sections I will focus on only a
part of the question of justice-the issue of assignment. There
are other components of the theory of justice and other legiti-
mate moral demands quite apart from those specified by the
theory of distributive justice. Hence, while the resources de-
voted to implementing any of the principles discussed below
may be substantial, they must compete with other legitimate
demands-health care, national defense, care for the aged,
environmental protection, international redistribution or for-
eign aid, and a viable and fair system of criminal justice (the

about the realistic possibilities for such a proposaljmcenuve effects, mdoctn-

nauon, and so forthl. Hence, stnct compliance Will be mterpreted in terms of
institutional chmce. Realistic problems of enforcmg mdividual conformity
would still have to be faced withm ideal theory as thus defined.
2. Ibid., p. 127.
46 THE TRILEMMA OF EQUAL OPPORTUNITY

latter falls within retributive rather than distributive justice).

Improvements in the realization of our three principles might
be worth sacrifices in any of these other worthy goals. Such
improvements, however, would not merit substantially sacri-
ficing all of them. 3 There are, in other words, certain outer
limits on the resources that can be devoted to any of our three
principles, limits set by the fact that the government and other
social institutions face a host of other legitimate moral de-
mands that are separable from the particular policy area we are
focusing on here, the issue of assignment. These limits consti-
tute a realistic budget constraint. While they are vague at the
level of general application necessary for our argument, they
will permit us to rule out some proposals as clearly unrealistic
or utopian because their prohibitive expense would surpass
any realistic construction of the budget constraint. For exam-
ple, it would violate such a constraint to supply every student
in society his own full-time team of elite private tutors com-
parable to the best talent available in private education. 4 Pro-
vided that our use of the budget constraint is limited largely to
ruling out such clearly fantastic examples, its admitted vague-
ness will not affect the development of our argument.
In addition to moderate scarcity, let us assume "strict com-
pliance": "Everyone is presumed to act justly and to do his part
in upholding just institutions."5 More specifically, we might
suppose that there is good will, cooperation, and conscientious
agreement on implementing the principles under discussion.
Furthermore, we might assume that such good will is not only
operative at the time principles are implemented but that it
has also been characteristic of the relevant recent history. This
would conform to Rawls's assumption that "strict compliance
theory" does not have to deal with questions of "compensa-
tory" justice. 6
3. The argument below w1ll requue only the acceptance of some general
upper hmit. Because we will employ 1t only to rule out certam clearly utopian
proposals, ItS admitted vagueness w1ll not affect the argument.
4. See sectiOn 3.6 below.
5. Rawls, Theory of fustlce, p. 8.
6. Ib1d.
 T H E T R 1L EM M A 0 F E Q U A L 0 P P 0 R T U N IT Y 47

This notion of strict compliance will structure certain

ground rules in the argument below. My claim will be that
achievement of any two principles realistically precludes
achievement of the third. Hence, I will assume in this argu-
ment that the two principles under discussion have been
conscientiously implemented not only at the time of discus-
sion, but also in the relevant recent history. In addition, I will
assume that the remaining principle has also been the subject
of conscientious effort and cooperation within the limits re-
sulting from the comparatively more successful efforts to
implement the other two.
These conditions, moderate scarcity and strict compliance,
rule out some of the more obvious cases under which a serious
sacrifice in one or more of our three principles might be
acceptable. For example, under conditions of extreme (rather
than merely moderate) scarcity, it is commonly granted that
the private sphere of liberty might have to be seriously in-
fringed upon in order to regulate population growth. Similarly,
when there has not been a history of strict compliance, but
rather, a legacy of injustice to specific groups, such as Ameri-
can blacks or native Americans, there is a compensatory argu-
ment that could not otherwise arise for abridging the principle
of merit. Lastly, when instead of strict compliance, there is
organized opposition from entrenched power groups, it is pos-
sible that attempts to equalize life chances might not be worth
the costs of reform. While each of these sacrifices in one of our
three principles might be understandable and necessary under
the conditions of non-ideal theory, they would constitute a
more disturbing challenge to liberal assumptions if they were
required under the ideal conditions of both strict compliance
and only moderate scarcity. The general argument that follows
will confine itself to such ideal conditions.

.. Throughout the modern world,

3.2 Background Inequallt1es equality is generally pre-
scribed, yet inequality is generally practiced. Assertions of
moral equality lie at the root of all the ideologies commonly
48 THE TRILEMMA OF EQUAL OPPORTUNITY

accepted in the developed countries. 7 Yet all of these countries,

whether capitalist or socialist, whether democratic or hierar-
chical lor some hybrid combination of these) have very sub-
stantial economic and social inequalities. Some of the state
socialist systems of Eastern Europe have, at times, made im-
pressive gains in equalizing opportunities across class barriers.
Nevertheless, they remain highly stratified economically and
socially.R Similarly, no proposals for redistribution that have
been seriously entertained in any of the Western capitalist
systems would come remotely close to equalizing outcomes. 9
The deprivations of poverty at the bottom might be reduced by
such proposals and inequalities might be limited somewhat;
they would not, however, be eliminated.
Traditional societies, furthermore, have rarely been distin-
guished by any substantial degree of equality-economic or
social. Rather, they have only been more forthright in their
acceptance of inequality. They have more explicitly matched
inegalitarian ideologies to highly stratified outcomes. 10

7. See Douglas Rae et al., f.qualwes tCamhndge. Harvard Umversny Press,

1981), and Sanford Lakoff, Equality m Po/Jtical Philosophy tCambndge· Har-
vard Umversuy Press, 1964 ).
8. Charles E. Lmdblom, Po/liles and Markets: The World"s PolJtico-Eco-
nomJc System.~ tNew York: Bas1c Books, 1977), chapter 20. See also Dav1d
Lane, The End of Jnequal!tyl .'itrallficatmn under State Soc1a1Jsm (Har-
mondsworth, Eng.: Pengum Book, 1971), and Frank Parkm, Class Inequality
and Po/Jtical Order (New York. Praegcr, 1971). For claims about progress m
equal opportumty m state soc1ahst systems, see Parkm, chapter 5, especially
sectmn 4.
9. See, for example, Arthur M. Okun, Equal1ty and Efficiency: The 8Jg
Tradeoff (Washmgton, D.C.: The Brookmgs InstitUtiOn, 1975), chapter 4, and
Lmdblom, Politics and Market.~. chapter .t For an admittedly utopian but
more stnctly egahtanan proposal, see Joseph H. Carens, Equality, Moral
Incentive.~ and the Market· An h~ay m Utop1an Pol!tJco-EconomJc Theory
tCh1cago: Umvcrs1ty of Ch1cago Press, 19811. Carens, however, self-con-
scmusly avmds the problem of "transition" from the present Amencan policy
context to h1s utopian system. It 1s not, in other words, offered as a program we
might now try to Implement. It 1s hmned to the important quesuons of 1deal
theory. Sec sectmn 4.4 below. •
10. See, for example, Bamngton Moore, Jr., Imusuce. The Soc1al Bases of
 T H E T RILE M M A 0 F E Q U A L 0 P P 0 R T U N IT Y 49

If any modern economy must organize itself either through

the use of market incentives or through the use of command or
hierarchy, it should be clear that substantial inequalities, both
economic and social, always accompany both of these basic
mechanisms. Hybrid systems, furthermore, have never offered
any escape from the problem. Large-scale experiments with
market socialism, on the one hand, and democratic socialism,
on the other, have always, whatever their other merits, been
subject to the same charge. 11
The division of labor necessary for a complex, technologi-
cally advanced economy may, itself, provide one of the root
causes of both economic and social inequality. For it may carry
with it a host of social, psychological, and economic differenti-
ations that provide the basis for a hierarchy of status or esteem.
Such advanced economies may also require economic differen-
tiations to support the necessary investments in skill and
training and to motivate performance. Even those socialist
systems (in less developed countries) that have attempted to
replace economic incentives with "moral" ones have intro-
duced social inequalities as a result. 12
These observations all support my assumption, in the argu-
ment that follows, that any large-scale modern industrial soci-
ety will have substantial social and economic inequalities. In
Obed1ence and Revolt (Wh1te Plams, N.Y.· M E. Sharpe, 197R), and Louis
Dumont, Homo H~erarchicus: The Caste System and Its lmphcatwns (Chi-
cago: Umvemty of Ch1cago Press, 1980).
II. For a discussion of markets and h1erarch1es in alternatiVe pohuco-
economlc systems, see Lmdblom, PolJllcs and Markets.
12. For the complex system of pomts and social d1stmctions employed for a
ume m Cuba, see Robert M. Bernardo, The Theory of Moral Incentive.~ in
Cuba (Umvers1ty, Ala.: Umvemty of Alabama Press, 1971), chapter 3. For
arguments about the mevltabllity of inequality because of the diVIsion of labor
and various soc1al differenuatwns, see Lane, The End of Inequality!; Ralf
Dahrendorf, "On the Origin of Soc1al Inequality," m Laslett and Runciman,
eds., Ph1losophy, Polit1cs and Society, Second Senes, pp. RR-109, and the
debate between Melvm Tumm and W1lbert Moore in the Amencan SocwlDKl·
cal Review 2R ( 1963): 19-28. The latter concerns the extent to wh1ch a variety
of separable mequalities must amount, m total, to a system of soc1al straufica-
twn.
50 T H E T RI LEM M A 0 F EQ U A L 0 P P0 RT U N I T Y

additiOn to the empirical evidence that it would be unrealistic

to assume otherwise, there is a normative purpose served in
applying this assumption. It has been characteristic of liberal-
as distinct from radical-political theory to advocate equality
of opportunities rather than equality of outcomes. 11 If we were
to assume that strict equality of outcomes were achieved, then
there would be little point in worrying about the fate of equal
opportunity. Equal chances for equal outcomes would follow
in an empty and formal sense from the mere fact of equal
outcomes. As strict equality of outcomes is approached, the
salience of equal opportunity as an issue can be expected to
subside accordingly.
The argument that follows should be thought of as a test of
the coherence and viability of this liberal focus on equal
opportunity. Given the commitment of liberal theory to lib-
erty, can its reliance on equal opportunity as an ultimate
strategy of legitimation for distributional questions be main-
tained, even under ideal conditions? Hence, my focus will be
on liberal theories that are committed to equal opportunity but
not to equality of outcomes. Later I will explore the implica-
tions of relaxing this assumption. 14
In the trilemma to be developed below, both principles of
equal opportunity can be maintained only through severe
sacrifice in the autonomy of families. But this argument will
depend on the assumption just defined-that there arc substan-
tial social and economic inequalities in the society. If this
assumption were relaxed, the conflict among our three princi-
ples would disappear. It is worth emphasizing the role of this
assumption in the argument since it brings into focus a more
general claim-that liberal theory, if taken seriously, must be

13. Sec n. 5, chapter l. Also, see Lakoff, Equality. chapter 6, John Schaar,
"Equality of Opportumty and Beyond," m J. Roland Pennock and John W.
Chapman, cds., Nomos IX, Equality jNew York: Atherton Press, 19671, pp.
228-50; and Michael Young's fantasy, The Rise of the Mentocracy iHar-
mondsworth, Eng.: Pengum Books, 1961).
14 Sec sccuons 4 4 and 5 I.
 THE TRILEMMA OF EQUAL OPPORTUNITY 51

far more radical in its implications than has been imagined by

its major proponents. For it would be difficult to imagine a
defensible liberal ideal of equal opportunity that was not
committed to both of the principles defined here (merit and
equal life chances). Yet any such doctrine would require either
a systematic sacrifice in the autonomy of the family or a
systematic achievement of equal results (rather than merely
opportunities) throughout the society. Either of these two
latter options would be far more radical, and far less comfort-
able in its implications, than any doctrine of equal opportunity
seriously advocated in liberal theory. It is worth adding that I
hope to subject liberal theories to this kind of critical scrutiny,
not because I think liberalism is bankrupt, but because I
believe that any future, viable version of liberal theory must
face up to certain hard choices. 15 The trilemma of conflicts
arising for equal opportunity will provide a starting point for
that kind of re-examination.

. . under the ground rules just

3.3 The Fmt Optwn specified, implementation of
any two of these principles will render achievement of the
remaining one virtually impossible. For example, the first and
third principles rule out the second. The autonomy of the
family protects parental efforts to influence the development
of their children. Given background conditions of inequality,
children from the higher strata will have been systematically
subjected to differential developmental opportunities that can
reliably be expected to advantage them in the process of
meritocratic competition. Under these conditions, the princi-
ple of merit-applied to talents as they have developed under
such unequal conditions-becomes a mechanism for generat-
ing unequal life chances.
It is just as if the continuing inequality of life chances in the
15. This exammauon can be viewed as a contmuation of the one I began m
Tyranny and Legitimacy. For a general characterization of liberalism see
section 5.1 below.
52 T H E T R IL E M M A 0 F E Q U A L 0 P P 0 R T U :'\ I T Y

warrior society example came about through parental influ-

ence on talent development. If the autonomy of the family
protects the process whereby parents influence their children,
and if the principle of merit is employed, no matter how
scrupulously, to select the best warriors, these unequal condi-
tions for talent development-when combined with equal con-
sideration for talents as developed-produce unequal life
chances. Children of the present warriors can be expected to
win in the competition disproportionately.
Furthermore, if the relevant recent history involves consci-
entious efforts at strict compliance, then the principle of merit
will itself tend to exacerbate these unequal conditions for
talent development. For parents from the higher positions will
be systematicaly distinguishable by their greater skill, their
competence, and their familiarity with all the desirable charac-
teristics that are taken to constitute "qualifications" in that
society. When the process of talent development is protected
by the autonomy of the family, advantaged parents can be
expected to have systematically greater success inculcating
these characteristics among their children.
Consider, in our society, all of the advantages that families
from the higher strata can give to their children: private
schools, culture in the home, a secure home environment,
trips abroad, private lessons, an advantaged peer group, and
successful role models. These are only the most obvious exam-
ples of differential developmental opportunities affecting both
competence and motivation so as to give children from the
higher strata systematic advantages in any system of merito-
cratic competition compatible with our second principle. 16
Consider the overwhelming effect of family background on
life chances that emerges from Jencks's study of the determi-
nants of economic success among men in America:

16. See Jencks, Inequality, pp. 29-32, 76-81, 89-92, 135-246, Kenneth
Kemston, All Our Children: The Amencan Family Under Pressure (New
York: Harcourt Brace Jovanovich, 1977), chapter 2, and Jencks, Who Gets
Ahead!, chapters 3 and 12.
 THE TRILEMMA OF EQUAL OPPORTUNITY 53

If we define "equal opportunity" as a situation in which

sons born into different families have the same chances of
success, our data show that America comes nowhere near
achieving it. If, for example, an omniscient social scientist
were to predict the economic standing of sons from differ-
ent families he would find that sons from the most favored
fifth of all families had predicted Duncan scores [a scale of
occupational status running from 0 to 96J of about 64,
while sons from the least favored fifth of all families have
predicted scores of about 16Y
As Jencks notes, this is about the difference in status between a
social worker and an auto mechanic. Or, if the outcomes are
evaluated in terms of income,
the sons of the most advantaged fifth could expect to earn
150 to 186 percent of the national average, while the sons
of the least advantaged fifth could expect to earn 56 to 67
percent of the national average. 18
Of course, evidence from American society is only indirectly
relevant to our problem of ideal theory since the other princi-
ple under discussion, the principle of merit, is only partially
instituted in this society. Jencks, however, usefully identifies
five basic causal mechanisms that generate unequal life
chances in America; it is worth noting that all but one of them
would be compatible with the fullest institutionalization of
the principle of merit:
1. Men from advantaged backgrounds have cognitive
skills that employers value.
2. Men from advantaged backgrounds have noncogni-
tive traits that employers value.
3. Among men with similar cognitive and noncognitive

17. Jencks, Who Gets Ahead!. p. 82.

18. lbtd., p. 83.
54 THE TRILEMMA OF EQUAL OPPORTUNITY

traits, those from advantaged families have more educa-

tional credentials. Employers appear to value these creden-
tials in their own right, even when they are not associated
with measurable skills or behavior.
4. Among men with similar skills and credentials,
those from advantaged families seek jobs in higher status
occupations than those from disadvantaged families.
5. Even among men with similar skills and credentials
who enter the same occupation, employers seem to pay
men from advantaged families slightly more than men
from disadvantaged families. 19
The cognitive skills, noncognitive traits, and educational
credentials referred to in mechanisms I to 3 can all count as
qualifications compatible with our principle of merit. 20 Mecha-
nism 4, the motivation to aspire to higher positions, is equally
compatible with the principle. Only the fifth mechanism !un-
less it depends on factors of performance or motivation un-
measured in Jencks's data) might be interpreted as discrimina-
tion in terms of family background and, hence, as unacceptable
in a society attempting to institute the principle of merit.
While empirical conditions will vary from one society to the
next, the general proposition is difficult to deny: children from
advantaged families in a given society will have greater oppor-
tunities to develop the skills, credentials, and motivations
valued in that society. If the autonomy of the family protects
the process by which parents provide those greater opportuni-
ties, and if the principle of merit sorts people accurately in
terms of their skills, credentials, and motivations-as devel-

19. Ibid., pp. 70-71.

20. However, to the extent that credentials do not support the JUStificatory
burden of prcd1ctmg competence or motivation for performance m g1ven
pos1t1ons, they should not be used as cntena, compatible w1th the pnnc1ple of
ment, for assignment to those pos1t1ons. See Jencks, Who Gets Ahead!, p. 184,
for a d1scussmn of the overreward g1ven the B.A. degree. Hence, some aspects
of the operation of mechamsm .~ m the quoted passage would not be compati-
ble With the pnnc1ple of ment.
 THE TRILEMMA OF EQUAL OPPORTt:NJTY 55

oped under those unequal conditions-then systematic in-

equality of life chances will result.l'

3.4 The Second Option S

uppose one were to keep the
autonomy of the family in
place but attempt, nevertheless, to equalize life chances? Insti-
tutionalizing the last two principles in this way would require
sacrifice of the first, the principle of merit. Given background
conditions of inequality, the differential developmental influ-
ences discussed in the last section produce disproportionate
talents and other qualifications among children in the higher
strata. If they must be assigned to positions so as to equalize
life chances, then they must be assigned regardless of these
differential claims. "Reverse discrimination" in favor of those
from disadvantaged backgrounds would have to be applied
systematically throughout the society.
By reverse discrimination, I mean any procedure of assign-
ment that consistently places crucial weight on some charac-
teristic other than qualifications (as defined earlier). 22 If qualifi-
cations have been appropnately defined, in a sense sufficiently
relevant to performance, then widespread assignment of those
who are less qualified can be expected to have a great cost in
efficiency. 23 It would also have a substantial cost in fairness.
21. The vJsJbJhty given to extreme upward moh1hty should not obscure
such statistical proposJtJons. As Harold Lasswell noted m h1s 19.36 classic,
Politics: Who Gets What. When. How (New York: Mcnd1an Books, 19SRI:
"Although any bnght and talkattvc lad m the Umted States may be told that
one day he may be president, only eight boys made It m the last generatiOn."
jp. 14). Lasswell then goes on to note the hm1tcd numbers of persons who
actually make it to other highly valued posltwns, despite the wtdcly shared
character of the aspiration for those posJtJons.
22. Such a procedure would d1scnmmatc agamst those who arc more
qualified and m favor of those from disadvantaged backgrounds. For some of
the many contested uses of reverse discnmmatwn sec Barry R. Gross, lJiscnm·
mation m Reverse: Is Turnabout Fa11 Play! !New York: New York Umvcmty
Press, 1978), chapters 2 and 3. Sec also sectiOn 3.7 below.
23. See Damels, "Merit and Mentocracy" and Goldman, fustice and Re·
verse Discnmmatwn. chapter 2, for the efficiency goals served by ment. The
systematic sacrifice of merit considered here would sacrifice efficiency for
56 T H E T R IL E M M A 0 F E Q U A L 0 P P 0 R T U N I T Y

For the analogue of a competition in which skill and effort

provide a sense in which people earn or merit their payoffs
would no longer apply.24
Now the sacrifice in efficiency might be avoided if on-the-job
training could make up entirely for deficiencies in earlier
preparation. It is hard to believe, however, in a modern indus-
trial society, with a complex differentiation of tasks, that
qualifications that are performance-related could not be de-
fined so as to predict better performances. Perhaps, in some
actual societies, the definition of qualifications has not carried
this justificatory burden. In those societies, apparently merito-
cratic competition has amounted to no more than an empty
credentialism. That is just to say, the principle of merit has not
actually been implemented in those cases, 25 for the principle
required an appropriate competition in terms of qualifica-
tions-criteria that are job-related in that they fairly can be
interpreted as indicators of competence or motivation for an
individual's performance in a given position. To assign persons
to positions regardless of qualifications in this sense is, by
definition, to assign persons who are not as likely (as those
more qualified) to perform well-insofar as that matter can be
judged in terms of any prior indicators of talent or motivation
as these might be displayed in a suitable competition. If there
is a system of significant on-the-job training, then that must be
taken into account in the prior definition of qualifications.
Those who are most likely to profit or improve their perform-
ance from such training have to be given correspondingly
"arrays" of fob assignments (as m Damels's ment pnnciplc) or for mdtvidual
job assignments (as m Goldman's formulation)
24. For the appeal of thts model of a fair competition, embodied m the tdcal
of pure procedural JUStice, sec Rawls, Theory of fustJce, sections 12, B, 14. It IS
worth notmg that any msutuuonahzauon of such a fau competition can be
expected to leave a substantial role for luck or chance. Thts ts not wornsomc
so long as luck is not systematically maldtstnbuted by class, race, or cthmc
background.
25. For an inflated clatm that thts kmd of credenuahsm has been the rule
rather than the exception in apparently mentocrauc assignment sec George
Gilder, Wealth and Poverty (New York: Bastc Books, 1981 I. chapter B.
 THE TR ILEM MA OF EQUAL OPPORTUNITY 57

greater consideration. Demonstration that one will be more

educable on the job should itself be considered a kind of
qualification. .
I can think of one imaginary scenario that appears, at first
glance, to evade this argument. Suppose there were a lottery
system at birth that randomly assigned babies to families. The
system of differential opportunities for talent development
could be kept in place and the system of meritocratic assign-
ment for talents, as developed under those conditions, could be
fully instituted. The example is of interest because equal life
chances, in a quite precise sense, would result from the ar-
rangement. For the random system of assignment of newborn
babies to families would serve to equalize life chances when
judged from the perspective of newborns before the lottery.
Any newborn infant's chance of reaching any highly valued
position would be precisely equal to that of any other newborn
infant. Should this admittedly bizarre and unpalatable arrange-
ment be considered a relevant counterexample? Does it fulfill
all three principles simultaneously?
The example does not overturn the argument, I believe,
because it depends on a violation of the autonomy of families.
The liberty of parents to raise their own children would have
been interfered with coercively. The state would have forcibly
determined which children are to be raised by which parents.
By our initial principle, the liberty of parents to raise their own
children can be interfered with only in order to ensure for
children the essential prerequisites for adult participation in
the society. While removal of a child from its parents might be
justified then in some isolated cases of extreme deprivation (in
order to shield the child from clearly harmful developmental
conditions), such a wholesale process of reassignment clearly
could not rest on such a basis. While some children might be
benefited by the changes, others would be harmed. 26

26. For a careful appraisal of the many risks m any reasstgnment of chlidren
to famthes see Goldstem et al., Beyond the Best Interests of the Ch1ld, part 2.
58 THE TRILEMMA OF EQUAL OPPORTU:\IITY

Perhaps some of this harm might be avoided through decep-

tion. Reassignment, however, even when coupled with decep-
tion in order to minimize experienced harm, 27 would still
constitute coercive interference violating the autonomy of
families. In defining family autonomy I assumed that natural
parents have the mitial option of creating families (subject to
this principle) if they do so in a consensual manner, that is, one
not subject to sustained and intense disagreement. This does
not mean, of course, that parents may freely contribute to
unrestrained population growth. For under some empirical
conditions, such growth would endanger the prerequisites for
childrens' future participation as adults in the society and,
hence, may be coercively restricted in accordance with the au-
tonomy principle. 2R Furthermore, this principle would even leave
room for the state to intervene to prevent certain persons from
procreating when it is clear that their offspring could not
reasonably aspire to even the essential prerequisites for adult
participation in the society (because of extreme disabilities or
deformities). This principle would not require such govern-
ment intervention, but it would place such procreation outside
the sphere of action protected by the autonomy of families. 29
In including within family autonomy the "initial option" of
natural parents to create families, I explicitly identified the
basis for rejecting the lottery system. The autonomy of fami-
lies, if it is to be meaningful, must not only protect fully
formed families, it must also protect the process by which they
come into being. Wholesale reassignment of children to par-
ents, or wholesale intrusion into the process by which parents
consensually choose each other would violate the autonomy of

27. By expenenced harm I mean harm withm the subjective awareness of

the persons mvolved. They may also be harmed, m the sense of theu interests
bemg sacrificed, Without their ever knowing it. For a further cnuque of lottery
proposals see sectwn 4.2 below.
28. See Garrett Hardm, "The Tragedy of the Commons" and Michael D.
Bayles, "Limits to a Right to Procreate," m Bayles, Ethics and Populatwn
29. I have m mmd a case such as Tay-Sachs d1sease, where the harm to the
possible child is both extreme and predictable.
 THE T RILE M M A 0 F EQUAL 0 P P 0 R TUN IT Y 59

families as much as would intrusion into established fami-

lies.3o Perhaps such a bizarre lottery scheme would realize
equal life chances and the principle of merit. But like other
systems that achieve these two principles, it would violate the
remaining one, the autonomy of families.
There is a sense in which this scenario literally realizes the
familiar metaphor of the "lottery" of birth. If that metaphor is
taken seriously, one wonders how important it would be to
actually undertake the random process of assignment to par-
ents after birth. It is this latter process, of course, that violates
the autonomy of families. But this metaphor of a "lottery of
birth" depends on the relevance of an imaginary perspective
from which we can coherently view ourselves as potential
persons. 31
Rawls's famous device of the "original position" depends, of
course, on the coherence and viability of such an imaginary

30. I am indebted to James Bt!bngton for pomting out that the Soviet Umon
has used wholesale reassignment of children to suppress rebgwus sects such as
the Pentecostalists. The autonomy of the family defines an often unrecogmzed
condition for the transmission of rebgwus and ethmc pluralism from one
generation to the next
31. The potential persons mterpretatwn creates anomahes for populatiOn
policy. See R. M. Hare's cntique of Rawls, based on the mterests of potential
persons m actually commg to exist, m "Rawls's Theory of Justice," Norman
Daniels, ed., Readmg Rawls: Critical Studies of A Theory of {ustJce (New
York: Basic Books, 1975), pp. 31-107, especially pp. 97-101. Hare argues that If
persons m the origmal positiOn were not already sure of a place m society, theu
first interest would be to require procreation on a mass scale m order to
maximi~e theu chance of commg to be born We might add that th1s anomaly
could be avoided by an assumption that all members of the ongmal positiOn
know that they are among the persons who happen to exist m the society. This
strategy though would be subJect to the problem that the contmuation of the
human race would not raise issues then that could be considered Withm the
framework of the original posltlon. If a given generation decided not to
procreate, persons in the ongmal posuwn (choosmg, out of self-mterest, by
hypothesis) would not have grounds for objection smce they are to assume that
they already have a place among the extant generations. Rawls has argued that
individual interests continue over two generations, but this somewhat myste-
rious assumptiOn has not been interpreted so as to produce a duty or require-
ment that one generation create the next. See Hare, "Rawls's Theory," p. 97,
and Rawls, Theory of Justice, p. 128.
60 THE TRILE.'\t.'\\A OF EQUAL OPPORTUNITY

perspective. Yet, we can grant the relevance of a hypothetical

construction of ourselves as potential persons (before we actu-
ally enter into existence) for certain theoretical purposes but
not necessarily for others. When combined with certain other
assumptions, it might offer a useful account of fairness for
hypothetical choice (as in Rawls's original position), without
also constituting the basis for a substantive principle of equal
opportunity. The principle of equal life chances requires some
actual, not hypothetical, benchmark from which our expecta-
tions are equalized. In this respect, it can be sharply distin-
guished from the original position. 12
Accepting the hypothetical benchmark for the principle of
equal life chances would trivialize the principle, for any soci-
ety, no matter how stratified, oppressive, and inegalitarian,
would embody equal life chances in this merely formal sense.
If we can imagine ourselves reasoning about our life chances
before we come into existence in a given society, there is a
coherent sense in which we might think that our chances of
being any given person are randomized:B Hence, my chances of
being lord of the manor or a serf on a feudal estate could be
thought of as the same as anyone else's from this perspective. 34
Such a feudal society could then be interpreted as equalizing
life chances in this abstract, hypothetical sense. The principle
interpreted this way would clearly have entirely different im-

.U. See the d1scusswn of the onginal posmon and other moral deciSIOn
procedures in section 5.2. Sec also n. 34 below .
.B. In the absence of further information, we can employ the pnnc1ple of
msufficient reason to treat the vanous possibilities as equally hkely. As Rawls
admits, there 1s then an argument that the expected value of our payoffs will be
maximized by a society orgamzed according to the pnnc1ple of average uuhty.
Sec Rawls, Theory of /us !lee. pp. l(iS-66.
34. Rawls offers the example of a slaveholder who cites gambling from the
ongmal pos1t10n as a justification for a practice of slavery (conformmg to the
pnnc1plc of average uuhty). "In the imtial contractual Situation he would
choose the average principle even at the nsk of 1ts subsequently happemng
that he IS justifiably held a slave" (p. 167). Rawls's cla1m IS that the gambhng
argument 1s obJectionable, although the hypothetical character of the agree-
ment IS not.
 THE TRILEMMA OF EQUAL OPPORTUNITY 61

plications from the principle requiring equal life chances from

a benchmark identifying actually existent persons. In this
latter version, systematic implications follow for social mobil-
ity throughout the society. The core of its appeal would be
sacrificed then by the alternative version.
Hence, a lottery among newborns, while it might succeed in
equalizing life chances, would do so at the expense of the
autonomy of the family and its connected private sphere of
liberty-in particular, the liberty of parents to raise their own
children. If the autonomy of families remains undisturbed,
given background conditions of inequality, unequal develop-
mental influences will produce a differential development of
qualifications among those from the higher strata. These con-
ditions force a hard choice between equality of life chances, on
the one hand, and the principle of merit, on the other. The first
option explored in the last section sacrificed the former. What I
am calling the second option, outlined in this section, sacri-
fices the latter. Only a process of assignment that was applied
regardless of meritocratic factors could equalize life chances
despite differential talent development. Such a system of re-
verse discrimination, 35 systematically applied across the soci-
ety, would appear inevitably to conflict with both efficiency
and fairness.
We might imagine another hypothetical system that could
conceivably ameliorate some of these conflicts. Suppose a
nationwide lottery for job assignment were instituted not
among newborns but among fully developed adults. The pri-
vate sphere of liberty would remain in place with all of its
effects on the unequal development of talents and other qualifi-
cations. Life chances would be equalized since the lottery

35. See n. 22 above and the dtscussion m the text. I am using the term
"reverse discrimination" more broadly than some defimtions that would
restrict it to preferential treatment in favor of spectfied groups that had
prevtously been dtscnmmated agamst. For an example of this more restncttve
defimtton, see the "working defimtion" m Barry Gross, ed., Reverse Discnml·
natwn !Buffalo, N.Y.: Prometheus Books, 1977), p. 3.
62 T H E T R IL E M M A 0 F E Q U A L 0 P P 0 R T U /'1: I T Y

would strictly randomize the process of assignment to posi-

tions. Again, the pnnciple sacrificed would be merit since
assignment would be accomplished entirely in disregard for
any qualification relevant to job performance.
Even though sacrificing meritocratic processes would di-
vorce the system of assignment from one kind of claim to
fairness, instituting the lottery would introduce another. Each
person would have an equal chance for any given position. No
one would have a special claim, based on any contingent
factors. Note that the principle of merit distinguished relevant
factors (qualifications) from irrelevant ones and based its claim
to fairness on insulating the system of assignment from influ-
ences by any of the irrelevant factors. In a sense, the job
assignment lottery would take this argument a step further by
simply considering all contingent factors-any basis for differ-
entiating one adult from another-as irrelevant to the process
of job assignment. In doing so, it precludes applicants from
claiming, in any sense, that they have "earned" or "merited"
their positions through skill or effort. The model of a fair
competition, when supplanted by a lottery, treats all persons as
indistinguishable recipients of an equal chance.
Of course, it is likely that the sacrifice in efficiency would be
severe. Perhaps it might be ameliorated through systematic on-
the-job training. As noted before, however, in a modern society
with complex task differentiation, it seems undeniable that
there would be a substantial cost in efficiency. Even if primary
training were on the job, applicants could surely be differenti-
ated in terms of their potential educability. If the private
sphere of liberty and background conditions of inequality were
maintained, then differential performance in a complex variety
of positions could be predicted. Therefore, compared to an
alternative system of merit, this system could be expected to
require a sacrifice in efficiency.
It might be objected that our presumption that parents will
differentially influence the talent development of their chil-
dren rests, in part, on the expectation that those talents will be
 THE TRJLEMMA OF EQUAL OPPORTUNITY 63

of use in the later process of job assignment. If a society

replaced meritocratic competition with a job lottery, perhaps
this new institutional background would affect the motivation
of parents to inculcate skills that eventually would be relevant
to job performance. Since we are raising an empirical question
about a hypothetical scenario, the answer is, to a large degree,
indeterminate. Parents might be imagined to switch their
focus to "well-roundedness," adaptability, and educability in
unexpected circumstances, and other characteristics that
might better serve their children in such a system. The essen-
tial point, however, is that regardless of the extent of the cost
in efficiency, the principle of merit would have to be sacrificed
by any such attempt to equalize life chances while maintain-
ing the autonomy of families.
It is also worth noting that a job lottery alternative to merit
would produce a conflict with liberty of a different kind. I
defined the autonomy of families and the private sphere of
liberty narrowly so as to clarify the role of the family in the
trilemma of equal opportunity. Another kind of liberty, which
falls outside the private sphere, is the liberty to seek and
compete for employment. This liberty might be formulated in
a strong way so that any interference with mutual consent
between a potential employer and employee is unjustified.
Even the principle of merit, in this view, might constitute
unjustified coercion. Or the interests of competing potential
applicants might be taken into account so that liberty to seek
employment land the corresponding liberty of employers to
seek employees) might be constrained by requirements of fair
competition like those specified by the principle of merit. This
is a weak formulation of the liberty of employment.36
Liberty of employment, even m this quite weak sense,
conflicts sharply with the job lottery proposal. While the
proposal would leave the private sphere intact in an effort to

36. See the discussiOn of hberty of employment under perfect equality in

section 4.4 below.
64 THE TRILEMMA OF EQUAL OPPORTUNITY

equalize life chances, it would produce conflicts with liberty in

another recognizable and important sense, even when tha,
liberty is defined modestly. If I want to be a pediatrician, yet
my lottery ticket requires that I become a corporate lawyer, or
if I want to be a violinist, yet my lottery ticket requires that I
become an auto mechanic, my liberty of employment, indeed,
my liberty to determine the basic direction of my life plan has
been severely constrained.
Since my focus here has been on the private liberties in-
volved in the family, I note this additional difficulty only in
passing. Within the framework of our more general discussion,
the point to note about this job lottery proposal is not its
sacrifice of liberty of employment, but rather, its sacrifice of
the principle of merit. Like all efforts to equalize life chances
while maintaining the autonomy of families intact, it would
sacrifice the third, remaining principle.

3.5 The Third Option Suppose one were to attempt to

equalize life chances while
maintaining the system of meritocratic assignment. Given
background conditions of inequality, it is the autonomy of
families that protects the process by which advantaged fami-
lies differentially affect the development of talents and other
qualifications in their children. Only if this process were
interfered with could both the principles of merit and of equal
life chances be achieved. In other words, if equality of life
chances is to be achieved through processes consistent with
the principle of merit, then conditions for the development of
talents and other qualifications must be equalized. Given
background conditions of inequality, this can be done only
through some mechanism that systematically insulates the
development of each new generation from the unequal results
achieved by the last. Coercive interferences with the family
would be required if advantaged parents were to be prevented,
systematically, from passing on cognitive, affective, cultural,
and social advantages to their children. Perhaps a massive
 THE TRILEMMA OF EQUAL OPPORTUNITY 65

system of collectivized child-rearing could be devised to

achieve such a resultY Anything short of such a large-scale
alternative to the autonomous nuclear family would probably
provide only an imperfect barrier between the inequalities of
the parental generation and the developmental processes af-
fecting its children.
From the communal child-rearing in Plato's Republic to the
test tube nurseries in Huxley's Brave New World, the replace-
ment of the family with some alternative strategy of child-
rearing has been the centerpiece of any social engineering that
required complete manipulation of human development. 38 As
long as the private sphere of liberty is in place, crucial develop-
mental factors are entrusted to the autonomous decisions of
families and are, by that very fact, insulated from social con-
trol. Whether the efforts at social engineering are aimed at
equalization or hierarchy, the family constitutes a crucial
barrier to the manipulability of the causal factors affecting
human development.
Whatever the precise institutional design, if developmental
factors are to be equalized, systematic intrusion into the au-
tonomy of families would be required. Recall that coercive
interference into consensual family relations can be justified,
according to the principles stated here, only to ensure the
essential prerequisites for adult participation in the society-
to assure a child's physical or mental health, his literacy, or his
knowledge of the necessary social conventions. This principle
defines a restrictive paternalistic burden that must be met if
coercive interference is to be justified. This paternalistic bur-

37. For a mtxed assessment of the comparatively small-scale Israeb expen-

ment with vanous forms of collectlvtzed chtld-reanng (among self-selected
groups), see Bruno Bettelheim, The Children of the Dream (London: Thames
and Hudson, 1969).
38 For an mterestmg argument that Plato's utopian proposals on the family
and child-reanng were meant to be taken senously, see Susan Moller Okm,
Women in Western Polltlcal Thought !Pnnceton: Pnnceton Umversity Press,
1979), part 1 and also the appendtx to chapter 2 (for a cnttque of Bloom's
vtews).
C.C. T H E T R I I. E ,\\ ;\\ A 0 F E Q U A I. 0 P P 0 R T U 1\i I T Y

den is obviously not met by strategies that simply level down

or equalize developmental influences m order to lessen the
advantages of children in the upper strata. The interests of
those children are not served at all by such efforts. And,
provided that scarcity is not so extreme that such intrusions
are necessary to ensure the essential prerequisites for children
from other strata, it would have to be considered an intrusion
into the autonomy of families. In other words, under the stated
ground rules of ideal theory-for which we assumed only
moderate scarcity-these leveling down strategies cannot be
reconciled with the autonomy of families.
In the next section I will look into various strategies of
intervention, by the government or other social institutions,
that might "level up," rather than down, that might, in other
words, increase the developmental opportunities of the lower
strata, and still leave the autonomy of families intact. My
general claim will be that such strategies of leveling up are
either so paltry in their efficacy or so utopian in their expense
that they cannot be expected to provide a solution to our
problem:~ 9 While there is a compelling need to pursue them as
far as possible, they cannot be expected to equalize develop-
mental opportunities to the levels offered by more advantaged
families. The only strategies of intervention that might offer a
hope of the required massive effects would amount to such a
wholesale change in the child's environment that their obvi·
ously prohibitive expense would violate any realistic construe·
tion of the budget constraint. Hence, leveling up strategies of
intervention-that leave the private sphere intact-are either
prohibitive in expense or insufficient in effect. On the other
hand, by making many of the children affected worse-off (those
who would have been advantaged without the equalization
efforts), leveling down strategies clearly violate the autonomy
of families if those provisions are to be universally enforced.
.w. They arc paltry when compared to the aspuation of full equalization.
Many such programs, however much they fall short of the ideal, are urgently
needed. See section 4.5 below.
 THE TRILEMMA OF EQUAL OPPORTUNITY 67

They do not satisfy the restrictive, paternalistic burden re-

quired to justify such interventions.

3.6 Strategies of Intervention T hejustlinchpin of the trilemma

outlined is the connec-
tion between the autonomy of families and the unequal devel-
opment of talents and other qualifications that takes place
under background conditions of inequality. If qualifications
tend to develop unequally in this way, then we are faced with a
hard choice between the principle of merit and equality of life
chances. Those who have unequal opportunities to develop
qualifications will have unequal life chances, if assignment to
positions is governed by the principle of merit. Correspond-
ingly, they can have equal life chances only if assignment to
positions violates the principle of merit in a systematic man-
ner. These two corners of the trilemma, the autonomy of the
family and background inequalities, can be viewed as a di-
lemma of equal opportunity-a forced choice between two
basic formulations of the concept.
It is worth pausing to explore further whether this crucial
causal connection is unavoidable. Perhaps governments, or
other organizations, might be able to intervene in a manner
compatible with the autonomy of families, in order to equalize
developmental opportunities, and thereby equalize life
chances. Considerable evidence has recently accumulated
that, while compensatory efforts can indeed improve the de-
velopmental opportunities facing the least advantaged, any
realistic efforts will not be substantial enough to alter the basic
pattern of conflict in the trilemma.
In the last section I alluded to two general strategies of
intervention aimed at greater equality in the developmental
influences on the formation of talents, motivations, and other
qualifications. One strategy was called leveling up and the
other, leveling down. The former achieves greater equality by
improving the resources and opportunities available to the less
advantaged strata, the latter achieves greater equality by lower-
68 THE TRILEMMA OF EQUAL OPPORTU!'IITY

ing the resources and opportunities available to the more

advantaged strata. If background conditions of inequality and a
realistic budget constraint are both assumed, then the effects of
both leveling up and leveling down strategies are predictable.
Under these conditions, leveling up strategies must be insuffi-
cient to equalize developmental conditions. Leveling down
strategies, on the other hand, must violate the autonomy of
families. To be successful, they would have to intervene coer-
cively to separate upper strata children from all of the advan-
tages that their parents might attempt to give them. 40 Hence,
the trilemma is not affected by such strategies since either the
autonomy of the family must be sacrificed (by leveling down)
or the inequality of developmental conditions basically is not
affected. As we have seen, this inequality of developmental
conditions requires, in turn, that we sacrifice either equality of
life chances (if we institute the principle of merit) or the
principle of merit (if we institute equality of life chances).
Under any of the possible options, at least one of the three
principles is sacrificed.
A great deal of recent research supports the view that
schools, by themselves, must be insufficient to redress the
developmental inequalities created by differing home environ-
ments. For example, one of the central conclusions arising
from James Coleman's landmark study was:
Altogether, the sources of inequality of educational oppor-
tunity appear to lie first in the home itself and the cultural
influences immediately surrounding the home; then they
lie in the school's ineffectiveness to free achievement from
the impact of the home and in the school's cultural homo-
geneity which perpetuates the social influences of the
home and its environs. 4 I

40. Coercion would be necessary because universal comphancc with such a

disadvantageous soluuon could not be expected. Sec the dtscusswn below.
41. James S. Coleman, "Equal Schools or Equal Students?," The Public
Interest 4 (19661: 70-75. The quotation ts from pages 73-74.
 T H E T R I L E M M A 0 F E Q U A L 0 P P 0 R T U N IT Y 69

Coleman amplifies this conclusiOn in the report:

One implication stands out above all: That schools bring
little influence to bear on a child's achievement that is
independent of his background and general social context;
and that this very lack of an independent effect means that
the inequalities imposed on children by their home, neigh-
borhood, and peer environment are carried along to be-
come the inequalities with which they confront adult life
at the end of school. 42
Of course, the mere fact that schools, as they now exist, have
not equalized achievement across disparate home environ-
ments does not mean that some institutional innovations in
the future might not provide for more perfect and more effica-
cious school systems. We cannot resolve questions of ideal
theory merely by direct reference to the conclusions of empiri-
cal studies of actual institutions.
Yet such institutional innovations, if they were to affect our
main conclusions, would have to take account of the freedom
granted families, by the autonomy principle, to make decisions
about which schools and other institutions their children will
be subjected to. This freedom greatly restricts the manipulabil-
ity of a crucial causal factor in the school environment-peer
group influences.
The option of private schools is by itself a crucial barrier. It
provides advantaged families an escape route from leveling
down strategies of equalization. Such leveling down efforts
could only be universally imposed on advantaged families at a
cost in family autonomy. Families could not be expected to
agree voluntarily and universally to an educational system
designed explicitly to disadvantage their children (while com-
paratively advantaging children from less fortunate back-
grounds). But unless participation in such a system were vol-
untary, it would violate the autonomy of families; unless it

42. Coleman, EqualJty of EducatiOnal Opportumty, p. 325.

70 T H E T R IL E M M A 0 F E QUA L 0 P P 0 R T U :'II I T Y

were universal, it would fall short of equalizing developmental

conditions across the society. 43
As long as the autonomy principle is maintained, the basis
for control over the crucial developmental influences-home
environment, school, peer group, location-will rest within
the family. Families sufficiently advantaged to exercise that
control can be expected to opt out of any leveling down
strategies of equalization, precisely because such strategies
would otherwise disadvantage their children. By opting out
they can ensure desirable peer groups for their own children in
favorable educational environments; their decision, of course,
will have an effect on the peer groups available to other
children, those who have not been able, or who have not
chosen, to opt out.
But suppose the option of opting out were universalized
through a publicly supported voucher plan? Such proposals
have many attractions. Yet, I believe, they cannot be expected
to resolve the root problem I am focusing on here. First, they
would not affect the crucial role of home environments, a
factor to which I shall return in a moment. Second, a crucial
characteristic of the schools one might imagine people opting
into requires that their option of selectivity be maintained.
Because differential home environments will affect the aca-
demic resources and qualifications students bring with them
to the selection process, upper strata children will maintain an
advantage-even if a voucher plan were so munificent as to
equal the tuition of the very best private schools (a possibility
that raises serious questions for any realistic construction of
the budget constraint). Furthermore, parents will vary greatly
in their ability to select schools and schools will vary greatly in
the criteria they use, in turn, to select students.
43. For one proposal that would sacnfice family autonomy by ehmmating
pnvate schools and other parental mvestments m child development, see
Bruce Ackerman's recent mnovauve book, Socwl fustJce in the Liberal State,
chapter 5. Wh1le Ackerman's proposal of an equal "liberal education" would
requue systematic efforts to counterbalance parental mfluences, there IS no
indication that It would yield equal life chances.
 THE TRILEMMA OF EQUAL OPPORTUNITY 71

If, as advocates believe, a voucher plan is likely to improve

the fit between the values and educational philosophies of
parents and the school environments of their children, such an
improvement, by itself, is unlikely to equalize developmental
opportunities across sectors of society. 44 As Coleman points
out in the citation above, the "cultural homogeneity" of public
school environments Ia by-product of districting and residen-
tial patterns) "perpetuates the social influences of the home
and its environs." Voucher plans that protect the voluntary
character of choice-both on the part of parents and on the part
of schools-can be expected to produce even greater cultural
homogeneity. Meanwhile, any efforts to eliminate this process
of mutual choice so as to require exposure to equal develop-
mental influences would clearly violate the autonomy princi-
ple.
Simply stated, only a mandatory routing of students to
school environments, that was designed, systematically, to
compensate for differential home environments, would affect
the basic mechanism at work in our trilemma. According to
such a plan, disadvantaged children would receive proportion-
ately more resources and attention, while advantaged children
would receive proportionately less, in order to equalize overall
developmental conditions regardless of family background.
Systematic leveling down, in other words, would be required.
Such a routing of children to school environments could not
realistically be expected of any voucher plan since parents who
have exposed their children to more advantaged environments
could not universally be expected (indeed, they would rarely be
expected) to expose their children voluntarily to such system-

44. Thoughtful proponents, such as Coons and Sugarman m Educatwn by

Cho1ce, argue on grounds of choice and liberty; their equality claims are
carefully hmited. They argue that a voucher plan will equahze the avmlability
of choice, extending It to poor famihes, they do not argue that a voucher plan
w1ll equalize developmental conditions (p. 21. Furthermore, the voucher plan
must be carefully designed If it 1s to avmd "economic class segregauon." See
theu critique of the Friedman plan m chapter 11.
72 T H E T R I L E M M A 0 F E Q lJ A L 0 P P 0 R T U N I T Y

atic disadvantages. And if the exposure were not voluntary

then it would violate the autonomy of families principle.
It may seem that in emphasizing the connection between
the autonomy of families and the option of private schooling
we are burdening an appealing principle with the weight of
indefensible privilege. Yet in modern developed societies, the
choice of which institutions outside the family a child will be
subjected to is a crucial part of the influence families may
exercise over their children. To the extent that parents lor
families operating consensually) lose control over such
choices, they have been supplanted by the state lor other
institutions) as key paternalistic agents. 4' Perhaps ultimately
this kind of change would be justified in the interests of
equalizing developmental opportunities. Its impact, however,
on the autonomy of families should not be underestimated.
Thus far, the policies mentioned would affect choice of
schooling either by closing it off through forced attendance at
certain schools or by widening it through the use of voucher
plans. I have not yet considered preschool efforts aimed at
ameliorating disadvantages in home environments.
Uric Bronfenbrenner, in a systematic review of research on
early childhood interventions, concludes that the crucial de-
velopmental factor in their success or failure is their effect on
what he calls the "mother-child dyad as an interactive sys-
tem."4~ This dyad (or a comparable one between a child and
some other relatively permanent adult such as a father or a
surrogate parent) is a crucial component in the child's cogni-
tive and emotional development:

45. For a htstory of instttuuonal encroachments upon the famtly, see

Chnstopher Lasch, Haven m a Heartless World: The Family Besieged !New
York: Baste Books, 19771.
46. U.S., Department of Health, Education and Welfare, A Report on Longi-
tudmal Evaluations of Preschool Programs: Is Early Intervention Effective?
voL 2, by Une Bronfenbrenner (Washmgton, D.C.: Office of Chlld Develop-
ment, 19741, p. 26.
 T H E T R I L EM M A 0 F E Q U A L 0 P P 0 R T U N I T Y 73

Any force or circumstance which interferes with the for-

mation, maintenance, status or continuing development
of the parent-child system in turn jeopardizes the develop-
ment of the childY
Strategies of intervention intended to improve the develop-
mental opportunities of the less advantaged might usefully be
divided into:
a. those that do not attempt to affect this mother-child
system but rather attempt to supplement it by offering the
child other developmental opportunities;
b. those that do attempt to affect this mother-child system
by strengthening it; and
c. those that attempt to replace it with some alternative
compatible with the child's existing family assignment.
Generally, early preschool education in a group setting falls
into the first category. Certain efforts at intervention in the
home environment fall into the second. And "ecological inter-
ventions" that provide the child a virtually new environment
fall into the third.
One difficulty with preschool interventions in a group set-
ting is that substantial gains, at least in IQ, have not been made
beyond the first years of intervention, no matter how long the
program is continued. Furthermore, the results have been
transient:
In general, one year after intervention is terminated, the
IQ of the "graduates" begins to drop, the difference be-
tween the experimental and control groups gradually de-
creases, the once impressive gains are reduced to a few
points, and, what is most crucial, the average IQ of the
experimental group often falls back into the problem range
of the lower 90s and below. 48

47. Ib1d., p. 31.

48. Ibid., p. 14.
74 THE TRILEMMA OF EQUAL OPPORTl'I\:ITY

Since Bronfenbrenner rendered these pessimistic conclu-

sions in 1975, a new wave of somewhat more optimistic
findings has emerged. Some long-term effects on scholastic
achievement (as contrasted to IQ) do seem to result from early
participation in programs such as Head Start. As the preface to
the most enthusiastic volume of new findings notes, however:
Early intervention does not assure that the children will
perform as well as the average child in the population. The
benefits of early intervention over non-intervention arc
striking, but comparison of the children in early interven-
tion programs with grade norms for 4th, 5th and 6th grades
is not impressive. The children are behind the average
child in the population. 49
Compared to control groups of children from similar hack-
grounds, children from early intervention programs do show
some long-term scholastic gams. But these gains are largely a
matter of the prevention of loss that would otherwise have
occurred. Children from these programs fall behmd less rapidly
than control groups from similar backgrounds, though they
still fall behind the norms for their grade levels.
Furthermore, the most striking positive effects have been on
the assignment of children to special education classes (for the
learning disabled) and on retention in grade. Compared to
control groups, early intervention children arc far less likely to
be held back and are less likely to be assigned to special
remedial classes. While these are important improvements,
they are far short of the effects that would be required were
early intervention to equalize developmental conditions re-
gardless of family background.
Home-based programs directed explicitly at affecting pat-
terns of parent-child interaction have achieved more lasting
gains in IQ than have preschool programs in a group setting.
49. Bernard Brown, ed., Found: Long· Term Cams From Early Intcn·cntJon
!Boulder, Co.: Westview Press, 1978); the quotatiOn IS from the Preface hy
Edith H. Brotberg, p. xviu
 THE T RILE M M A 0 F EQUAL 0 P P 0 R TUN IT Y 75

Yet, these home-based programs are also notable for what

might be called the "climbing control group" !untreated sub-
jects who also improve notably). These studies involved an
"admission requirement that parents be interested and willing
to enroll their child in the program even at the risk he might
end up in the control group." 50 As Bronfenbrenner concludes:
The climbing control group resulted from the self-selec-
tion of families in terms of their motivation to provide
educational experience for the child. The more motivation
was required, the more selective the sample of parents,
and the more likely their children were to make a gain in
IQ even if not admitted to the intervention program. 51
There is no doubt that home-based strategies of intervention
aimed at strengthening patterns of parent-child interaction
have been comparatively successful in producing more lasting
gains in IQ. Yet, the resultant gains are also subject to erosion
and have not been reflected in school performance measures,
for example, reading programs. 52 Furthermore, they depend
crucially on the self-selection of parents in terms of motiva-
tion. The generalizability of their results in any large-scale
application throughout the society is, therefore, extremely
dubious. So this approach, while helpful, should not be viewed
as a solution to our basic problem since it is only applicable to
families in which parents are especially motivated.
The most impressive gains in IQ have been achieved by
massive ecological interventions, those that replace a child's
home environment with a new one. For example, a Milwaukee
project matched twenty newborn children from deprived back-
grounds with twenty paraprofessionals who acted as mother
replacements for each child throughout the working day.
Meanwhile, the children's mothers underwent training rele-
vant to both child-rearing and employment. In each case, both
50. Bronfenbrenner, Early Interventwn p. 21.
51. Ibid.
52. Ib1d., p. 37.
76 T H E T R I L EM M A 0 F E Q U A L 0 P P 0 R T U N I T Y

mother and child were exposed to a carefully sequenced phase

of educational activities. By the age of five and a half the
children showed impressive gains in IQ compared to a control
group. 53 Instead of the eight to thirteen points in IQ gained
from the most successful group interventions, or the fourteen
to sixteen points in IQ gained by the most successful parent
intervention programs, ecological intervention programs like
this one achieved gains of over twenty-five points. 54
Two caveats should be noted immediately about this strat-
egy: First, "the program 1s, and will continue to be, as expen-
sive as it is effective, perhaps more so. And in terms of large-
scale applicability, the costs are prohibitive." 55 Second, the
long-term psychological effects of the program are unknown.
The improvement in children's cognitive development may
have to be balanced against harmful effects:
Whatever happens to them intellectually, serious ques-
tions arise about their development in other spheres, espe-
cially in terms of identity formation in their relation to
their family or to other children in the neighborhood from
whom they are partially isolated so long as they continue
in the program. s6
Given the potential for harm to the child, such a strategy of
radical separation and virtual replacement of the parent could
be justified-in a manner compatible with the autonomy of
families-only in cases where the alternative of leaving the
child in the care of its parents clearly involved severe harm,
53. Ibid., p. 45. See also R1ck Heber, Richard Dever, and Juhanne Conry,
"The Influence of Environmental and Genetic Vanables on Intellectual Devel-
opment," m Herbert J. Prehm, Leo A. Hamerlynck, and James E. Croson, eds.,
Behavwral Research m Mental Retardation (Eugene, Oregon: University of
Oregon, Rehabilitation Research and Trammg Center m Mental RetardatiOn,
monograph no. 1, 1968) pp. 1-22, especially pp. 19-21. For some methodologi-
cal cnticisms see Ellis B. Page, "Miracle m Milwaukee: Ra1smg the IQ,"
Educational Researcher 1, No. 10 (October 1972): 8-16.
54. Bronfenbrenner, Early Intervention, p. 49
55. Ibid., p. 46.
56. Ib1d.
 THE TRILEMMA OF EQUAL OPPORTUNITY 77

endangering some of the essential prerequisites discussed ear-

lier. Hence, the applicability of this strategy would have to be
limited to those in the severest circumstances, even if its
expense could be supported. The expense, we might add, of
full-time care on a one-to-one basis would be prohibitive, at
the large-scale, by any plausible construction of our realistic
budget constraint.
Setting aside the question of expense, any large-scale appli-
cation of this strategy would presumably violate the autonomy
of families. Parents-even those who are relatively disadvan-
taged in the judgment of social policy makers-could not be
expected to relinquish, universally and voluntarily, their chil-
dren to virtually full-time care by paraprofessionals. They
would retain their parental role then in name only. Even when
others might judge the children to be marginally better-off, this
judgment is inevitably controversial and the parents cannot be
expected, except in extreme and unusual cases, to agree.
Furthermore, practical experience in this field supports the
conclusion that disruption of intact families usually does more
harm than good to the child, except in the most extreme cases
of deprivation or abuseY Hence this last strategy, which really
amounts to parent replacement by paraprofessionals, cannot be
considered a solution that avoids our trilemma since it re-
quires sacrifice of one of the three essential principles, the
autonomy of families.
In other words, even if the realistic budget constraint were
set aside, it is safe to assume that no system of wholesale
replacement of parents with paraprofessionals would be en-
tered into voluntarily on a universal basis by all those whose
children it would benefit. Yet coercive intervention, compati-
ble with the autonomy of families, could be justified only
where lack of essential prerequisites for the children was in
question. Otherwise, if parents chose to take primary responsi-
bility for child-care themselves, they could be prevented from

57. See Goldstein et al., Beyond the Best Interests. pp. 3-8, 48-52.
78 T U E T R I L E M M A 0 F E Q lJ A L 0 P P 0 R T ll N I T Y

doing so only at the cost of sacrificing one of our three princi-

ples. Given the centrality of parental roles in most of our
lives, this result is only to be expected. One might object that,
if indeed the children would be benefited, parents in those
cases should be forcibly supplanted from most of their influ-
ence. Yet, I believe that the autonomy of families assumption,
as I have defined it here, accurately captures the extent of
protection from such social policies that we would customar-
ily give to the family. Parents deserve protection from social
interference (given that they satisfy certain essential prerequi-
sites for their children) even when others might judge that
their children could be raised better by someone else.
In summary, while both group and home-based programs
produce some significant gains, even in the long-term, they fall
far short of equalizing developmental conditions throughout
the society regardless of class background. The massive eco-
logical intervention exemplified by the Milwaukee experiment
might so transform the total environment of the child as to
approach equalization of developmental conditions. It could
not be adopted universally, however, without violating the
autonomy of families (and any plausible construction of a
realistic budget constraint as well). It seems reasonable to
conclude, therefore, that the recent wave of institutional ex-
perimentation with early childhood intervention does not offer
any results that would affect the basic trilemma of relations
outlined earlier. While those experiments suggest many ave-
nues for social policies that are urgently needed, they all fall far
short of equalizing developmental influences, regardless of
family background, through a strategy of intervention that
would be feasible as a large-scale solution.
The general problem is that leveling up strategies compati-
ble with the autonomy of families can expose children to new
developmental influences only with the voluntary cooperation
of their families (given the caveats noted earlier in the auton-
omy principle). The more far-reaching the developmental in-
fluences required, the closer the intervention would approach
 T H E T R I L EM M A 0 F EQ U A L 0 PP0 R T U N I T Y 79

the parent-supplanting example of the Milwaukee experiment

cited above. Complete manipulability of causal factors could
be achieved only by something approaching replacement of the
parents altogether, and it cannot be expected that this sort of
extreme dislocation of intact families would be entered into
voluntarily by all of those whose children would be benefited.
On the other hand, forced participation would violate the
autonomy principle. And nonuniversal participation would
fail to resolve the problem since it would fail to equalize
developmental conditions throughout the society.
An alternative strategy of intervention atmed at some di-
rect effect upon the character of parent-child interaction in the
home has already been mentioned. If parents arc not to be
replaced, they may be transformed. This strategy's success
depends heavily on the willingness and ability of parents to
cooperate. For this reason experiments involving self-selected
parents, who are especially motivated, cannot be generalized to
the population at-large.'H Home-based strategies confront the
problem noted above; if the modification of developmental
influences is to be substantial enough to approach equaliza-
tion, universal participation (at least among all those whose
children might be thought to benefit) cannot be expected to
come about voluntarily. In so intimate and salient an area of
private liberty as parent-child relations, any transformation of
behavior on a mass scale is hound to encounter opposition. The
coercion, however, that would be required to bring about
universal compliance would violate the autonomy principle.
On the other hand, selective rather than universal participa-
tion would fail to achieve equalization of developmental con-
ditions throughout the society.
The autonomy of families principle therefore provides a
significant impediment to leveling up strategies just as it does

58. See Bronfenbrenner, Early Jnterventwn. pp. 35-38. For more on the
home-based strategy see P. Levenstein, "Cogmuve Growth m Preschoolers
through Verbal InterventiOn with Mothers," Amencan /ournal of Orthopsy-
chwtry 40 11970): 426-32.
80 THE TRILEMMA OF EQUAL OPPORTUNITY

to leveling down strategies. Except when certain essential

prerequisites are at stake, it would prevent the state from
manipulating developmental conditions without family con-
sent. Systematic leveling down strategies, if made universal,
would clearly violate the autonomy principle. It would be
hopelessly unrealistic to expect advantaged parents to consent
voluntarily and universally to conditions that were explicitly
designed to render their children worse off.w But any effective
leveling down strategy would, by definition, have that result.
Conversely, leveling up strategies are either insufficient to
close the gap in developmental conditions across different
family backgrounds, or they must violate the autonomy of
families in order to achieve, universally, a full manipulation of
the required developmental conditions, either through replace-
ment or through transformation of conditions within the fam-
ily.
Suppose, however, that all potential parents were subjected
to systematic indoctrination, or thought control, so that by the
time they reached procreation they could universally be ex-
pected to comply "voluntarily" with systematic government
intervention into their child-rearing practices. Or they would
systematically conform without any further government inter-
vention to whatever behavior was required and, if they be-
lieved themselves unable to conform, they would refrain from
havmg children. A hypothetical reform like this might appear
to remedy all the disadvantages now suffered by some children
because their parents lack motivation or competence, or have a
distinctive conception of appropriate child-rearing practices
that places their children at a disadvantage. While economic
irregularities would remain, there is no doubt that many devel-
opmental disadvantages might be overcome by such a reform.
Would it escape the objection, however, that it violates the
autonomy principle?

59. Sec note l of this chapter for the limits of the stnct compliance
assumption.
 THE TRILEMMA OF EQUAL OPPORTUNITY 81

To answer this question, I must return to the definition of

"coercive interferences" in the autonomy principle. 60 If univer-
sal and apparently voluntary compliance were to come about
through a campaign of indoctrination that suppressed certain
alternatives and promoted others, it would count as coercive
interference and would violate the autonomy principle. Ac-
cording to the earlier definition, if knowledge of the alterna-
tives to action A has been suppressed, 61 a person has been
coerced into doing A just as effectively as if he had been
threatened with dire consequences for performing any of those
alternatives. Suppression and thought control may be as effec-
tive as punishment. The autonomy principle could be trivial-
ized unless it also provided protection against such efforts at
manipulating a person's awareness and understanding of alter-
natives.
It might be objected also that I have rested the analysis of
this section too heavily upon recent American experience.
Perhaps other strategies of intervention would be more effec-
tive than those that have been tried recently in this country.
Furthermore, there is certainly no reason to focus meritocratic
criteria for assignment on IQ. After all, our definition of
qualifications was quite general and efforts to boost IQ (an
admittedly controversial measurej 62 should be considered only
a particular case of a more general problem.
In response, it is worth emphasizing that I invoked recent
American experience for illustrative purposes only. The exam-
ples cited serve to reveal the general impediment posed by the

60. See section 2.4 above.

61. Suppress10n m this defimuon presumes mtenuonality. And It apphes
with respect to known alternatives and our evaluation of them. One should
not expect to consider synoptically all of the possible alternatives. See David
Braybrooke and Charles E. Lmdblom, A Strategy of Decision: Policy Evalua-
tiOn as a Social Process (New York: The Free Press, 1963), part 1.
62. For a systematic review see Block and Dworkm, The IQ Controversy,
part 4. For a discussiOn of operatiOnal cntena for test fairness see Robert L.
Linn, "Fair Test Use in SelectiOn," Rev1ew of Educatwnal Research 43(1973):
139-61.
82 THE TRILEMMA OF EQUAL OPPORTU:\fiTY

autonomy of the family principle to any systematic efforts to

equalize developmental conditions regardless of family back-
ground. Some leveling up efforts will undoubtedly accomplish
a great deal. There is no reason to believe, however, that any
strategy might be devised that would leave the autonomy of
families in place and, at the same time, equalize developmen-
tal conditions across families from different socioeconomic
strata. Leveling down strategies can expect to encounter resist-
ance from advantaged families because, whatever the precise
form, such strategies are designed to render children worse off.
On the other hand, leveling up strategies, whatever the precise
form, must be insufficient to close the gap unless they involve
a complete manipulation of all the significant causal factors
affecting the disadvantaged child. Such a manipulation could
be accomplished, I have speculated, either through wholesale
supplanting of the parental role, as in the Milwaukee experi-
ment, or through some complete transformation of parent-
child patterns of interaction. Whatever the precise mechanism
of intervention, universal compliance with such a radical
change in so intimate and central an area of private liberty as
parent-child relations could only be expected were coercion to
be employed. Hence the impediment posed to strategies of
intervention by the autonomy principle.

3. 7 Preferential11eatment
and Compensation for the My central argument for the
Past trilemma of equal opportu-
nity presumes the conditions of ideal theory. Realization of
any two of our principles reasonably can be expected to pre-
clude the third under the ideal conditions of both strict compli-
ance lin the present and relevant recent past) and only moder-
ate scarcity. No reference to any legacy of injustice is necessary
to produce a forced choice among the three stated principles. In
this sense, simultaneous commitment to all three of our prin-
 THE TRILEMMA OF EQUAL OPPORTUNITY 83

ciples, given background conditions of inequahty, produces an

incoherent ideal for public policy, even under favorable condi-
tions.63
If a coherent and defensible position is to be reached, this
pattern of conflict requires some modification of our usual
commitment to all three principles. One strategy would ex-
plicitly establish priority relations justifying the required sacri-
fice in one of the three principles. Another strategy would
require abandoning one or more of the principles entirely and,
perhaps, replacing others with a quite different principle. A
third strategy would modify the character of our commitment
to these principles, rather than the character of the principles
themselves. If we regard them only as prima facie principles, to
be traded off intuitionistically, one against another, in each
particular case of conflict, then we have arrived at a defensible
if limited kind of non-theory. While such a nonsolution to the
problem of equal opportunity would be dissatisfying to many,
it would avoid the difficulties raised by any commitment to
fulfill simultaneously all three principles. 64
I will later return to these various strategies available for
ideal theory. For the moment, however, it may be useful to
place this theoretical discussion in the context of recent Amer-
ican policy. What does this analysis suggest about recent
efforts in the United States to provide preferential treatment
according to native characteristics such as race, sex, or ethnic
origin, motivated in part to compensate for a legacy of past
discrimination? To what extent are these policies compatible
with the liberal framework of individualistic competition em-

63. It would be mcoherent as a gu1de to possible social choices when this

empmcal analysis IS accepted and when our commitments to these pnnc1ples
are strong rather than weak or pnma fac1e. See n 12, chapter 2 above.
64. See sectiOn 5.2 below. For a general d1scuss1on of responses to conflict
among ethical criteria see Brian Barry and Douglas Rae, "Poht1cal Evaluation,"
section 2, m Fred I. Greenstein and Nelson W. Polsby, eds., Handbook of
Pobtical Science, 9 vols. (Readmg, Mass: Addison-Wesley, 1975), 1:337-401.
84 T II E T R II. E 1\\1\\ A 0 F E Q U A I. 0 P P 0 R T lJ N I T Y

ployed here? How docs the departure from ideal theory, be-
cause of our actual history of discrimination against various
groups, affect the analysis?
Starting from the basic liberal notion of a fair competition
among individuals for unequal positions in society, two princi-
ples explicitly concerned with equal opportunity were
reached-merit and equality of life chances. It would be sur-
prising and disturbing were any recognizably liberal theory
forced to relinquish its commitment to either of these. Given
background conditions of inequality, however, we are forced to
choose between realization of both these principles and the
autonomy of the family. As indicated earlier, the latter princi-
ple identifies the core area of negative liberty that is most
central to the way most of us have structured our lives. It
would be equally disturbing and surprising were any liberal
theory forced to relinquish its commitment to realizing this
principle, even under ideal conditions.
This trilemma of forced choice under ideal conditions sug-
gests a minimum requirement for acceptable public policies,
even under less than ideal conditions. It seems reasonable to
require that no sacrifice in one of these three central principles
should be undertaken unless the sacrifice is necessary for a
gain in one or more of the other two, or unless there is some
other valuable gain (according to other principles that would
then have to be argued for, in turn). It is one thing to sacrifice a
principle when the sacrifice is absolutely necessary to realize
one or more other precious goals. It is quite another thing to
sacrifice it when there is not such a corresponding gain lor even
a prevention of loss) in some other important value.
This minimum condition seems extremely weak, since it
prescribes nothing for the enormous range of cases where we
must choose between sacrificing one or more of these values
and preserving or improving the realization of one or more
others. Despite this obvious incompleteness, this minimum
condition docs support a clear line of argument against certain
practices of preferential treatment that are actually employed
 T H E T R I L E M M A 0 F E Q lJ A L 0 P P 0 R T lJ :\i I T Y 85

in the present American policy context. Some of the practices

that proved most controversial in the recent debates over the
DeFunis and Bakke cases would, in fact, fail this minimum
condition. 6'
Before turning to such cases of reverse discrimination, I will
distinguish them from several weaker varieties of preferential
treatment. It is worth emphasizing that many practices labeled
"affirmative action" do not even imply preferential treatment
for minority group members in the actual process of assign-
ment. By forcing institutions to report certain kinds of infor-
mation, they are meant only as procedural safeguards for
ensuring nondiscrimination. A similar point can be made
about requirements that positions not be filled through "word-
of-mouth contact" 66 or through systems that grant preference
to relatives of current employees (a practice once common
among unions and employers in the construction trades in the
United States).n 7 Other requirements ensure procedural fairness
through an extremely weak kind of preferential treatment in
the solicitation and recruitment of applicants. Special efforts
to encourage a wider and more diverse pool of applicants can
increase access to the competitive process for groups that
might not be represented otherwise. Whether preferential
treatment in any stronger sense is involved would depend on
how those applicants are evaluated later in the assignment
process, compared to applicants who are not from the desig-
nated special groups.
Other practices of apparent preferential treatment may
really amount to no more than guarantees of procedural fair-

65. See the arguments below in this section agamst preferential treatment
merely on grounds of race and in favor of preferential treatment when tt is
directed at the dtsadvantaged. Strangely enough, whtle the program the Su-
preme Court refused to rule on m DeFunis would fat! the mmimum condition
proposed here, the program the court struck down in Bakke would pass tt.
66. U.S. Commission on Civtl Rights, Statement on Affirmative Action.
Cleannghouse Publication 54 (Washmgton, D.C.: U.S. Government Printmg
Office, 1977), p. 3.
67. Ibtd.
86 THE TRILEMMA OF EQUAL OPPORTUNITY

ness when they involve special care in the interpretation of

data from persons who come from disadvantaged backgrounds
or from differing cultural milieu. Some standardized tests may
fail to measure accurately an applicant's real competence;
apparent differential treatment may really amount in such
cases to no more than conscientious efforts to make up for
biases in the instruments of evaluation. I will set aside such
measurement problems here since efforts to overcome them do
not truly constitute cases of preferential treatment. Such ef-
forts are, of course, entirely compatible with the principle of
merit. In fact, they would be required by it.
It is also worth mentioning some very weak forms of prefer-
ence that might be given in the actual assignment process. A
ceteris paribus preference to break ties is sometimes employed.
Members of a specified group are chosen, in other words,
provided they are equally qualified. This kind of preferential
practice does not involve a serious sacrifice of merit since, by
definition, those chosen must be equally qualified. Neverthe-
less, it is worth noting that such practices are far from trivial in
their implications, since they may substantially affect the
probability that members of other (nonminority) groups will be
chosen. Instead of an equal chance in cases of ties, nonminor-
ity group members must lose out in such cases. They must, in
other words, do better than equally well in the competitive
process in order to gain a position. To see that such a weak
form of ceteris paribus preference is nontrivial, consider our
reaction to it when it is practiced against rather than in favor of
a given minority group. If an employer consistently refused to
hire equally qualified blacks this would constitute objection-
able descrimination-even if he were willing to hire blacks
when they were more qualified than their competitors. Even
when discrimination against a minority group takes this weak
form, we still rightly consider it discrimination. Instead of
equal consideration, in the sense of an equal chance for those
with equal qualifications, members of the designated group
must do better when they are discriminated against in this
way. A ceteris paribus preference, in other words, is a denial of
 THE T R I L E M M A 0 F E Q U A L 0 P P 0 R TUN IT Y 87

equal consideration of one's qualifications and should be con-

sidered more than a trivial departure from the principle of
merit, even though those chosen must be equally qualified.
The kind of preferential treatment now often practiced is far
more potent than this weak, ceteris paribus variety. Let us say
that a practice of preferential treatment amounts to 11 reverse
discrimination" if it produces significant and widespread sacri-
fice of the principle of merit in assignment in order to favor
some specified group defined in terms of arbitrary native
characteristics (such as race, sex, or family background). 6H By a
significant sacrifice in the principle of merit in any given case,
I mean simply the choice of those who are substantially less
qualified over those who are more qualified, as those qualifica-
tions would be impartially assessed for the position in ques-
tion.
Such reverse discrimination was, of course, the issue in the
DeFunis and Bakke cases where preferential treatment went
beyond the various weak measures cited above. It went beyond
ensuring nondiscrimination, an appropriately broad applicant
pool, and an accurate assessment of qualifications despite
possible biases in the instruments of measurement. Preferen-
tial treatment in these cases clearly involved a substantial
sacrifice in merit considerations as they might be applied
regardless of race, sex, or other arbitrary native characteristics.
For example, in the DeFunis case all but one of the minority
students admitted to the University of Washington Law School
ahead of DeFunis rated lower on the university's 11 predicted
first year average," a composite of grades and the law school
admissions test that provided the primary basis for evaluating
applicants. Furthermore, thirty of the minority students had
scores so low that 11 but for their racial classification !they]
would have been summarily rejected" (emphasis added). 69 And
considering all indicators of qualifications, in addition to
68. Reverse discnmmatwn is thus preferential treatment that 1s both sys-
tematic and widespread.
69. Bnef of the Anti-DefamatiOn League of B'na1 B'nth as amicus cunae,
DeFums. p. 9. See also Smdler, Bakke, DeFunis, and Mmonty Adm1sswns.
SS T H E T R I L E M M A 0 F E Q U A L 0 P P 0 R T U N IT Y

grades and test scores, the university admitted the sacrifice in

meritocratic criteria resulting from this process but contended,
nevertheless, that it was justified by the greater racial diversity
that resultedJO
Preferential treatment in this case fit a pattern that is now
quite common in highly competitive meritocratic contexts.
Even though some of the arguments for the program made
reference to economic, social, and cultural disadvantages, pref-
erential treatment was instituted merely on the basis of racial
classifications. As Bickel and Kurland pointed out in their
amicus curiae brief:
There is no pretense in this record that the assumption of
cultural and economic disadvantage as applied to any
particular individual applicant rested on anything but his
race, or that the contrary assumption, namely, that an
applicant had the normal cultural and economic advan-
tages, in turn, rested on anything but the applicant's race .
. . . And the evidence is clear that no effort was made to
examine the mass of applications for individuals of cul-
tural or economic deprivation, whatever the race of the
applicant. 71
Although socioeconomic disadvantages purportedly pro-
vided a justification for preferential admissions, the actual
practice adopted separated applicants merely by race or ethnic
origin. Marian Wright Edelman in another amicus curiae brief
argued that these classifications were entirely appropriate:
For it was the racial minorities generally-not just their
"disadvantaged" or "poor" or "culturally deprived" mem-
bers-that had suffered in the past from pervasive racial
discrimination and de facto exclusion from law school and
Sindler concludes: "In short, had these 30 mmonty admmees been wh1te, they
would have been turned down summanly." (pp. 37·38)
70. Testimony from the Dean of the Law School 1s c1ted by B1ckel and
Kurland m Bnef of the Anu·Defamatwn League of B'na1 B'nth as am1cus
curiae, pp. 9·10.
71. Ib1d., pp. 12-13.
 THE T R 1L EM M A 0 F E Q U A L 0 P P 0 R T U N l T Y 89

the legal profession, which were the problems the School

sought to remedy. In that sense, all members of the minor-
ity groups were relevantly "disadvantaged" for the pur-
poses of the University policy. 72
If this practice of preferential treatment merely according to
race were to be defended, it would require a sharp departure
from the liberal framework I have been employing here. For the
significant sacrifice in the principle of merit was not counter-
balanced by any gain in the other two principles I have been
employing-equality of life chances or the autonomy of the
family. In this sense it stands in stark contrast to another kind
of preferential treatment program, one designed to help only
those who had themselves suffered developmental disadvan-
tages. If preferential treatment were awarded according to
developmental disadvantages that had actually been suffered,
as these might be roughly ascertained by income and class, or
even by more specific information, then the cost in merit
might arguably be counterbalanced by an improvement in
equality of life chances. Persons who might otherwise never
have a realistic chance of attaining certain positions because of
unequal developmental opportunities, would see their chances
improved by such a system. 73
Such a program would be entirely compatible with the basic
liberal framework of individualistic competition analyzed
here. It would be designed to redress the inequality in develop-
mental opportunities that, arguably, renders the competitive
process objectionable because of its lack of background fair-
ness. It would make up for the fact that in the first place some
people never really would have had a chance to develop the
skills and qualifications demanded by meritocratic competi-
tion.
This improvement in background fairness stands in sharp

72. Brief for the NatiOnal Council of Jewish Women and others as amiCI
cunae, DeFums, p. 59.
73. For a Similar proposal see Goldman, Justice and Reverse Discrimina-
tiOn, pp. 194-95.
90 THE TRII.EMI\tA OF EQl!AL OPPORTli~ITY

contrast to the effects of a program of preferential treatment for

racial or ethnic groups as such. Even proponents of such
programs admit that many of the beneficiaries will be persons
who are not disadvantaged themselves. As the University of
California at Davis acknowledged in its Bakke brief, only
about one third of the underrepresented minorities accepted in
our nation's medical schools in 1976 had parental incomes
under $10,000. 74 There was also considerable speculation
whether DeFunis, a Sephardic Jew from a relatively poor back-
ground who had to work his way through college and law
school, might reasonably have been considered more disadvan-
taged than many of the middle-class minority students admit-
ted ahead of him.'"
Terence Sandalow, an advocate of preferential treatment
identifies the general issue:
It is reasonable to suppose that a substantial percentage of
the lnot otherwise admissable) minority applicants who
have the best chance of success in law school and the
professions come from backgrounds that cannot plausibly
he considered disadvantaged. Opponents of minority pref-
erences have found it especially galling that the child of a
wealthy and well-educated professional should receive
preferential treatment simply because of his or her race.76

74. Bncf for Petitioner, The Regents of the Umvcrsuy of Cahforma v. Allan
Bakke, p..U, n. 46. Th1s po10t 1s also made 10 the Bncf of the Amcncan Jew1sh
Comm1ttce and others as am1c1 cunae, Rakke. p.•H, n. Ill Of course, I do not
mean to 1mply that all of these students were admmcd as a result of preferen-
tial treatment.
75 Sec the Wash10gton Supreme Court op1010n, Dcfums v. Odegaard, 82
Wn.2d II, 507 p.2d 1169!19HI, C. I. Hale !d1ssent10g), repnnted in Robert M.
O'Ne1l, IJJ.~cnmmatmg Agamst Ihscnminatwn IBloommgton: Indtana Um·
vcrslty Press, 1975), p. 2m-05. Sec also M1chacl J. Malbm, "The Court Ought
to Dcc1de for Mr. Defunis" The New York Times. 12 April 1974, p. 31. On
Defums's financ1al s1tuatmn sec S10dler, Rakke. IJeFums, and Mmonty Ad-
lliJS.mm.~. pp ..~H-.W.
76. Terence Sandalow, "Rac1al Preferences 10 H1gher Education· Pohucal
Rcsponsih1hty and the Jud1c1al Role," llnivemty of Chicago Law Review 42
!Summer 1975\:691.
 THE TRILEMMA OF EQUAL OPPORTUNITY 91

Despite this difficulty, Sandalow would oppose reshaping spe-

cial admissions programs "to focus upon disadvantaged appli-
cants" because this "would also force the exclusion of some of
the ablest minority applicants." 77
Kent Greenawalt, another advocate of preferential treatment
by race, acknowledges the same problem, which he calls a
serious problem of "o~erinclusiveness." 7 R He offers, however, a
novel argument for preferring such minority group members,
even if they are not from disadvantaged backgrounds:
If one effect of discrimination is that fewer blacks can
qualify without preference, then one proper form of com-
pensation is to eliminate the effect of denial of admissions
on those who would otherwise have qualified. The closest
a law school may be able to come to identifying the class
of those who in the absence of discrimination probably
would have qualified without preference is to choose the
best prepared blacks who now apply.7 9
There are several complexities in this counterfactional claim
that I shall discuss below. For the moment, it is worth noting
that if one were to attempt consistently to imagine a version of
American society in which previous discrimination toward
blacks had never occurred, there is little reason to believe that
those particular blacks who now are in a position to benefit
from preferential treatment (merely according to racial charac-
teristics) are "those who would otherwise have qualified" if
there had been no legacy of injustice against blacks. Surely the
compensatory claim is far stronger for present members of the
ghetto underclass. Who knows where they would be now but
for the "lengthening shadow of slavery" and its effect on their
family structure and the social prison of poverty.Ro There is no
77. lb1d.
7R. Kent Greenawalt, "judicial Scrutiny of 'Bemgn' Rac1al Preference m
Law School Admisswns," m Gross, ed., Reverse Discnmmatwn, p. 222.
79. lb1d., p. 224.
80. Sec John E. Fleming et al., The Lengthenmg Shadow of Shn·ery· A
Historical fusll{icatwn for Affirmallve Actwn for Blacks in H1xher Education
(Washmgton, D.C.: Howard UnivefSlty Press, 1976).
92 THE TRILEMMA OF EQUAL OPPORTUNITY

reason to believe that those blacks who are presently "best

prepared" offer even a remote approximation to those blacks
"who in the absence of discriminations probably would have
qualified" (emphasis added). I will explore in some detail some
of the more surprising implications of taking these counterfac-
tuals seriously. In the meantime, it is worth noting that this
strategy for dealing with the overinclusiveness problem is open
to serious objections.
The main point is that given the differential developmental
effects upon more advantaged minority group members-ad-
vantages resulting from the income, family stability, peer
groups, and schooling that their parents can make available to
them-it is only reasonable to expect them to profit dispropor-
tionately from policies of preferential treatment applied
merely according to membership in the minority group. It is
also only reasonable to expect them to be differentially repre-
sented among the most qualified minority students. And if
colleges and universities conceive of preferential treatment in
terms of race or ethnic background only, rather than in terms
of disadvantages actually suffered by the preferred individuals,
then a fair application of meritocratic criteria to the pool of
minority applicants can be expected to result in these same
disproportions.
Richard Posner argues from census data that, since "most
poor people are white/' and since "most members of generally
disadvantaged minorities are not poor/' we can expect prefer-
ential admission by race to produce a common pattern of
advantages for relatively affluent minority students over poor
whites who are actually more disadvantaged. This is particu-
larly so because we can "expect the nonpoor members of the
minority groups to be over-represented, relative to the poor of
their groups, among law school and other university appli-
cants."81
81. Richard A. Posner, "The DeFums Case and the Constitutwnahty of
Preferenual Treatment of Rac1al Minonues," m Ph1hp B. Kurland, ed., The
Supreme Court Rev1ew 1974IChicago: University of Chicago Press, 1975), p.
IS,n.31.
 THE TRILEMMA OF EQUAL OPPORTUNITY 93

Of course, once women are added to the pool of "minorities"

!blacks, Indians, Orientals, and Hispanics) given group prefer-
ence per se, the sheer number of relatively advantaged persons
eligible for preferential treatment is expanded enormously.
Thomas Sowell has calculated that:
With the addition of women to the groups entitled to
preferential lor "remedial") treatment, all the persons so
entitled constitute about two-thirds of the total popula-
tion of the United States.s2
Even if minority group membership may serve usefully as an
empirical proxy for actually suffered disadvantages in some
empirical contexts, it would be difficult to defend the use of
mere membership in minority groups when the groups are,
together, so numerous as to cover two-thirds of the population.
The essential point is that practices of reverse discrimina-
tion applied merely according to native characteristics may
benefit the relatively advantaged members of the designated
group without substantially helping those who are truly de-
prived. Without administrative practices designed explicitly to
guarantee their application to the disadvantaged, they stand
open to the charge that they sacrifice merit without ensuring
any significant gain in equality of life chances.
Of course, such practices do serve some worthy purposes.
They may provide role models for other minority students,
they may provide a critical mass of minority students in a
given institution !raising the likelihood of survival in profes-
sional school for members of the minority group in general),
and they may affect the development of aspirations on the part
of other minority group members. 83 Yet these positive second-
ary efforts have to be balanced against negative ones; they may

82. Thomas Sowell, Knowledge and Deciswns (New York: Bas1c Books,
1980), pp. 251-52.
83. See Greenawalt, "Jud1c1al Scrutmy," in Gross, ed., Reverse Discrimma-
tion, pp. 228-32, and J. Harvey W1lkinson, From Brown to Bakke· The
Supreme Court and School lntegratwn 1954-1978 (New York: Oxford Umver·
sity Press, 1979), pp. 274-89.
94 T H E T R I L E M M A 0 F E QUA L 0 P P 0 R T U N I T Y

also have a major effect on the self-esteem of those admitted

and a stigmatizing effect on all minority students, even those
who would have been admitted without any preferential pro-
gram.84
It is also worth noting that these programs may provide
minority group members who are more likely to serve particu-
lar communities in their professional life. If the latter purpose
were truly central though, it is arguable that special programs
open to all races, for students willing to bind themselves to
practice in particular areas or for particular groups, would
serve the goal more effectively and in a nondiscriminatory
way. 85 There is no doubt that nonminority group members can
provide effective professional service to minority groups. In the
United States, the history of civil rights cases and, indeed, the
record of the Legal Services Corporation demonstrates this for
the legal profession. The effective provision of professional
services is, I believe, a less telling argument than the one based
on role models and aspirations.
However these various secondary effects are evaluated, it is
doubtful that they can be formulated strongly enough to justify
a sacrifice in merit when that sacrifice is not accompanied by a
gain in one of the other principles, such as equality of life
chances. For unless preferential admissions were directed at
those who had, themselves, suffered significant socioeconomic
disadvantages, the preferential practices would be open to the
charge that they render the competition unfair by making it
turn on morally irrelevant native characteristics. Furthermore,
while some of the secondary effects noted above may be
significant-through providing role models for other students

84. See Thomas Sowell, Black Educatwn: Myths and Traged1es !New York:
McKay, 1972), p. 292, and Midge Deeter, "On Affirmative Action and Lost Self·
Respect," The New York T1mes, 6 July 1980.
85. Note that the Umvers1ty of Califorma at Dav1s Med1cal School actually
had a program of preference !only occasiOnally operational) for students who
were prepared to practice m areas of northern Cahforma that were short of
phys1cians. See Bakke v. Regents of UmveTSJty of Cahforma, 553 P.2d 1152, at
1157-5811976).
 THE T RILE M M A 0 F EQUAL 0 P P 0 R T U N IT Y 95

or potential professionals, or through raising the level of aspira-

tion among minority group members-there is no reason to
believe that these secondary effects will reach, much less,
significantly affect, the disadvantaged so as to bear on equality
of life chances. In the United States, where the disparity
between the black middle class and the ghetto underclass has
been growing disturbingly, visible benefits to the former may
have little effect upon the latter. 86
Some proponents of affirmative action would defend prefer-
ential treatment, even when those benefited were not disad-
vantaged or had not been subjected to previous discrimination.
The Department of Justice, for example, in its amicus curiae
brief in the Bakke case argued that "the consequences of
discrimination are too complex to dissect on a case-by-case
basis." Schools and employers "ought not to be confined to the
choice of either ignoring the problem or attempting the Sis-
yphean task of discerning its importance on an individual
basis." The separation of applicants on racial grounds alone,
into separate admissions processes, was therefore held to be
defensible because "when individual measurement would be
impractical, a state may properly use categorical means." 87 A
general category, race, provides a practical approximation to
the results of a case-by-case assessment.
If this kind of approximation is the central justification, I
have already noted reasons for questioning its applicability to
practices of preferential treatment based merely on race. At
least in highly competitive meritocratic processes (such as
admission to professional schools), class, income, and other
advantages will clearly affect the development of the relevant
qualifications. Members of the designated racial or ethnic
category who achieve the highest scores and other qualifica-

86. For a d1stressmg account of the contrast between the ghetto underclass
and the black middle class see the cover story, "The American Underclass:
Destitute and Desperate in the Land of Plenty," in T1me 110 (August 29, 1977):
14-27.
87. Bnef of The Umted States of America as am1cus cunae, Bakke, p. 56.
% THE TRILEMMA OF EQUAL OPPORTUNITY

tions are likely to be disproportionately those who have not,

themselves, come from relatively disadvantaged backgrounds.
Hence the irony of the DeFunis case. Even a perfunctory
investigation of income and family background would prevent
preferential treatment from being awarded to relatively advan-
taged members of minority groups, instead of to the truly
disadvantaged whose qualifications may appear less impres-
sive. As I will show later, a further irony was that the program
the court struck down in the Bakke case did include an effort
to target preferential treatment toward relatively poor appli-
cants. From our standpoint, the court's insensitivity to this
factor was regrettable.
It is important to distinguish the issue of compensation for
past discrimination from the issue of ameliorating develop-
mental disadvantages in the present that may result from
many other causes distinguishable from past discrimination.
Low income, class background, inadequate housing, a broken
home, or cultural or linguistic differences may be only tangen-
tially related, if they arc related at all, to previous acts of
discrimination. Yet children growing up in homes affected by
these disadvantages may be denied an equal life chance be-
cause their environments effectively inhibit the development
of their talents and aspirations.
Efforts to correct these developmental disadvantages-
through economic redistribution, intervention in the home
environment, or provision of special compensatory school-
ing-do not require any references to past discrimination as
the basis for their justification. A similar point can be made
about efforts to equalize life chances despite the persistence of
these developmental disadvantages, when the equalization
would result from programs of preferential treatment for those
who have actually suffered the disadvantages. In either case,
such efforts could be justified by an aspiration to realize
equality of life chances in order to bring about a fair system of
individualistic competition. The equality of life chances prin-
ciple would not distinguish cases where developmental disad-
 T H E T R I L E M M A 0 F E Q U A L 0 P P 0 R T ll N I T Y 97

vantages result from identifiable acts of discrimination from

cases where the disadvantages result merely from socioeco-
nomic inequality.
Yet perhaps compensation for past discrimination or other
past injustice may be interpreted to require the kind of pro-
gram I have been criticizing-preferential treatment for groups
classified merely by race or other arbitrary native characteris-
tics. However, it is the relatively poor and disadvantaged
members of a given minority group-precisely those least
likely to be benefited by preferential treatment applied merely
by race in competitive contexts-who are likely to have a
strong compensatory claim. Consider the compensatory argu-
ment that a person X is worse-off than he would otherwise
have been because his parents were discriminated against, or
because his distant forebears were enslaved. One approach
would be the systematic investigation of historical and genea-
logical claims based on the background of each applicant, in
order to arrive at some reasonable probabilistic construction of
the required counterfactual; in other words, how well-off X
might otherwise have likely been. Here the objection quoted
above from the Justice Department brief in the Bakke case has
considerable force. To investigate historical claims of compen-
sation on behalf of each applicant surely would amount to an
impractical burden in most of the social contexts where this
issue has been raised.
On the other hand, race alone surely offers an inadequate
proxy for identifying those subject to compensation since it is
far from clear that the more advantaged members of a racial
minority generally are worse-off than they would otherwise
have been, were it not for discrimination practiced against
their forebears in previous generations. A much more compel-
ling probabilistic claim could be made on behalf of the ghetto
underclass. Actual income and class background of one's par-
ents, when combined with race or ethnic background, might
provide a better, if still distant, approximation. It would more
clearly identify victims of historical injustice than would race
98 T H E T R I L E M M A 0 F E Q U A L 0 P P 0 R T U :-.1 I T Y

alone; it would identify present-day victims about whom it

could be said that they would, in some sense, have been better
off if their forebears had not suffered from discrimination or
other injustice.
If this compensatory argument were pursued seriously, it
would face two challenges. First, the list of groups previously
subjected to discrimination is far broader than the groups now
demanding compensatory benefits. Given the sad history of
professional schools with respect to Jews, DeFunis himself
might have had as valid a historical argument as did those
admitted ahead of him. Consistent pursuit of the argument
would produce a host of other historical ethnic claims-Irish,
Polish and Italian, Catholic zs well as Jewish, in addition to
Hispanics, native Americans, and Orientals, to cite only the
most obvious examples.
This Pandora's box is not fanciful. There has been public
agitation for affirmative action to be extended to many other
white ethnic groups presently underrepresented in American
higher education, for example, Italians, Greeks, Slavs, and
Poles. 88 Even narrow constructions of the compensatory argu-
ment have led to a proliferation of ethnic classifications. In its
survey of post-Bakke professional school admissions, the Anti-
Defamation League discovered one American law school that
classified its applicants into the following ethnic and racial
classifications:
(1) American Indian or Alaskan Native, (2) Chinese, (3)
Filipino, (4) Hawaiian, (5) Korean, (6) Japanese, (7) Pacific
Islander, "including Fijian, Micronesian, Samoan, Tahi-
tian, Tongan, etc." (8) Other Asian, "including Cambodian,
Sri Lankan, Laotian, Thai, Vietnamese, etc." (9) Mixed
Asian and/or Pacific Islander, (10) Black, not of Hispanic
origin, (11) Portuguese, (12) Puerto Rican, (13) Spanish, (14)
Mixed Hispanic, (15) White, "including any persons hav-

88. Smdlcr, Bakke. DeFums. and Mmnnty Adm1sswm. p. 266, and Wllkm-
son, From Brown to Bakke. pp. 268-69.
 THE TRILEMMA OF EQUAL OPPORTUNITY 99

ing origin in any of the original peoples of North Africa,

the Middle East, Indian Subcontinent or Europe Iexcluding
Spain and Portugal)" and 116) Mixed Ethnic Background-
"a combination of ethnic backgrounds which includes two
or more of the five major categories-American Indian or
Native Alaskan, Asian or Pacific Islander, Black, Hispanic
and White."89
Furthermore, if historical claims about discrimination in
this country provide the basis for compensatory programs, why
should recently immigrated Hispanics benefit from preferen-
tial treatment? They can make no such historical claims
unless the locus of responsibility is widened far beyond Ameri-
can domestic institutions. 90 While they may have a strong case
on the basis of actually suffered developmental disadvantages,
these recent immigrants do not have a similar claim based on
compensation for historical injustice.
An additional complication arises for those subjected histor-
ically to discrimination whose descendants fail to share the
characteristics that singled out members of an earlier genera-
tion for discrimination. Perhaps I am far worse-off now than I
would otherwise have been, because my grandmother was
discriminated against because she was a woman. It is clearly
possible for the counterfactual standard to apply, namely, I
would be better-off if she had not been discriminated against,
even though I do not share the characteristic that singled her
out for discrimination, namely, her sex.
There are equally perplexing issues about the locus of re-
sponsibility for paying the compensation. Some corporations
and institutions guilty of past discrimination still exist. Others
do not. However, many persons are now better-off than they
would otherwise have been, just as others are now worse-off.

89. See Anti-Defamation League of B'na1 B'nth, "A Study of post-Bakke

Admtss10ns Polic1es m Medical, Dental and Law Schools," Rights 10, no 1
(Summer 1979): 11-12.
90. Wilkmson, From Brown to Bakke, p. 278.
100 THE TRILEMMA OF EQUAL OPPORTUNITY

Some present-day advantaged persons are now better-off be-

cause their parents, grandparents, or great-grandparents did not
have to compete with groups discriminated against in previous
generations. Others made fortunes out of slavery or other
exploitative practices, fortunes that, in some cases, survive
intact to the present day. If the compensatory arguments were
to be taken seriously, some special responsibility for the pay-
ment of compensation might be placed on these persons, were
it possible to identify them nonarbitrarily.
In addition to the proliferation of claims, there is a challenge
of a different sort also facing the compensatory argument. An
individual, X, is supposed to be compensated by returning him
to the level of well-being he would have reached, were it not
for some identifiable injustice in the past committed against
his forebears. In tracing back through the generations, how·
ever, it soon becomes clear that for many of the cases for which
we might think compensation was owed, X would not have
existed, were it not for the injustice in question. In our con-
struction of the counterfactual, the historical chain of events
following from the hypothetical nonoccurrence of the stated
injustice, it is quite reasonable to suppose that we would find
no individual, X, existing in the alternative world thus envis·
aged.91
If Kunta Kinte, Alex Haley's ancestor portrayed in Roots, had
not been brutally kidnapped and sold as a slave, there is
virtually no likelihood that the author of Roots would have
come to exist in the twentieth century. The mating and repro-
duction of each generation, in turn, would have depended on a
host of contingent circumstances. A different genealogical line
would have been created were the chain to have been broken at
any point, for instance, if a parent, grandparent, or great·

91. I have been influenced m this argument by Derek Parfit's work on

identity and the moral claims of possible persons. See the literature cited m nn.
40 and 41, chapter 2. I have also benefited from George Sher's "Ancient Wrongs
and Modern Rights," Philosophy and Public Affairs 10 no. 1 (Wmter 1981):
3-17.
 THE TRILEMMA OF EQUAL OPPORTUNITY 101

grandparent were to have mated with someone else. Kunta

Kinte's descendants would have been native Africans, perhaps
residents of Juffure (Kunta Kinte's village in West Africa), had
the initial injustice not occurred.
The particular circumstances of slavery in one generation,
freedom in another, determined a host of contingencies affect·
ing the choice of mates and the identities of each generation in
turn. Any realistic construction of an alternative world in
which the enslavement of a given person X's ancestors had not
occurred would not be likely to include X, on whose behalf
present compensation is being demanded. How then, could
compensation for injustices to previous generations satisfy the
strict notion of compensation and, thereby, return X to the
level of well-being X would have enjoyed, had the injustice in
the past not occurred?
Or, to take a simpler case, while my grandmother may have
suffered from sex discrimination and while certain disadvan-
tages to her descendants may have causally followed, she
would clearly have had different descendants had it not been
for the sex discrimination. If we imagine her pursuing a profes-
sional career rather than that of a full-time housewife, she
would surely have had fewer than the five children she did
have; she might not have had any at all and the timing of her
children would surely have been different. If I attempt to
envision the alternative world that would have resulted, had
discrimination not occurred, I cannot plausibly locate myself
in it.
Perhaps this strict criterion of compensation should be loos-
ened. Perhaps X should be compensated to the level of well-
being that some closely related, identifiable person X' would
have reached, had the injustice not occurred. A viable theory of
individualistic compensation would require well-developed
criteria for identifying X' in the counterfactual world-for
identifying, for example, a descendant of Kunta Kinte who
would have come to exist had the injustice not occurred and
for comparing the level of well-being achieved by Alex Haley
102 THE T RILE M M A 0 F E Q ll A L 0 P P 0 R TV NIT Y

(before the writing of Roots) with that of the alternative

hypothetical descendant of his ancestors.
There are enormous challenges, both theoretical and empiri-
cal, facing the development and application of such a theory of
compensation for intergenerational injustice. Perhaps an ade-
quate version of this kind of theory might be developed,
although my argument here will not depend on it. I need only
note that were such a theory to be developed, it would have
implications that parallel, in essential respects, the conclusion
reached above. Plausible constructions of the required counter-
factuals would lead to the same objection presented earlier to
preferential treatment applied merely to racial or ethnic cate-
gories. Since there are clear pressures, in competitive contexts,
to admit the most qualified members of the specified racial or
ethnic groups, the most qualified members are likely to come
from relatively more advantaged rather than less advantaged
backgrounds. Yet it is precisely the relatively more advantaged
about whom it can least plausibly be said that they are worse-
off than they would have been (or than identifiable relatives
would have been) had not injustices occurred to previous
generations. Similarly, it is the less advantaged for whom the
counterfactual claim is most plausible and for whom the harm
(the gap between their present position and the counterfactual
supposition) is likely to be greatest (or at least far greater than
any claim that might be lodged on behalf of the more advan-
taged).
Hence, the compensatory argument, like the argument from
developmental disadvantages, leads us to reject preferential
treatment merely applied to racial or ethnic categories and
instead leads to requirements that preferential treatment, if it
is to be adopted at all, should be directed only toward the
disadvantaged. Income, class background, and whatever other
information could be handled administratively in a reasonable
manner would have to he taken into account. In addition, if it
were possible to identify more precise compensatory claims for
each individual such efforts would clearly be supported by this
kind of argument.
 THE TRILEMMA OF EQUAL OPPORTUNITY 103

One central difference between the compensatory argument

for past discrimination and the argument from developmental
disadvantages in the present is that the latter would clearly
require special efforts for all persons now suffering from specifi-
able and extreme disadvantages, regardless of their racial or
ethnic background. For example, preferential treatment for
poor whites would be required if the principal effort were to
overcome developmental disadvantages in the present in order
to equalize life chances. By contrast, the compensatory argu-
ment for past discrimination would not, presumably, include
poor whites. It would, by contrast, include consideration of
racial and ethnic classifications for all groups subjected to past
discrimination; these classifications would provide part of the
basis for identifying those now eligible for compensation ac-
cording to some counterfactual consideration of how well-off
they would have been lor how well-off appropriately related
persons would have been) had the discrimination not occurred.
It is worth noting that both of these versions of preferential
treatment targeted toward the disadvantaged-the version
aimed at compensation for past injustice and the version aimed
at ameliorating present developmental disadvantages-should
satisfy our proposed minimum condition. In either case, the
sacrifice in merit would clearly be accompanied by a gain in
one of our other principles, in this case, equality of life
chances. Such policies would benefit persons from disadvan-
taged backgrounds who could not otherwise realistically aspire
to the higher positions. Of course, satisfaction of our mini-
mum condition, no sacrifice in one of our principles without
an improvement in one or more of the others, 92 does not settle
the question of how these conflicting claims are to be weighed
or balanced one against the other. It does, however, remove the
easy objection to the mistargeting of benefits Ito the nondisad-
vantaged) from preferential treatment merely applied to racial

92. An Improvement m one or more of the other pnnciples, or an improve-

ment in some other morally relevant factor that would then have to be argued
for. See the earlier d1scusswn of the minimum condition at the beginning of
this section.
104 THE TRILEM MA OF EQUAL OPPORTUNITY

or ethnic categories, in competitive meritocratic contexts.

While our minimum condition does not resolve the general
problem by any means, it is far from trivial in its implications
for actual policy.
Let me return to my original thought experiment. The point
of imagining attempts to realize all three principles under ideal
conditions was to argue that simultaneous commitment to
fulfilling all three amounted to an incoherent ideal under the
best conditions that might realistically be assumed for a large
industrial society, provided that background conditions of in-
equality were left in place.
Within the liberal framework of individualistic competition,
the departure from ideal theory required to approach the actual
American policy context raises the problem of what would be
necessary, at any point in time, to implement "a fair competi-
tion among individuals for unequal positions" when the imme-
diately relevant recent history has produced assignments of
persons to positions that were unfair because they were based
on discrimination.
The question is whether a person is worse-off than he ought
to have been in the competition because one or more of the
principles of fair competition being implemented !merit or
equality of life chances) has been violated in the past. As we
have seen, compensation for previous violations of merit, that
is, discrimination, may require present violations of merit,
that is, preferential treatment. Similarly, attempts to achieve
equality of life chances jwith the autonomy of the family in
place) would require sacrifice of merit (whether under ideal or
under less than ideal conditions). In either case, the benefits of
defensible preferential treatment must be directed at those
individuals who have suffered from past discrimination or
from unequal developmental conditions. Preferential treat-
ment when targeted carefully in that way would be compatible
with the basic aim of implementing a fair competition among
individuals for unequal positions. Either by compensating for
previous violations of merit or by achieving greater equality of
 THE TRILEMMA OF EQUAL OPPORTUNITY 105

life chances lor both), such specially designed programs of

preferential treatment would lead to an improved realization of
one or more of the principles in this individualistic framework.
By contrast, it would not be an improvement in fair individu-
alistic competition to sacrifice merit, through programs of
preferential treatment based merely on race or ethnic back-
ground, without also accomplishing greater equality of life
chances. In empirical contexts where they produce substantial
mistargeting, such programs do not compensate individuals for
past violations, nor do they contribute directly to realizing
in the present either of the principles of fair competition !merit
or equality of life chances). 93 Their secondary effects on the
framework of individualistic competition-provision of role
models, effects on aspirations-are not likely to substantially
affect the life chances of the less advantaged. And they come at
the cost of a severe sacrifice in merit. As we shall see in the
next chapter, one would have to depart strikingly from our
original liberal framework of individualistic competition in
order to justify such programs.

93. Of course, in some empuical contexts preferential treatment by race

may not turn out to produce substantial mistargetmg because virtually all
members of the group, even the most disadvantaged, have the requisite skills.
That is why the objection developed here was restncted to highly competitive
mentocratic contexts. My obJection is to the mistargetmg phenomenon and
not to the use of preferential treatment per se.
4. FAIRNESS, GROUP
COMPENSATION,
AND EQUALITY:
RESPONSES TO
THE TRILEMMA

I nclaim
the last chapter, my central
was that given back-
ground conditions of inequality, commitment to both princi-
ples of equal opportunity produces an incoherent ideal when it
is combined with our customary commitment of the auton-
omy of the family. The principles of merit, equal life chances,
and family autonomy conflict according to a systematic pat-
tern. Like a three-cornered stool with only two legs available
to prop it up, this combination of commitments is unsupport-
able, even as an ideal for guiding public policy under the best
conditions that might realistically be imagined for a modern
industrial society. One strategy would be to accept one of the
three unpleasant options in the trilemma-accepting the re-
quired sacrifice in merit, equal life chances, or the autonomy of
the family. Another strategy would be to reformulate either
our basic commitment to these notions of equal opportunity or
our basic commitment to the autonomy of the family. The
only remaining alternative compatible with these commit-
ments would be to relax the assumption of background condi-
tions of inequality and, hence, to prescribe equality of out-
comes rather than merely equality of opportunities.
This last alternative confirms our earlier conclusion with
respect to the family. Equal opportunity, far from being the
weakly reformist notion sometimes alleged, 1 is a startlingly
1. See Schaar, "Equahty of Opportumty and Beyond," m Pennock and
Chapman, eds., Nomos IX Equahty. See also Bowles and Gmtis, Schoolmg m
Capitalist America, for a cnuque of the leg1t1matmg function performed by
those verswns of "equal opportunity" compatible with a capitalist economy.

106
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 107

radical idea. Either it would require systematic intrusions into

the family and a vast revision in the way of life we commonly
take for granted, or it would require the even more radical
reforms required for the systematic elimination of both social
and economic background inequalities. In either case, equal
opportunity could be fully implemented only at a considerable
cost in liberty-either in the private sphere of liberty at issue
in the family, or in the liberties of property, contract, and
employment that would have to be mfringed in order to main-
tain equality of outcomes.
I interpreted the principles of merit and equal life chances as
essential components of the basic competitive assumption-
that there should be a fair competition among individuals for
unequal positions in society. Let us consider possible revisions
in the crucial components of this definition. These may be
identified as: "fairness," "competition," "individuals," and
"unequal positions." I will consider these in the sections
below.

4.1 Fauness and Merit N otions of fairness were in-

volved explicitly in the argu-
ments for merit and equal life chances. The principle of merit
is, itself, the expression of a kind of procedural fairness. More
demanding notions of "background fairness" can be employed
to support equal life chances. Various modifications of these
notions of fairness might be introduced in order to avoid any
initial commitment to the principles yielding our trilemma.
Recall the earlier definition:
THE PRINCIPLE OF MERIT.There should be widespread proce-
dural fairness in the evaluation of qualifications for posi-
tions.

The limited implications of equal opportumty wnh respect to race are probed
m Edwm Darn's Rules and Racial Equality !New Haven: Yale University
Press, 1979). See my discussion of some of those contentions below.
108 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

And by qualifications I meant: criteria that are job-related in

that they fairly can be interpreted as indicators of competence
or motivation for an individual's performance in a given posi-
tion.
If this definition of qualifications were to be revised so that
membership in a given race, family, class, or ethnic group
could, in itself, count as a qualification to be considered
impartially in the process of meritocratic competition, then
our trilemma of equal opportunity could be avoided. For then
both equal life chances and merit, in the revised sense deter-
mined by such a new definition of qualifications, could be fully
instituted without the autonomy of the family being affected.
Despite the unequal development of talents and other qualifi-
cations protected by family autonomy, qualifications could be
defined in such a way that persons could be routed into the
higher positions in the precise proportions required by equality
of life chances. By redefining qualifications, the resulting prin-
ciple of merit could be formulated to mirror the precise effects
of the kind of preferential treatment considered earlier. This
new definition simply would make explicit the new terms of
competition.
It would be difficult, however, to consider the new competi-
tion a fair one in the same sense. Family background, race, or
sex do not support the same justificatory burdens of job rele-
vance as do competence and motivation. Furthermore, they are
not earned; they are not subject to the same process of compet-
itive acquisition. And if technological innovations permitted
them to be acquired-permitting one to change races, for
example, as sexes are now sometimes changed, that would not
increase equality of life chances for native members of such
groups for it might only permit newly altered proxies to take
advantaged places instead.
The notion of merit, as originally defined, captured a claim
to procedural fairness that carefully insulated the assignment
process from factors that were irrelevant to performance in the
positions to be filled. One could view the results of meritocratic
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 109

competition as merited or entitled by competitive efforts at

acquisition and then by fair evaluation of the results. Redefin-
ing merit so as to explicitly include race or family background
would constitute a disturbing departure from this position for
most liberals. One cannot earn one's race or family background
in the same way that one is supposed to earn the qualifications
evaluated by merit.
Perhaps the principle of merit needs to be sacrificed in order
to fulfill other important goals such as equal life chances or
compensatory claims for past injustice. But the cost of sacrific-
ing the principle only would be obscured were qualifications to
be redefined so that merit could include membership in a given
race, class, or ethnic group. 2
This redefinition of merit sacrifices the claim to procedural
fairness bult into the original principle, but might be held to
serve background fairness.3 Let us now turn to the alternative
strategy of sacrificing background fairness but maintaining the
commitment to procedural fairness. Suppose I interpreted "a
fair competition for unequal positions" (our initial "fair com-
petition" assumption) to require only adherence to merit in our
original sense?
To reduce equal opportunity to the limited procedural re-
quirements defined by merit would constitute, I believe, a
drastic and disturbing truncation of an otherwise inspiring
ideal. Such a limited conception is open to the obvious objec-
tion that is would be satisfied even by the limited reforms
2. Justice Powell's optmon in Bakke is a step m thts direction; race may be
constdered expbcitly as "one element to be wetghed fairly against other
elements m the selection process." See Alfred A. Slocum, ed., Allan Bakke
versus Regents of the Umversity of Callfornia 6 vols. !Dobbs Ferry: Oceana,
1978), "Optmon, Majority by Mr. Justice Powell," 6:355. See Anti·Defamatlon
League, "A Study of post-Bakke Admissions Polictes," for evidence that race
has been commonly treated smce as a major qualification in its own right.
3. Dependmg on how much weight mere membership in a given race or
class were accorded m such a new definition of merit, this strategy might
mcrease equabty of bfe chances. But it would do so in the same way that the
second opuon m our tnlemma dtd so, by sacrificmg the clatm to procedural
fairness offered by ment m our origmal sense.
110 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

exemplified in our imaginary warrior society. Recall that in the

warrior society a competition was instituted to determine the
best warriors in order to regulate admission to the upper strata.
Children of the present warriors continued to dominate the
competition because they were well-nourished, while children
from other strata verged on starvation. There is no doubt,
however, that the competition did, indeed, select the best
warriors. Yet the specter, let us imagine, of 300 pound Sumo
wrestlers vanquishing 90 pound weaklings offers a disturbing
ideal of equal opportunity. The reason, of course, is that grossly
unequal developmental conditions in this simplified case (dif-
ferential nourishment) yield similarly unequal life chances.
The conditions underlying a competition can be judged unfair
if we can predict how people will do in it merely from knowl-
edge of the strata into which they were born. Hence, the notion
of background fairness that yields equal life chances seems just
as essential as did the notion of procedural fairness that yielded
merit. Sacrificing either principle would produce a truncated
and inadequate ideal of equal opportunity. Both types of fair·
ness, as I have interpreted them, seem to provide essential
components of the liberal doctrine of equal opportunity.

4.2 Competition and

Lotteries E ach of these departures from
our earlier analysis of a fair
competition would require a disturbing revision of liberal
assumptions. Perhaps, instead of modifying notions of fairness,
we might modify notions of competition. A lottery system
might be formulated to preserve procedural fairness. Unbiased
by morally irrelevant factors such as race, sex, class, or family
background, it could be instituted to achieve equal life
chances. In that way, it would offer a strategy for realizing both
procedural and background fairness without requiring the sac-
rifice in liberty identified by the autonomy of the family. This
strategy is an example of the kind of revision of fundamental
assumptions that would evade our trilemma by abandoning
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY Ill

one of the central principles, merit, and replacing it with an

alternative claim to procedural fairness.
In this kind of lottery system, persons would be regarded as
passive recipients of statistically equal chances rather than as
active participants in a competition. No person's actual behav-
ior would affect where he or she would end up. The result
would represent a severe challenge to liberty in employment
although it would not affect the private sphere of liberty
involved in the family. If I wanted to be a doctor but the lottery
required that I be an auto mechanic, or if I wanted to be a
lawyer but the lottery required that I be an orchestra conduc-
tor, my liberty to seek and compete for employment would be
forfeited by the lottery system. Concurrently, there would also
be a severe cost in efficiency. If people were assigned regardless
of aptitude and developed competence, some gross mismatch-
ings could be presumed to result. My assignment to be an
orchestra conductor would surely constitute a somewhat fan-
ciful example of this kind of mismatching. I speculate that no
amount of on-the-job training could ever be expected to lead to
my performing that function competently. Meanwhile, many a
nascent conductor might be assigned to positions where I
would blossom.
Of course, we might also imagine a pure lottery system
combined with provisions for trades or even a market. The
sacrifice in competence and efficiency again would be severe.
Imagine X, who has always wanted to be a conductor but who
has no talent in the field whatsoever, trading his lottery enti-
tlement for mine. He might be happier in the position than I
would but no improvement in competence need result. And if
a market so as to allow side payments were permitted lor if a
black market in lottery tickets were not prevented), then
background inequalities would enable those who were willing
and able to spend more to compromise the principle of equal
life chances. It would permit the rich, or their children, to buy
up the most desirable lottery entitlements.
Any market in lottery entitlements, or any system permit-
112 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

ting trades in positions, could be expected to undermine the

equality of life chances produced by the lottery initially. Some
people will be better traders or have better connections, and
some entitlements will be far more tradeable (depending on the
distribution of preferences and skills). Even if explicit side
payments are ruled out, a host of arbitrary factors would
permit some people to parlay their lottery entitlements into
much better than equal life chances, if trades on markets were
permitted.
On the other hand, when this lottery notion is interpreted
strictly so as to realize equal life chances, it offers a stark
contrast to the liberty of employment presumed by the princi-
ple of merit. Merit permits each of us to participate actively in
the process of assignment. We are free to decide what we wish
to compete for; we are free to attempt to acquire the appropri-
ate qualifications (restricted only, I assume, by meritocratic
competition for the opportunities to undergo the requisite
training).
The lottery, by contrast, would substitute equal chances for
equal opportunities; it would substitute equal chances, pas-
sively, to receive positions for equal opportunities, actively,
to compete for those positions. The lottery would, admittedly,
offer procedural fairness by ruling out discrimination or prefer-
ential treatment based on race, sex, class, or ethnic or family
origin. Because it would also be insensitive to individual com-
petence and preference, however, it would require sacrifices in
both efficiency and liberty of employment-valu es that are
both promoted by the principle of merit.
Thus the lottery might achieve equal chances in a procedur-
ally fair way, but at a severe cost in individual initiative and
liberty. People would become passive recipients of equal
chances rather than active seekers of equal opportunities. If
taken seriously, the lottery notion would require a drastic
departure from our basic competitive assumption. Although it
offers a strategy for evading the trilemma, it does so, I believe,
at a cost that most liberals would find deeply disturbing. While
 FAIRNESS, G R 0 U P C 0 M P ENS AT I 0 N, A N D E Q U ALIT Y ll3

it must be counted among the theoretically possible solutions,

the costs in both efficiency and liberty render it a less-than-
inspiring ideal for liberal theory.

4.3 Group ComP_etition Another strategy for avoiding

and CompensatiOn the trilemma would be to re-
lax the assumption that the fair competition must take place
among individuals. If it were conceptualized as a group compe-
tition, then there would be a defense for precisely those poli-
cies of preferential treatment we objected to earlier.
The notion of a fair competition among groups leads natu-
rally to claims that members of the group should get a repre-
sentative, or even proportional, share of the desirable social
positions. If blacks, Hispanics, or American Indians represent,
respectively, X,, X2, and X3 percent of the total population, it is
natural to question why they do not also constitute the same
percentages in any given desirable profession, class, or ranked
social position. And if the reply is based on reference to their
percentages in the relevant qualified pools of applicants, then
the same question can be raised about group percentages in
each of the preceding steps in the sorting process. The focus
changes to the percentages of high school and college gradu-
ates, or applicants to professional schools, who turn out to be
black, Hispanic, or American Indian. 4
Receiving less than their proportionate share in any of these
cases provides groups with a basis for claims that the terms of
competition among groups must, in some way, be unfair.
Perhaps there is discrimination or some systematic inequality

4. See Dorn, Rules and Racwl Equality, chapter 4, where th1s group percent·
age notion is offered as an ideal for racial and ethnic equahty. He does not,
however, mean 1t to be applied too ngidly to each particular occupation. If
blacks do not compnse "exactly 11 percent avocado farmers or psychiatrists,"
he would not have cause for complaint (p. 126). Rather, the focus should be on
substantial departures from proportionate representation. See also, for exam-
ple, Smdler, Bakke, DeFums, and Mmonty Adm1sswns. pp. 35, 55-57, and
26fr67 for some other examples of group proportwn calculatwns.
114 FAIRNESS, GROUP COMPENSATION, AND EQUAliTY

in causal conditions. Or perhaps cultural differences have an

effect on the development of aspirations. Whatever the root
cause of disproportionate shares, the notion of a fair competi-
tion among groups, rather than among individuals, would lead
naturally to some redress via a closer allotment of proportional
shares. Hence, an argument would follow for a strong version
of preferential treatment for members of the group.
As long as fairness claims are conceptualized among groups,
rather than among individuals, there would be no compelling
claim for any commitment to the principle of assignment by
merit. The latter notion was defined, of course, in terms of
individual qualifications. Assignment by group membership
might seem unfair to individuals but it could be managed so as
to achieve a recognizable notion of fairness among groups,
namely, each group receiving what is, in some sense, held to be
its "representative" or "fair" share of the desirable social posi-
tions.
Such a departure from our basic assumption about competi-
tion among individuals would avoid the trilemma by avoiding
the commitment, in the first place, to the principle of merit
that is sacrificed by preferential treatment. Owen Fiss offers a
particularly well-developed and influential example of argu-
ments for the explicit consideration of groups. He reasons that:
There are natural classes or social groups in American
society and blacks are such a group. Blacks are viewed as a
group; they view themselves as a group; their identity is in
large part determined by membership in the group; their
social status is linked to the status of the group; and much
of our action, institutional and personal, is based on these
perspectives. s

5. Owen M. F1ss, "Groups and the Equal Protection Clause," in Cohen,

Nagel, and Scanlon, eds., Equality and Preferentwl Treatment, p. 125. Ftss also
mcludes two conditiOns that must be satisfied by such a social group: "II) It 1s
an entity, . . you can talk about the group without reference to the particular
individuals who happen to be Its members at any one moment.j2) There 1s also
a condition of Interdependence. The 1dent1ty and well-bemg of the members of
the group and the Identity and well-being of the group are hnked." lp. 125).
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 115

Fiss argues that when this conception of blacks as a "social

group" is combined with a realization of their long-term status
as an "underclass," the basis is laid for preferential treatment
according to group characteristics. He mentions two crucial
factors:
One is that blacks are very badly off, probably our worst-
off class (in terms of material well-being second only to
the American Indians), and in addition they have occupied
the lowest rung for several centuries. In a sense, they are
America's perpetual underclass. It is both of these charac-
teristics-the relative position of the group and the dura-
tion of the position-that make efforts to improve the
position of the group defensible.li

Fiss considers preferential treatment (or, more generally,

"redistribution") as justified by two possible arguments. One is
compensation, the other is the ethical objection to "castes."
Consider the first:
This redistribution may be rooted in a theory of compen-
sation-blacks as a group were put in that position by
others and the redistributive measures are owed to the
group as a form of compensation. 7
This notion of group compensation would face the difficul-
ties we encountered earlier with individualistic compensation.
It is worth noting, however, that the identity issue does not
arise in precisely the same way. While it is arguable that a
particular present-day black would not be here (or would not
exist at all) were it not for injustices committed in the past,
(hence, attempting to compensate him to the level he would
have reached were it not for those injustices raises the identity
problem, namely, that he would not otherwise have been) the
same cannot be said for blacks as a group. The persons to
whom the injustices were committed would, in general, have
6. Ib1d., p 127.
7. Ibid.
116 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

had other descendants and the well-being of that alternative

group can be considered in the required counterfactual.
Of course, those descendants might not have been Ameri-
cans had it not been for the initial act of enslavement and
transportation to America. For these purposes, however, we
can distinguish the initial injustice committed against blacks
in bringing them to this country from all of the subsequent
injustices committed against them, once here. It should be
clear that the latter were so severe as to provide a basis for
compensatory arguments, were such arguments to be admitted
at all, without any necessity to take account of the former as
well.
At first glance, then, group compensation appears to evade
the identity problem because the same group (if not the same
individuals) could be presumed to exist in the counterfactual
world that would have arisen had the injustices for which
compensation is being demanded not taken place. The problem
of maintaining individual identity in the counterfactual world
does not appear to apply to the group.
However, the problem still does arise but in a different way.
While there would still be American blacks, it is arguable that
had the injustices to their ancestors not occurred, American
blacks would not constitute a social group in the same sense.
Without a legacy of slavery, discrimination, and other injus-
tices, blacks might not constitute a social group or "natural
class" in Fiss's sense. In other words, were it not for these
previous injustices, it is likely that their identity and status
would not be linked, as they are now, to their group member-
ship. It is likely that they would not satisfy one of the neces-
sary conditions for being a social group, a condition Fiss calls
"interdependence":
Members of the group identify themselves-explain who
they are-by reference to their membership in the group;
and their well-being or status is in part determined by the
well-being or status of the group. 8
8. Ibid., p. 125.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 117

Were it not for a history of injustice and discrimination

rendering Americans color conscious, blacks might very well
not constitute an "interdependent" group in this sense. Absent
such a tragic history, the mere existence of physical differences
surely would not be enough to fulfill this condition. Blue-eyed
persons are physically distinguishable from the rest of the
population, but they do not define their identities in terms of
eye color. Neither do they view their well-being or status as
determined, even in part, by the well-being or status of other
blue-eyed persons. On the other hand, we might speculate that
had there been a history of discrimination and injustice di-
rected at blue-eyed persons, consciousness of eye color might
have grown to the point that blue-eyed persons would have
come to constitute a social group in Fiss's "interdependent"
sense.
Similarly, if we were to embark upon the admittedly difficult
task of imagining the alternative world lor worlds) that might
have occurred had injustices to blacks not been committed, it
is arguable that we might find a society in which race
functioned somewhat the way eye color does now. 9 In a ra-
cially neutral society, blacks would not constitute a social
group or natural class. Their status, identity, and welfare
would not be tied to their group membership. Therefore, but
for the injustices for which compensation is being advocated,
the group to be compensated would not exist as a social group.
And in this way, we see that a theory of intergenerational
group compensation faces a version of the identity problem-
the group cannot be returned to the level of well-being it would
have enjoyed, had the injustices not occurred, because had the
injustices not occurred, it would not have been a group, at least
in the same strict sense.
One might contend that compensating the group, per se, in
the alternative world is not as important as compensating the

9. I ta~e the eye color analogy from R1chard A. Wasserstrom, Philosophy

and Soc1al Issues: Five Stud1es (Notre Dame: University of Notre Dame
1980), p. 15. '
118 FAIRNESS, GROUP COM PEN SAT I 0 N, AND EQUALITY

individuals who would have comprised the group. But this

would return us to individualistic compensation, a principle
for which identity issues arise even more dramatically. There
is a clear basis for claiming that the particular individuals who
comprise the group in the world actually resulting, in part,
from previous injustices, would not exist in the alternative,
hypothetical world that would have resulted had the injustices
to past generations not occurred. 10 Similarly, the individuals
we would find in an alternative world, that lacked our present
social groups, would not be the same individuals as the mem-
bers of the group for which compensation is now being advo-
cated.
Because we find neither the same group nor the same indi-
viduals in the alternative world, identifying the level of well-
being that would have resulted had the injustices not occurred
becomes problematical. Group compensation arguments face
the challenge of providing a theory of group identity for such
counterfactuals. In what sense ought we to compensate groups
for their loss of well-being, had injustices not occurred, when
had it not been for the injustices, neither the group nor the
particular individuals within it would likely exist today?
One kind of compensation argument that might be envi-
sioned would avoid this identity problem in formulating the
counterfactual. It is arguable that some forms of existence are
so terrible that it would be better never to have existed at all
rather than experience such misery. If an individual experi-
ences so deprived an existence that we can plausibly say of
him, "it would have been better, had he never been born," then
his nonexistence provides a coherent construction of the coun-
terfactual. 11

10. For some interesting calculations showmg how even small divergences
m population in one generation will produce vast differences m the population
membership of later generations, see Thomas Schwartz, "Obhgations to Pos-
tenty," m Sikora and Barry, Obhgatwns to Future Generatwns, pp. 3-13,
especially pp. S-6.
11. See chapter 3, n. 31.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 119

Two points should be made about this variant of the com-

pensation argument. First, although it avoids the identity
argument, it does so in a manner that renders it similarly
difficult to determine the appropriate benchmark for compen-
sation. What is the level of actual well-being that corresponds
to nonexistence? The persons in question are so miserable that
bringing them into existence to live such a life was a form of
harm-a harm they should be compensated for by returning
them to a level of well-being that somehow corresponds to the
one they would have experienced, if the harm had not oc-
curred. But if the harm had not occurred, they would not have
experienced anything at all, since they would not have existed.
Death, the interruption of an existence already begun, should
not be confused with the nonexistence of those who were
never born (or conceived). It should be clear that intergenera-
ti.onal compensation arguments are as problematical in this
construction as they are in the vanants already considered.
A second point about this version of the compensation
argument is that it only supports a compensatory claim on
behalf of those who actually suffer such an extraordinarily
miserable existence. Hence 1t would not support compensa-
tion for persons identified merely on the basis of group charac-
teristics such as race or ethnic origin, unless empirical circum-
stances made it reasonable to suppose that virtually all
members of the group actually suffered, themselves, an exis-
tence so miserable that the counterfactual claim applied to
them-namely that it would have been better had they never
been born.
Hence the strategy of tying compensation to hypothetical
nonexistence is not only subject to some extraordinary diffi-
culties, it also lends little support to group compensatory
arguments in the present context. And as we have seen, any
viable theory of group compensation would face the identity
issue in the distinctive version that arises for groups. Let us
suppose, however, that this difficulty is somehow overcome.
Suppose that advocates of group compensation provide us with
120 FAIRNESS. G R 0 l' P C 0 M PENS AT I 0 N, AND EQUALITY

a theory of group identity for counterfactual worlds such that

we can determine the hypothetical level of well-being that a
group, or at least a collection of persons, would have reached,
had it not been for the injustices committed to previous
members of an identifiable group.
Were it available, such a theory would permit us to deter-
mine the amount of compensation owed, or at least, the level
of well-being that we ought to attempt to bring to the group
being compensated. However, such a theory would still require
a justification for compensating all present members of the
group (defined, for example, on racial or ethnic grounds) rather
than merely particular persons who had, themselves, suffered
some deprivation requiring compensation.
One rationale for such a policy of group compensation might
be found in the interdependence condition proposed by Fiss as
part of the definition of a social group or natural class. Remem-
ber this condition provides that the "well-being or status" of
any member of the group is "in part determined by the well-
being or status of the group." Depending on the extent of this
interdependence, it might be argued that improvements in the
well-being or status of the whole group might be interpreted as
improvements to the individual members, since the individual
status or well-being of each member is, by hypothesis, "in part
determined by the well-being or status of the group." Of
course, the issue turns on what is meant by "in part." A quite
extraordinary degree of interdependence would be required
were this condition to provide the rationale for group compen-
sation.
Would any of us seriously maintain that members of the
"perpetual underclass" (to use Fiss's term) are compensated for
past injustice by policies of preferential treatment that do not
affect their own life chances directly, but that affect, instead,
the life chances of other persons who are members of the same
group? The average or total level of well-being of a given racial
or ethnic group could be improved by such policies of preferen-
tial treatment. However, if those policies are not directed so as
 FAIRNESS, G R 0 UP C 0 M PENS ATI ON, AND EQUALITY 121

to help precisely those persons for whom individualistic com-

pensation might be demanded (those who have, themselves,
actually suffered significant deprivations), then any claim that
the persons left out are compensated through "interdepen-
dence" becomes implausible. Whatever satisfaction or pride
unemployed ghetto blacks might take in the visible income
and status of upper-middle class blacks, it must be quite paltry
compared to what they would feel were they to reach the same
positions themselves. It is unrealistic to expect the interdepen-
dent aspect of group welfare functions to make up for the fact
that compensation policies based merely on group member-
ship are not targeted so as to reach those who have, them-
selves, suffered greatly from the injustices for which compen-
sation is being demanded.
Such a theory of group compensation, then, must depend on
an assumption about the appropriateness of groups, rather than
individuals, as the unit of analysis such that the group as a
whole can be regarded as having been compensated for a past
injustice even when many of the individuals actually receiving
compensation never suffered from the injustice, and even
when many of the individuals who have suffered from the
injustice have not actually received compensation. According
to such a theory, so long as the group as a whole is sufficiently
better-off, mistargeting of the compensation to individuals
would not undermine the claim that compensation for past
injustices has been fulfilled.
Unlike individual compensation, group compensation theo-
ries would be vulnerable to a basic objection now commonly
offered against utilitarianism-the objection that they fail to
take account of the separateness of persons. Just as utilitarian-
ism may tolerate extreme disadvantages to some persons pro-
vided that enough others are benefited, this form of group
compensation would tolerate persistent injustices to some
persons, provided that other persons (who have not themselves
been victimized but who are members of the same group)
receive sufficient benefits. Just as utilitarianism might even
122 FAIRNESS, GROUP COMPENSATI0:\1, A!IID EQl!ALITY

justify slavery under some possible conditions (provided that

the rest of the society receives sufficient aggregate utility from
the practice), the group compensation argument could lead us
to tolerate the persisting effects of slavery and injustice im-
posed on some persons, on the condition that other members
of the same group receive sufficient benefits. 12
An aggregative principle, as noted in chapter two, must be
indifferent between differing distributions of goods or welfare
provided that the total (or with population held constant, the
average) is the same under the differing distributions being
considered. A policy of group compensation, merely concerned
with total or average levels of welfare, income, or status-for
example, for blacks or Hispanics as groups-would have this
same aggregative feature.
As long as preferential policies raised the average or total
level of the group, that would be justification enough according
to this kind of principle. How the benefits of preferential
treatment were distributed within the minority group being
compensated would not affect the basic issue. This is an
implication that Fiss accepts forthrightly. He considers the
possibility that programs of preferential treatment might be
administered so as to avoid favoring "rich blacks." Such an
exclusion is not appropriate, we are told: "wholly apart from
considerations of administrative convenience, the decision not
to exclude the rich blacks (even once identified) can be justi-
fied."13 He explains:
The claim is that the preference of the rich blacks may be
justified in terms of improving the position of the group.
Even if the blacks preferred happen to be rich, a benefit
abounds to the group as a whole. Members of that group
have obtained these positions of power, prestige and influ-
12. For the w1de mfluence of this "distinctively modern cntic1sm of utihtar-
lamsm," see H. LA. Hart's d1scuss1on of the separateness of persons argument
m "Between Uulity and R1ghts," m Alan Ryan, cd., The Idea of Freedom:
Essays m Honor of lsawh Berlm IOxford: Oxford Umvemty Press, 19791 p. 7H.
13. F1ss, "Groups and the Equal Protection Clause," p. 140.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 123

ence that they otherwise might not have and to that extent
the status of the group is improved. 14
Preferring poor persons, Fiss adds, would not help "the poor
conceived as a group-the preferred individual merely leaves
the group/' presumably, because he is no longer poor. Fiss's
conclusion:
Even if there were group benefits entailed in a preference
for the poor, certainly legislators or administrators are
entitled to rank the improvement of blacks as a group as a
social goal of first importance, more important than ele-
vating the poor conceived as a social group. 15
A group preference principle, like the one advocated by Fiss,
has the property that as long as the position of blacks, taken
generally as a group, is improved, it does not matter whether
the particular persons benefited are the already advantaged, the
"rich blacks," or members of the ghetto underclass. This kind
of group preference seems particularly open to objection when
it is justified by a notion of compensation for past injustice.
Consider, for example, the contrast between such a policy and
German efforts after World War II to provide restitution to the
victims of Nazi persecution. Accordmg to a 1949 West German
law, these efforts,
provided compensation for (a) loss of life, (b) damage to
limb or health, (c) deprivation of liberty, (d) damage to
property and possessions and (e) damage to economic
advancement, to persons persecuted in Germany because
of political conviction, or on racial, religious or ideological
grounds. 16

14. Ib1d., emphasis has been added.

15. Ib1d.
16. These proviSions are quoted m Derek L. Phillips, Equality, fustJce and
Rect1{icatwn· An Exploratwn m NormatiVe Socwlogy (New York: Academic
Press, 1979), p. 285.
124 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

Of course, Jewish survivors of the Holocaust were, over-

whelmingly, the focus of this compensation program. Even the
separate program of compensation paid to the state of Israel
was justified in order 11to assist in the integration of uprooted
and destitute refugees from Germany and from lands formerly
under German rule. 11 I7
It is important to note the careful routing of compensation
to actual victims. Of course, this routing was not perfect. In
1956, for example, the criteria were loosened slightly to permit
11 probabilistic" causal connections between persecution and

damage to health as a sufficient condition for compensation. 18

Likely victims were, in other words, given the benefit of the
doubt.
The occasional mistargeting of benefits in individual cases
that undoubtedly followed from the admission of probabilistic
connections is, however, a minor matter compared to the
massive mistargeting that could follow from any principle of
group compensation, per se. Would any reasonable person have
supported compensation to Jews, as such, regardless of their
personal history of persecution? Would it not have been outra-
geous if payments had been made, for example, to American
Jews untouched by the Holocaust-in 11 compensation" for
Nazi crimes-while many of the actual victims were uncom-
pensated by the program?
Jews are, to a significant extent, an "interdependent 11 social
group in Fiss's sense; yet the parallel principle of aggregative
group compensation would have been an indefensible response
to Nazi atrocities. Merely raising the average or total level of
income or welfare, of Jews as a group, to the level they would
have reached had it not been for the persecutions, would be an
unacceptable substitute for compensating those individuals
who had, themselves, suffered from Nazi atrocities. For one
could raise the average or total level of the group without
compensating any of the actual victims.
17. Ib1d., p. 286.
18. Ib1d., p. 285.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 125

The same charge Rawls directs against utilitarianism can be

directed against such an aggregative group principle: it "does
not take seriously the distinction between persons." 19 It per-
mits harm or deprivation to some persons to be tolerated for
the sake of benefits to other persons, just as if there were one
person or social entity that, somehow, experienced them all on
behalf of the group. Nozick makes a similar argument against
utilitarianism, an argument that we can apply, with equal
appropriateness, to the group compensation principle:
Individually, we each sometimes choose to undergo some
pain or sacrifice for a greater benefit or to avoid a greater
harm.... Why not, similarly, hold that some persons have
to bear some costs that benefit other persons more, for the
sake of the overall social good? But there is no social entity
with a good that undergoes some sacrifice for its own good.
There are only individual people, different individual peo-
ple, with their own individual lives. Using one of these
people for the benefit of others, uses him and benefits the
others. Nothing more.2o
As Nozick concludes about the person whose misery is toler-
ated for the sake of benefits to others: "that is the only life he
has. He does not get some overbalancing good from his sacri-
fice, and no one is entitled to force this upon him."2I
Group compensation, as contrasted to individual compensa-
tion, would permit benefits to some persons to count as
compensation to other persons. Any theory that takes seri-
ously the separateness of persons will be driven away from
such a principle. It is individuals who experience these benefits
and harms. The average or total levels prescribed by the group
principle are statistical artifacts. One must look behind them
to see whether those individuals who now benefit are the same

19. Rawls, Theory of fustJce, p. 27.

20. Nozick, Anarchy, State, and Utopia, pp. 32-33.
21. Ib1d., p. 33.
JU, FAIRNESS, GROUP COMPENSATION, AND EQUALITY

as those individuals who were previously harmed, if one wants

to compensate for past mjustice.
The notion of compensation to the victims of injustice, like
the notion of punishment of the perpetrators of injustice, is
most defensible when it meets an agent-specific requirement:
claims that a given person should be either compensated or
punished for a given injustice require reference to that particu-
lar agent's life history to establish that he-rather than merely
someone similar to him in certain respects-had the alleged
role in the injustice, as either victim or perpetrator. Principles
of compensation, like principles of punishment, become ex-
tremely difficult to defend when they relax this agent-specific
requirement and permit substitution of other members of a
group for persons who played the actual role in the alleged
injustice jwhether that role be victim or perpetrator).
How would we react to the group principle if it were em-
ployed for punishment as well as for compensation? Would it
not be extraordinarily objectionable if a criminal justice sys-
tem were to view blacks, for example, as interchangeable for
purposes of punishment-substituting any black for any other
in the punishment of crimes committed by blacks? The blithe
substitution of one person for another in the "justice" system
that sent many Russians to the Gulag constitutes one of the
most disturbing aspects of Solzhenitsyn's account of the Soviet
prison camp system. 22 Persons are simply not substitutable for
purposes of punishment, even when members of a group are

22. Aleksandr I. Solzhcmtsyn, The Gulag Archipelago: 1918-1956, !New

York: Harper and Row, 197.~ ), chapters l and 3. For example: "By and large, the
Organs had no profound reasons for their chmce of whom to arrest and whom
not to arrest. They merely had over·all assignments, quotas for a specific
number of arrests. These quotas mtght be filled on an orderly basts or wholly
arbitranly. In 1937 a woman came to the reception room of the Novocherkassk
NKVD to ask what she should do about the unfed unweaned mfant of a
nctghbor who had been arrested. They satd "Stt down, we'll find out." She sat
there for two hours-whereupon they took her and tossed her mto a cell. They
had a total plan whtch had to be fulfilled m a hurry, and there was no one
available to send out mto the ctty-and here was thts woman already m theu
hands!" IP Ill.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY I27

interdependent. It is, therefore, difficult to see why they should

be interchangeable for purposes of compensation.
It is possible however, to argue for programs of preferential
treatment for groups, as such, without basing those programs
on a claim of compensation for past injustice. For example, Fiss
offers a second, independent argument based on an ethical
objection against "castes." We are told that "a redistributive
strategy need not rest on this idea of compensation, it need not
be backward looking." The alternative argument runs as fol-
lows:
The redistributive strategy could give expression to an
ethical view against caste, one that would make it undesir-
able for any social group to occupy a position of subordina-
tion for any extended period of time . . . a variety of
justifications can be offered and they need not incorporate
the notion of compensation. Changes in the hierarchical
structure of society-the elimination of caste-might be
justified as a means of (a) preserving social peace; (b)
maintaining the community as a community, that is, as
one cohesive whole; or (c) permitting the fullest develop-
ment of the individual members of the subordinated group
who otherwise might look upon the low status of the
group as placing a ceiling on their aspirations and achieve-
ments.23
I will not quarrel with Fiss's reasons for objecting to "castes."
The instrumental goals just cited are worthy ones, although
their precise connection to the persistence of castes would he
open to debate. Nevertheless, the persistence of anything re-
sembling a caste system seems unjust, in itself, regardless of
how it relates to these other goals. Here, the interesting issue
for our purposes is whether preferential treatment for social

23. F1ss "Groups and the Equal Protection Clause," p. 128. F1ss also makes
an Instrumental argument m terms of the distnbuuon of pohucal power (see
pp. 128-29).
128 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

groups, in Fiss's sense, would constitute a remedy for the

persistence of "castes" jat least to the degree this notion might
apply to the United States! that is either necessary or effective.
Consider, for example, the result of implementing equal life
chances, without any introduction of a preference for groups,
as such. If equal life chances were realized, either through an
equalization of developmental conditions, or through preferen-
tial policies targeted explicitly toward the disadvantaged, then
a persisting racial underclass would be impossible. Over time,
equal life chances would serve to randomly scatter any particu-
lar social group into positions throughout the society. Equal
life chances between individuals would be an extraordinarily
effective remedy for any persisting caste system. The lingering
effects of a caste system could, in other words, be eliminated
without any need to introduce a preference for groups.
Edwin Darn, in his provocative recent book, Rules and
Racial Equality, offers an apparent counterargument to this
latter claim. He defines "equal opportunity" in a sense very
close to the notion of equal life chances emphasized here:
What equal opportunity means is that both A and B have
the same a priori probability of receiving the good !whose
distribution is in question]. Put differently, it means that
certain factors such as race, religion, and sex cannot be
used to bias outcomes.24
From a perspective of "a priori probabilities" in which people
are not differentiated by race, religion, and sex jand, presum-
ably, class! this strict definition should hold: "Equal opportu-
nity means that, given some scarce indivisible good, X, and N
persons wanting X, the probability that any person will get X,
is 1/N." 25
Darn argues that this rule, ambitious as it is, will not
necessarily improve the overall position of blacks, even were it

24. Dorn, Rules and Racwl Equa]Jty, p. 112.

25. Ibtd.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 129

to be fully instituted. Essentially, the argument comes down to

this:
Equal opportunity, if it existed, would mean that for every
substantive gain whites make, blacks could be expected to
make the same gain (relative to population, of course). If
the races are unequal to begin with, that is, at the time
when bias is eliminated, substantive inequality will per-
sist.26
"The reason," Dorn explains, "is obvious: Equal opportunity
applies only to marginal allocation." 27 Equal (per capita) mar-
ginal allocations when added to unequal starting points will
produce unequal(per capita) results. Dorn's definition is not a
"global" one applying to all goods, but a "marginal" one apply-
ing to additional increments.
Hence, the ideal of equal opportunity that has the disturbing
implications identified in his argument is a less demanding
ideal than the one we have identified as equal life chances. The
"unequal starting points" presumed by Dorn's argument
would not be accepted by an adequate ideal of equal opportu-
nity. The transmission of inherited wealth, special advantages
for the privileged, grossly unequal developmental conditions
for the disadvantaged are all varieties of the unequal starting
points that would undermine equal life chances. By contrast,
the more demanding ideal of equal life chances applied globally
regardless of such initial inequalities would not have the
disturbing implication identified by Darn, namely, the perpet-
uation of inequalities between races. As Dorn acknowledges
later in his argument:
Racial equality can be achieved under a rule of global equal
opportunity. By global I mean that the rule applies to all
goods, not just to a portion of future allocations.2s

26. Ibid., p. 120.

27. Ibid., pp. ll?-18.
28. Ib1d., p. 139.
130 FA I R N E S S , G R 0 U P C 0 M P E N S AT I 0 N , A N D E Q U A L I T Y

If goods are an appropriate value for assessing the positions in a

principle of equal life chances, then this "global" principle is
equivalent to equality of life chances. And full implementa-
tion of this principle would, over time, eliminate any system-
atic inequalities between groups.
Were equality of life chances to be fully implemented, it
would offer an adequate response to the caste system problem.
Unlike the less demanding marginal principle considered by
Dorn, equality of life chances would define a version of equal
opportunity that is not subject to the objection that it perpetu-
ates existing inequalities between social groups, such as races
(or ethnic groups).
Yet equality of life chances is an extremely demanding ideal
and, perhaps, it might be argued, a preference for groups is an
appropriate and effective remedy when the ideal of equal life
chances is less than fully implemented. I believe, however,
that a more careful examination of its implications would lead
to the conclusion that it would not be an effective alternative
to equal life chances in eliminating the lingering effects of a
partially operative "caste" system, to use Fiss's term.
Recall Fiss's explicit inclusion of "rich blacks" in any pro-
gram of preferential treatment. So long as the required benefits
improve the position of blacks as a whole, it does not ulti-
mately matter, according to such an aggregative group princi-
ple, how the benefits are distributed within the group. And, as
we have seen, even proponents of such programs expect a
substantial mistargeting of benefits to those who are not them-
selves extremely disadvantaged. In any competitive merito-
cratic context, it is the more advantaged minority group mem-
bers who are likely to have the best qualifications-precisely
because of the unequal developmental conditions that their
more advantaged families are able to provide. Hence any pro-
gram of preferential treatment applied to members of the
group, as such, is likely to result in a substantial mistargeting
of benefits.
While benefits to minority group members who are not,
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 131

themselves, extremely disadvantaged may serve laudable sym-

bolic purposes, they will not be effective in freeing the ghetto
underclass from a persisting caste system. To the extent that
the life chances of the truly disadvantaged are untouched by
such measures, the caste system will persist. On the other
hand, to the extent that preferential treatment is directed, not
to members of the group, as such, but, more specifically, to the
truly disadvantaged, in particular, the caste system will be
broken down. In other words, preferential treatment directed
specifically to disadvantaged individuals so as to improve
equality of life chances would constitute a more appropriate
remedy for a lingering caste system. As noted earlier, such
programs of preferential treatment would differ markedly from
any program of group preference, as such. Such particularized
programs would satisfy the requirement that each individual
benefited, be someone who was, himself, from a disadvantaged
background. In this way, by avoiding the mistargeting objec-
tion, such programs would require that, in exchange for any
sacrifice in merit, there be some improvement in equality of
life chances. These policies are compatible with our basic
framework of individualistic competition. They would not
require any introduction of a preference for a particular social
group, as such.

.. Thus far, I have considered a

4.4 Unequal PoSltwns variety of revisions in our ori-
ginal liberal formula that there should be a fair competition
among individuals for unequal positions in society. Some revi-
sions focus on fairness, others on competition, and others on
the belief that competition should take place among individ-
uals rather than among groups. Another revision in the for-
mula that might avoid our trilemma would be to relax the
assumption of substantial background inequalities, both social
and economic. For then equal life chances would follow, trivi-
ally, from the mere fact of equal outcomes. The expected value
132 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

everyone would receive from the assignment process would be

the same. Presumably, some matching of talents and qualifica-
tions with tasks to be performed still could be devised so as to
conform to merit under such conditions. If so, then both
components of equal opportunity could be implemented with-
out any need to intervene in the family.
Our trilemma clearly could be avoided by such a move, but
only at the cost of an even more radical requirement than the
sacifice in family autonomy. Strict equalization of outcomes,
both social and economic, were it possible at all, could be
maintained in a modern industrialized society only at a sub-
stantial cost in liberty. One of this book's themes is that equal
opportunity, like certain other process equalities at the core of
modern liberalism, can only he achieved at a substantial cost
in liberty. This cost in liberty may arise because of efforts to
insulate social and economic inequalities from compromising
the process equality (in this case, equal opportunity). Interven-
ing intrusively in the family would constitute one form of
insulation. Or, the cost in liberty may arise because of efforts
to eliminate the social and economic inequalities directly. In
either case, insulation or elimination, a substantial cost in
liberty is unavoidable. I will pursue this more general argu-
ment in the next chapter.
It has long been a theme of libertarians that equality of
outcomes could be maintained only at a substantial cost in
liberty. It is more surprising, however, that even the apparently
less drastic process principles of equal opportunity produce the
same conflict. Once the role of the family is accounted for, the
conflict between liberty and equality becomes an unavoidable
problem at the core of liberal theory.
In order to cover all our alternatives, let us now examine
why libertarians have long contended that strict equality could
come only at a substantial cost in liberty. This will close off
the only escape route for proponents of equal opportunity who
might wish to avoid the conflicts with liberty directed at the
family.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY IB

The issue is captured incisively and dramatically by No-

zick's "Wilt Chamberlain" argument. We are asked to pick our
favorite "patterned" principle of justice and to imagine that a
distribution, 0 1, has been realized that perfectly conforms to
it. Wilt Chamberlain, "a great gate attraction," signs a contract
giving him "twenty-five cents from the price of each ticket of
admission." After a million people attend the home games
(each dropping a quarter in boxes marked "Wilt"), Chamber-
lain "winds up with $250,000, a much larger sum than the
average income and larger even than anyone else has." Let us
call this new distribution 0 2• Nozick's central claim is that:
If 0 1 was a just distribution and people voluntarily moved
from it to 0 2, transferring parts of their shares they were
given under 0 1 (what was it for if not to do something
with?), isn't 0 2 also just? If the people were entitled to
dispose of the resources to which they were entitled (under
Od, didn't this include their being entitled to give it to, or
exchange with, Wilt Chamberlain? Can anyone else com-
plain on grounds of justice? Each other person already has
his legitimate share under 0 1• Under 0 1, there is nothing
that anyone else has a claim of justice against. After
someone transfers something to Wilt Chamberlain, third
parties still have their legitimate shares; their shares are
unchanged. 29
Nozick's general point is that any patterned principle, and in
particular, any egalitarian end-state principle, cannot be "con-
tinuously realized without continuous interference with peo-
ple's lives."30 Even a perfect pattern such as 0 1 will be trans-
muted, through people simply exercising the property rights or
entitlements granted to them by the initially perfect distribu-
tion, into some other pattern, 0 2, that may differ markedly
from the original one.

29. Nozick, Anarchy, State. and Utopw. p 161.

30. Ibid., p. 163.
134 FAIRNESS, GROUP COMPENSATION, AND EQUAI.ITY

We need not follow Nozick in his overly strong conclusion

that 0 2 must be as just as 0 11 in order to realize that he is
correct in the basic claim that "liberty upsets patterns."·11
People merely exercising the entitlements granted to them by
some initial distribution will produce varying alternative dis-
tributions. It must be emphasized, however, that restrictions
on this liberty to exercise entitlements may lead to greater
justice according to some structural or patterned principle of
justice.32 Whether these latter conceptions are more or less
adequate than the bare entitlement conception cannot be set-
tled here. 33 For Nozick to reach the conclusion that 0 2 must be
"as just" as 0 11 he would have to assume that the voluntary
exercise of entitlements must be sufficient to settle the ques-
tion of the justice of a distribution. He must, in other words,
simply assume a particular and controversial answer to the
very problem at issue in the example.
While justice, in some sense that takes account of structural
or patterned considerations, may require a sacrifice in the
liberty of people to exercise their entitlements, it is important
to realize that the required sacrifice in liberty is a real one.
Nozick's argument dramatizes how the liberties of property,
contract, and employment (included in his entitlement con-
ception of rights distributed in D.), if left unfettered by other
moral constraints, will overturn initially just patterns, espe-
cially egalitarian ones. As Nozick concludes:
To maintain a pattern one must either continually inter-
fere to stop people from transferring resources as they wish
to, or continually lor periodically! interfere to take from
some persons resources that others for some reason chose
to transfer to them ..J4
31. Ibtd., p. 160.
32. "Patterned" pnnctples mclude all end-state pnnctples (the latter term
apphes to pnnctples such as equahty or uuhtanamsm answenng the structural
problem we dtscussed m secuon 2.2). See Noztck, Anarchy. State. and lltopw.
p. 156.
33. See my cnuque of the bare enutlcment conception m Tyranny and
Legitimacy, chapter 9.
34. Noztck, Anarchy. State. and Utopia, p. Hi~.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 135

This argument applies to "any distributional pattern with an

egalitarian component" and to any other distributional princi-
ples "with sufficient content so as actually to have been
proposed as presenting the central core of distributive jus-
tice."35
While maintenance of any egalitarian distributional pattern
would require continuous coercive interference, it should be
obvious that maintenance of strict equality would require even
more continuous and extreme interference than would most
alternative principles. Any consensual bargain between two or
more persons or groups that yields a departure from equal
shares for anyone would require prohibition or subsequent
redistribution undoing the effects of the bargain or contract.
Systematic suppression of liberties of property, contract, and
employment jso far as the latter involves any reference to
wages or benefits) would be required. The more closely a
principle attempted to maintain strict equality, the more sys-
tematic would this coercive interference and oversight have to
be.
Of course, liberties of property, contract, and employment
sometimes benefit the more advantaged far more than anyone
else. Yet this protection of privilege is not in any way essential
to the argument. One of the more interesting aspects of No-
zick's example is that it dramatizes how these liberties would
have to be suppressed, even if the pattern to be maintained
were one of perfect equality. In fact, the example works most
clearly if perfect equality is the principle under Dli then it is
clear that voluntary transactions between consenting groups
or individuals will serve, over time, to upset the pattern.
Furthermore, none of these voluntary acts need produce any
harm to third parties. Hence, these consensual relations can be
viewed as falling under a negative liberty claim, protected by
the harm principle !see section 2.4). Nozick's theory can be
viewed as protecting "capitalist acts between consenting
adults." Maintaining equality over time would require the
35. Ibid., p. 164.
136 FAIR NESS, G R 0 UP C 0 M PEN SAT I 0 N , A N D E Q U ALl T Y

continuous suppression of such capitalist acts !of exchange,

contract, and employment), as these might be carried out by
consenting adults without harm to third parties. 36 It is in this
sense that realizing equality of outcomes would require a
systematic and substantial sacrifice of liberty.
Of course even the principle of merit might be held to
involve a sacrifice of liberty in some sense. For example,
conservatives have sometimes attacked formal notions of
equal opportunity as an interference with the liberty of poten-
tial employees and employers to make whatever bargains they
like. I do not deny that the equality claim in the principle of
merit Ia procedural claim to equal consideration of one's quali-
fications) involves some sacrifice in a recognizable form of
liberty. Every other form of equality discussed in this book also
involves a sacrifice m liberty.
While the liberty sacrificed by merit takes a recognizable
form, it can be distinguished from the sphere of negative
liberty I have been employing that protects consensual acts
that do not harm third parties. The issue depends upon what
benchmark is accepted for assessing harm to third parties.
It is arguable that if people earn qualifications according to
publicly acknowledged practices of meritocratic assignment,
then they have acquired conditional entitlements; they merit
certain results depending on how their qualifications would be
evaluated in a fair competition. To depart from a fair competi-
tion and discriminate against persons arbitrarily is, in the
context of such meritocratic practices, to impose a harm on
those who have gone to the trouble to earn the qualifications.
It is to impose a harm on those who would have been assigned
36. Noz1ck often rehes on an analogy between consensual, mtimate acts
and consensual property relauons. H1s rhetoncal strategy lS often to extend the
protection hberals customanly prov1de to the former, to the latter as well. See
his comparison of d1stnbuuon problems to the chmce of mates m Anarchy,
State, and Utopw, pp. 150, 167, and 237. I beheve that th1s kmd of extenswn
must face the 1ssue of benchmarks for assessmg harm and the problem of
om1Ss1ons m the causauon of harm. I discuss the former m the paragraphs on
mentocracy m the text followmg th1s citauon. See my discuss1on of the latter
in Tyranny and Legitimacy, chapter 9.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 137

to the positions, had there been a fair competition conducted

according to publicly acknowledged practices. If we accept this
benchmark defined by meritocratic practices for the assess-
ment of harm, then arbitrary or discriminatory employment
contracts departing from merit do harm third parties. In that
case they do not fall within the protected sphere of negative
liberty defined by the harm principle.
The main issue, in this section, however, is whether equality
of outcomes, rather than equality of opportunities, would
require a substantial sacrifice in liberty. Libertarians such as
Hayek and Nozick37 have always argued that this must be the
case. It is worth pausing, however, to consider a novel counter-
argument to this claim, offered by Joseph Carens in a recent
book. 38 Carens constructs a hypothetical scenario for realizing
strict equality of income without, he argues, any substantial
cost in either efficiency or liberty. In an admittedly "utopian"
argument, he bids us to envision an imaginary egalitarian
society that conforms precisely to a private-property-market
(or PPM) system except for these essential modifications:

First, individuals in the second system [his proposal!

believe they have a social duty to earn as much pre-tax net
income as they are capable of earning.
Second, individuals in the second system derive satisfac-
tion from performing this social duty to earn as much pre-
tax income as they can.
Third, individuals in the second system place the same
relative value on the satisfaction derived from performing
their social duty to acquire pre-tax income (hereafter
called social duty satisfaction) as individuals in the first
system (the PPM system) place on the satisfactions de-
rived from acquiring income for consumption !hereafter
called income-consumption satisfaction). 39

37. See chapter I, n. 4.

38. Carens, Equality, Moral IncentJVes, and the Market
39. Ibtd., p. 25.
138 FA I R N E S S , G R 0 U P C 0 M P E N S A TI 0 N , A N D E Q U A LI T Y

Among the additional"logical" prerequisites for the argument,

the most important is the requirement that the tax laws
provide a perfectly equal after-tax income distribution without
any significant loopholes and that these tax laws be generally
obeyed. 411
Carens makes the case, with considerable clarity and detail,
that if individuals were successfully socialized so as to acquire
the preferences specified in the three conditions cited above,
the politico-economic system could be expected to function
successfully without any loss of efficiency, as compared to an
otherwise identical, private-property-market system employ-
ing substantial inequalities as incentives. I will not dispute
the efficiency claim Carens makes for these admittedly uto-
pian, if not implausible, conditions. Despite a few complica-
tions, it is fairly clear that if individuals in a hypothetical,
egalitarian system placed exactly the same value on the "social
duty satisfaction" to maximize pre-tax income that they now
place on "income-consumption satisfactions," then such a
system could operate rather like our present system. 4 I If their
preferences really were structured in that way, then their
efforts and motivations would not be undermined in the least
by the fact that confiscatory marginal income tax rates were
designed to leave them all with precisely the same after-tax
incomes. There would, in other words, be no loss of efficiency
because of lost incentive effects in a society composed largely
of individuals who held such preferences.
More central to our topic here, however, is Carens's claim
that his system could also be realized without any great cost in
liberty compared to the freedom available in a more conven-
tional private-property-market system. This latter claim is
defended in two steps: the first concerns freedom of employ-

40. Sec Carcns's hst of prerequisites, Ibid., pp. 174-76.

41. An issue might, however, he ra1scd about demand. Carens assumes
demand w11l be the same m the two systems, but It will surely be affected by
the much more lim1ted after-tax mwme ava1lable to consumers under h1s
proposal. Sec pp. 24-25.
 FA I R N E s s I G R 0 u p c 0 M p E N sAT I 0 N I A N D E Q u A L I T y 139

ment and consumption; the second concerns the conditions

for preference formation. Let us begin with Carens's claim that
his system would maintain "freedom of choice in occupation"
and "freedom of choice in consumption."42 Under Carens's
egalitarian plan people can enter the job market and pursue
careers in exactly the way they do in a PPM system-with the
one proviso that no matter what they do, they cannot keep any
more or less than the equal income they would have received
anyway. Similarly, they can spend their income on consumer
goods precisely as they do now-with the one proviso that no
matter what they do, they cannot spend more than their
alloted equal income share.
While there is a sense in which freedom of choice of both
occupation and consumption is preserved, we have only to
return to Nozick's Wilt Chamberlain example to see how
freedom of property, contract, and employment are actually
restricted by Carens's system. True, Wilt Chamberlain could
still enter the market for basketball players, he could still
negotiate a hefty contract for pre-tax income (if he happened to
share the preference for maximizing "social duty satisfaction"
in Carens's sense), and lastly, 1,000,000 spectators could still
place their quarters in boxes marked "Wilt." The one difference
is that Chamberlain could not keep any of the proceeds (above
the equal income allotment he would receive anyway, were he
not to play a single game). All one million quarters and any
other income from his hefty contract would have to be given
over to the government. It is undeniable that a central liberty
connected to employment and consumption has been lost: the
liberty to engage in consensual market relations so as to affect
one's material well-being and, in particular, so as to affect
one's after-tax income.
The spectators, presumably, are placing the quarters in boxes
marked "Wilt" in order to give the quarters to Chamberlain,
not in order to give the quarters to the government. Chamber-

42. Ibid., pp. 90-91.

140 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

lain, presumably, negotiates a contract and plays professional

basketball in order to earn a salary for himself, not in order to
add to aggregate government revenues. It is the liberty to
pursue acts that accomplish these objectives-purposes that
impose no harm on third parties-that must be thwarted by
Carens's system. Liberty of employment cannot remain unaf-
fected by programs that destroy the ability of employees to
have any effect whatsoever on their material well-being. The
wage bargain, and its connection to after-tax income, is an
intrinsic part of liberty of employment. Programs that would
completely suppress wage bargains-so far as they might affect
after-tax income in any way-must be considered a drastic
curtailment of liberty of employment.
Of course, income taxation is a familiar part of most private-
property-market systems. But income taxation at any of the
familiar marginal rates raises different issues from those raised
by 100 percent confiscatory taxation for any marginal incre-
ments above strict equality. Such a system would render it
impossible for persons to have any effect on their levels of
material well-being (after-tax income) by engaging in market
relations. To the extent that consensual market relations be-
tween persons or groups do not harm others, they can be
viewed as protected by the sphere of negative liberty discussed
earlier. Because Carens's system must thwart every possible
act of contract, gift, employment, or exchange intended to
affect after-tax income, the proposal would require systematic
coercive interferences banning, to use the phrase employed
earlier, all"capitalist acts between consenting adults" so far as
they are aimed at affecting the income of any of the partici-
pants.
In addition to this sacrifice in the liberties of property,
contract, and employment, another disturbing possibility must
be faced squarely. The only known experiments with "moral
incentives" as replacements for (after-tax) income incentives
have required "preceptoral education"-systematic efforts at
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 141

indoctrination and thought control. 43 Carens attempts to dis-

miss such indoctrination by accepting it as a requirement for
transition but not necessarily as a requirement for mainte-
nance once the egalitarian system was fully in place. He denies
"that the socialization process would have to be consciously
planned and managed by a highly centralized bureaucracy."44
Even though the socialization process that produced uniformly
high preferences for social duty satisfaction would have to be
"more intense" than the socialization process in a PPM sys-
tem:
The point is that it is quite possible to have a highly
intensive socialization process which attaches great social
importance to some particular goal and yet which is de-
centralized, informal, and even, for many of the agents of
socialization, unconscious and unintended. 45
Just as our present emphasis on socializing income incen-
tives is decentralized and not consciously coordinated, Carens
believes that it might be possible for an egalitarian system,
once established, to leave the process of passing on norms of
social duty satisfaction to the same decentralized process.
I believe, however, that there are reasons to be quite skepti-
cal that such a system could maintain itself without continued
and severe restrictions on the process of preference formation.
For Carens's argument to work, social duty satisfaction must
have the same widespread acceptance and relevance for behav-
ior that the desire for income has in more conventional market
systems. Families, schools, peer groups, advertisers, the elec-
tronic and print media, the example of opinion leaders would
all have to be marshalled for-or at least, prevented from
undermining-the required socialization process aimed at pro-
43. See Lmdblom, Politics and Markets, chapters 4 and 21; and Bernardo,
The Theory of MorallncentlVes m Cuba.
44. Carens, Equahty, Moral Incentives, and the Market, p. 119.
45. Ibid.
142 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

clueing persons who believe they have a social duty to earn as

much pre-tax income as possible. It is doubtful whether the
required uniformity and centrality for this particular norma-
tive preference could arise throughout the society without a
campaign of indoctrination and manipulation.
Note, for example, that without restrictions on liberty, the
very market system Carens hopes to maintain with his equal
income proposal would be the source of major disruptive
influences to the required socialization. Corporations in the
hypothesized large-scale market economy would have the
same incentives, that they have now in more conventional
market systems, to advertise and to do whatever they possibly
can to stimulate demand for their products. The pervasive
influence of advertising on preference development should not
be underestimated. 46 Unless there were restrictions on liberty
of expression, advertising, a major and systematic component
of the socialization process, would influence individuals to
maximize consumption of an enormous variety of consumer
goods. These familiar pressures would surely undermine com-
pliance with a social duty norm to consume only a precisely
equal share and hand over the rest to the government. The
already well-known compliance problems for major income
tax systems would be dwarfed by those faced by a confiscatory
100 percent tax-particularly if it were not bolstered by con-
certed efforts at indoctrination.
It is also worth noting that if a system of free expression
were in place, one could reasonably expect a diversity of
values, religions, and creeds to flourish. Indeed, one might
even carry the argument further and claim that without that
diversity, freedom of opinion and freedom of conscience
would, in an important sense, be fundamentally undermined.
Unless people are exposed to a sufficient variety of alterna-
tives, socialization becomes indoctrination and people lose any
autonomy they might otherwise have had in developing their

46. See Lindblom, Pohtics and Markets, chapter 16.

 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 143

opinions. 47 If this extension of the argument were accepted,

then a diversity of creeds, values, religions, and other view-
points on important moral and political questions would,
itself, be a precondition for meaningful freedom of opinion and
of conscience. But such a diversity of creeds, values, religions,
and political opinions would hardly be compatible with
Carens's hypothesis. For his argument would require homoge-
neous, intense, and nearly universal agreement on a particular
and controversial construction of everyone's "social duty." If
other moral, political, and religious opinions were sufficiently
diverse, could one really expect such an extraordinary con-
vergence on a particular normative issue to be maintained
without coordinated efforts at indoctrination? And such cam-
paigns of indoctrination and preference manipulation would,
of course, constitute an important part of the price in liberty
necessary for such a system. 48
Convergence on a particular normative issue, such as the
"social duty" preference required for moral incentives, should
be distinguished from the actual convergence of preferences
found in conventional market systems on the value of after-tax
income. As Rawls notes, income is a "primary good"; it is an
instrumental value that is likely to be of great use in any
market system, whatever one's particular theory of the good
and whatever one's particular normative perspective. 49 While
there are a few notable exceptions, such as asceticism, even
adherents to such positions may find a use for money in
proselytizing or protecting their cherished beliefs. This instru-
mental character of money, regardless of one's particular val-

47. See Mill, On Liberty, chapters 2 and 3. For a recent systematic argument
for a simtlar conclusion, see Ackerman, Socwl Justice, especially chapters 5,
10, and 11.
48. See sectwn 5.1 for the role of th1s sacnfice m hberty m the more general
argument.
49. Rawls, Theory of fustlce, pp. 90-95. Th1s general pomt can be accepted
Without the more particular assumptions about primary goods necessary for
Rawls's argument. See my "Tustice and Rationality" for a cntique of these
latter assumptwns.
144 FAIR :o-; E S S, G R 0 lJ P C 0 M PENS AT I 0 N, AND EQUALITY

ues or creed, thus should be distinguished from the kind of

normative consensus that would be required to systematically
socialize all members of society into agreement on a particular
normative issue with more substantive content, in other
words, that everyone has a "social duty" to maximize pre-tax
income, all additional increments of which ought to be given
to the government.
It might be objected that the need just identified for a sacri-
fice in liberty only arises if the system attempts to maintain
perfect equality without any cost in efficiency. Suppose, how-
ever, that we were to abandon attempts at indoctrination and
accept the resulting sacrifice in efficiency. The costs in liberty
of property, contract, and employment would still arise.
Furthermore, the widespread lack of compliance that would
result from lack of acceptance of the required social duty norm
could be expected to increase the need for coercive interference
to enforce compliance so far as possible. Would people work at
all if they were guaranteed an equal income and were not
indoctrinated to consider their contribution a social duty? In
any case, a great deal of coercion and surveillance would be
required to guarantee compliance with the confiscatory in-
come tax (for all increments above strict equality) and to
prevent black markets, underground economies, barter, and
the other strategies that could be expected to flourish if the
required social duty norms were not widely accepted and
deeply entrenched.
The sacrifice in liberty required by Carens's scenario illus-
trates one side of a more general dilemma. Inequalities can
only be eliminated at a substantial cost in liberty. Although I
have focused on economic inequalities, any thorough effort to
eliminate background inequalities would have to confront
social inequalities as well. It has been speculated that the
division of labor and the enforcement of norms required for a
complex modern society render social inequalities fundamen-
tally ineradicable. I will not enter this controversy here but
will only pause to note that the egalitarian strategy of avoiding
 FAIRNEss I G R 0 up c 0 M pENs AT I 0 N I AN 0 EQuALITy 145

difficulties in equal opportunity through the implementation

of equal results would confront, at some point, the challenge
that anything approaching social equality may not be possible
in a complex industrial society.so I would speculate, however,
that given a substantial sacrifice in liberty, it might prove
possible, no matter what the division of labor. With sufficient
thought control, a strictly egalitarian ethos might be main-
tained, no matter what the differentiation of skills and roles.
Of- course, such thought control would only increase the re-
quired sacrifice in liberty.
Either a substantial sacrifice in liberty or a substantial sacri-
fice in equal opportunity must be accepted. Stated simply, this
is the general problem underlying our discussion of equal
opportunity. In constructing the trilemma, I assumed the ap-
plication of background inequalities; if the door is opened to
efforts at eliminating (and maintaining the elimination of)
these economic and social inequalities, then a sacrifice in
liberty of a different kind results from any serious effort to
avoid the trilemma.
Given background inequalities, we must choose between the
strong doctrine of equal opportunity (merit and equal life
chances) and the autonomy of the family. To sacrifice either
component of the former would seriously compromise equal
opportunity, as we have seen. To sacrifice the latter would
represent a disturbing cost in the sphere of liberty most pre-
cious to the way most of us live. The only remaining alterna-
tive is to relax the assumption of background inequalities with
a resultant cost in liberty of a different kind. We must choose
between equal opportunity, on the one hand, and liberty, on
the other. In the latter case, the liberties are admittedly hetero-
geneous-the private sphere involved in the family can be
distinguished from the liberties of property, contract, and
employment at issue in the equalization strategy. Despite the
different spheres of application, these liberties can all be

50. See chapter 3, nn. II and 12.

146 FAIR NESS, G R 0 UP C 0 M PENS AT I 0 N, A N D E Q U A LIT Y

thought of as applying to consensual acts that do not directly

and significantly harm third parties. In that sense they can all
be subsumed within the broader framework of negative liberty
defined by the harm principle.
The general point is that the conflict between liberty and
equality arises dramatically even when the equality principle
falls short of demanding equal results and requires only equal
opportunities to be unequal. Once the family is taken into
account, as I have tried to do here, the general conflict comes
into clear focus. Equal opportunity, if taken seriously, is a
radical notion with far reaching implications. Even under the
simplifying conditions of ideal theory, its conflict with liberty
is unavoidable. If liberal theory is to offer us a coherent and
defensible ideal, then we must face squarely the conflicts that
were obscured heretofore by our blindness to the role of the
family. If we want more equal opportunity, we must be willing
to undergo some sacrifice in liberty. If we want to preserve
liberty, we must be willing to sacrifice, or significantly com-
promise, equal opportunity. If we blithely advocate the full
realization of both sides in this forced choice, then we are
failing to offer a defensible ideal that might coherently guide
public policy under realistic conditions.

4.5 Policy ImplicatiOns T hisquestion

book has focused on a
of ideal theory.
Given background conditions of inequality what would be the
implications, even under ideal conditions, of attempting to
institute a fair competition among individuals for assignment
to social positions? As I have shown, the implications are quite
radical. We must either sacrifice the autonomy of the family or
we must achieve equality of results. Either of these latter
options would require a substantial sacrifice in liberty. On the
other hand, to avoid the sacrifice in liberty, we must seriously
compromise one or the other component of our doctrine of
equal opportunity. This conjunction of liberal commitments
produces a series of disturbing scenarios even under ideal
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY I47

conditions (that is, even when strict compliance and only

moderate scarcity apply).
However, we do not directly face the problems of ideal
theory in actual public policy. This pattern of conflicting
principles only means that once the role of the family is
realistically taken into account common liberal commitments
do not add up to a unified and coherent ideal.
In actual public policy, however, we are very far from facing
these trade-offs of ideal theory. Each of the principles cited,
while often invoked in policy debates, is only imperfectly
implemented. We do not have now a meritocratic system of
assignment. We do not have now substantial equality of life
chances. And family autonomy is now far more a reality for
middle income and rich families than it is for poor ones. 51 We
could better realize each of these three principles without
producing any further requirement to sacrifice one of the
others. There is much that could be done to root out the
lingering racial and ethnic discrimination that undermines
merit. And there is much that could be done to endow poor
people with the same protection for family autonomy granted
to other strata.
Most dramatically, there is a great deal that might be done to
increase equality of life chances without requiring any sacri-
fice in either merit or family autonomy. We need to continue,
indeed to expand, the Great Society's wave of social experi-
mentation with efforts to improve the developmental condi-
tions of the least advantaged.52 Such efforts may not, as I

51. See Jusuce Blackmun's eloquent dtssent m the recent case Lassuer v.
Dept. of Social Servtces of Durham County, North Carohna 452 U.S. 1811980).
Blackmun defends the right of mdtgent parents to state-paid legal asststance m
child custody cases. Otherwtse, they may be virtually defenseless m confront·
mg efforts by state agencies to take away their children.
52. See Sar A. Levttan and Robert Taggart, III, The Prom1se of Greatness
ICambndge: Harvard UniveiSity Press, 1976); and Edward Ztgler and Jeannette
Valentme, eds., Pro1ect Head Start: A Legacy of the War on Poverty !New
York: Free Press, 1981). See also the home-based strategtes referred to herem
chapter 3, n. 58.
148 FAIRNESS, GROUP COMPENSATION, AND EQUALITY

argued in chapter three, equalize life chances regardless of

social class or family background. Nevertheless, there are
many government interventions, compatible with both family
autonomy and merit in assignment, that may lead to far greater
equality of life chances than has been achieved thus far. Year
after year, black teenage unemployment has hovered around 40
percent in many of our nation's cities. 53 The life chances of a
whole generation have been blighted by inadequate opportuni-
ties to acquire skills and jobs. We need to think more creatively
and to act more resolutely in implementing strategies of inter-
vention directed at the home environment, schools, peer
groups, and all local institutions having a direct impact on
child development. Youth employment opportunities, on-the-
job training, and nontraditional forms of education all need
greatly expanded funding and more creative experimentation.
Even our relatively successful programs have received lack-
luster support. Gerald Gill noted in 1980 about Head Start,
that despite its considerable growth:
A majority of children eligible for participation in Head
Start programs still do not have the opportunity because of
inadequate funds. Because of inadequate funding, not as
many children as are eligible take advantage of Title I
programs. It has been estimated that in New York City
alone there are 151,000 pupils who are eligible for Title I
programs but do not participate. 54
Recent cuts by the Reagan administration in food stamps,
school lunch programs, Medicaid, and a host of other services
are likely to affect directly the physical and economic well-
being of children in poor families.
In all of these cases, there are obvious opportunities to target
policies so as to improve the life chances of children born into
53. See "The American Underclass," T1me llO !August 29, 1977), and
Andrew Brimmer, "Youth m the Labor Market," Black Enterprise !May 1981),
p. 59.
54. Gerald R. Gill, Meanness Mama: The Changed Mood !Washington,
D.C.: Howard Umvers1ty Press, 1980) p. 22.
 FAIRNESS, GROUP COMPENSATIO N, AND EQUALITY 149

disadvantaged families. Such interventions may improve,

rather than undermine, the integrity of those families (without
affecting family autonomy in other strata). Furthermore, such
policies need raise no conflict with the principles of merit
employed for adult assignment. When realization of one of our
core principles can be furthered without any corresponding
sacrifice in one of the others, there is an obvious case for the
effort. The case is more controversial when one principle must
be sacrificed in order to realize another. My point is that the
easy, less controversial case is available to justify a host of
policies that might improve equality of life chances in our
present social context.
The other obvious policy implication is one I have discussed
already in some detail. Just as there is an easy case for improve-
ments in one principle that do not require sacrifices in another,
there is also an easy case against sacrifices in one principle that
do not bring improvements in another. Recall my earlier argu-
ment against policies of preferential treatment merely accord-
ing to group characteristics such as race or ethnic origin.
Unless the particular empirical context makes it reasonable to
suppose that such policies will be targeted primarily towards
persons who are themselves disadvantaged, such policies run
the risk of dramatically sacrificing merit without any corre-
sponding improvement in equality of life chances.
Ironically, while the program that the court refused to rule
on in DeFunis was subject to this obvious objection, the quite
different program of preferential treatment that the court
struck down in Bakke, whatever its other defects, was unusual
in being targeted toward disadvantaged blacks. As Sindler
notes about the special admission program at the University of
California at Davis: "the Davis program, unlike most others,
did not call for giving preference to solidly middle-class minor-
ity applicants over lower-class whites" (or whites of any other
background).ss Each application was searched "for such clues
as whether he has been granted a waiver of the application fee,
55. Sindler, Bakke, DeFunis. and Minority AdmJSSwns, pp. 57-58.
150 FAIRNESS, GROliP COMPE:-.ISATION, AND EQUALITY

which requires a means test, whether he had in the past

participated in programs for the disadvantaged, whether he
worked during school, and the occupational background and
education of his parents."'li While there is some controversy
about how this search was administered, there appears to have
been a clear effort to target preferential treatment toward
persons who were, themselves, disadvantagedY Whatever the
other defects in the Davis program, it is thus arguable that it
fulfilled our minimum conditions. The cost in merit was, at
least, accompanied by a policy that would improve equality of
life chances. From this perspective, the court's solution was
regrettable. For it struck down a program targeted toward
persons who were actually disadvantaged while it encouraged
the explicit consideration of race as a qualification, in itself,
under the banner of "diversity."
As we have seen, such straightforward racial preferences are
subject to the mistargeting objection. For that reason they are
vulnerable to the charge that they sacrifice one of our princi-
ples without producing any gain in the others. Furthermore,
compensatory arguments for past injustice lead to a similar
focus on the disadvantaged (rather than on race per se) when
those compensatory arguments are applied to individuals
rather than to groups. And, as we saw above, it is only the
individualistic version of such arguments that seems defensi-
ble. Thus within the framework developed here, current public
policies are vulnerable to some obvious criticisms, both be-
cause of the severe injustices that are blindly tolerated (the
inequality in life chances and developmental conditions that
might be lessened) and because of the remedies that seem
misdirected (preferences for racial or ethnic groups per se).
Even without an ideal solution to the general theoretical

S6. Bakke v. Regents of Umvcrstty of California, SS3 p.2d 1152, at 1158,

n. f!.
57. There 1s some controversy about the extent to which th1s pohcy was
actually apphcd. Sec the Bncf of the Amcncan Jcw1sh Committee as am1cus
cunac, Bakke. p. 5 I, n. 26.
 FAIRNESS, GROUP COMPENSATION, AND EQUALITY 151

problem we can argue, nevertheless, that there is much that

should be transformed in the current American policies aimed
at equal opportunity.
5. OPTIONS FOR
LIBERAL THEORY

5.1 Contemporary Lib·

eralism: The General Issue T
he ingredients for the tri-
lemma of equal opportunity
are a familiar part of American public ideology and political
theory. It is worth pausing for a moment to consider two
preeminent examples-Lyndon B. Johnson's famous speech at
Howard University in support of his Great Society program
and John Rawls's systematic expression of liberalism in A
Theory of fustice. I pick these two examples from many
possibilities because Johnson's Great Society was the most
important policy initiative in modern American history aimed
at improving the developmental conditions necessary for equal
opportunity. And Rawls's book is clearly the most influential
work of liberal political theory to appear recently in America.
Johnson employed the central metaphor of a footrace, a race
that not only must be procedurally fair, but that also must
realize background fairness in the conditions for acquiring the
abilities measured in the race:
You do not take a person who, for years, has been hobbled
by chains and liberate him, bring him up to the starting
line of a race and then say, "you are free to compete with
all the others," and still justly believe that you have been
completely fair. 1
Applying this metaphor to America's "20 million Negroes,"
the task is to give them:

l. Pubbc Papers of the Pres1dents of the Umted States, Lyndon B. fohnson:

Contaimng the Public Messages, Speeches and Statements of the President
1965 (Washmgton, D.C.: Government Pnnting Office, 1966), p. 636.

152
 OPTIONS FOR LIBERAL THEORY 153

The same chance as every other American to learn and

grow, to work and share in society, to develop their abili-
ties-physical, mental and spiritual, and to pursue their
individual happiness. 2
Jolmson acknowledged that this equalization of developmen-
tal conditions in training for the race, giving blacks "the same
chance ... to develop their abilities/ would require a transfor-
mation of a host of environmental factors. These include the
family:
Ability is not just the product of birth. Ability is stretched
or stunted by the family that you live with, and the
neighborhood you live in-by the school you go to and the
poverty or richness of your surroundings. It is the product
of a hundred unseen forces playing upon the little infant,
the child, and finally the man. 3
And among these forces, the family is crucial: "The family is
the cornerstone of our society. More than any other force it
shapes the attitudes, the hopes, the ambitions and the values of
the child."4
Johnson's historic speech contains all the elements neces-
sary for constructing our trilemma. To maintain what we have
been calling merit is simply to enforce the rules of competition
in the "race." To equalize chances to develop abilities for the
competition would be to equalize life chances for outcomes
determined by the race. But to do so would require manipula-
tion of the "hundred unseen forces," including the family, that
shape the development of abilities.
Of course, Johnson's purpose was not to construct a general
theory of equal opportunity for ideal conditions. Rather, it was
to find a direction for policies that might soon affect the crisis
conditions actually applying. Surely in this latter goal he set a

2. Ih1d.
3. Ih1d.
4. Ih1d., p. 639.
154 0 P T I 0 N S F 0 R LIB E R A L T H E 0 R Y

standard for presidential speeches that has not often been

equaled. And in formulating this direction, surely he was wise
to focus on immediate strategies for leveling up, strategies that
might improve the nutrition, health care, and preschool educa-
tion of the worst-off. The theoretical questions I have raised-
how such strategies must fall short of equalizing developmen-
tal conditions and of giving every American, black or white,
rich or poor, "the same chance as every other American to
learn and grow ... to develop their abilities"-would have
seemed irrelevant philosophical issues at the time. There was
too much immediate progress to be made to worry about how
we might eventually fall short of some ideal.
Yet it is worth emphasizing that this philosophical issue,
central for our purposes, arises simply from any effort to take
seriously common commitments that are an acknowledged
part of American public ideology. We cannot maintain these
simultaneous commitments with any pretense that together
they identify a coherent direction for public policy without
also confronting the hard choices explored in the last two
chapters. Under any realistic construction of the options avail-
able, even under ideal conditions, they cannot all be realized
simultaneously. We must either abandon one or more of these
commitments, or we must modify them by admitting their
susceptibility to sacrifice-by admitting that they can be
traded-off one for another or sacrificed according to certain
priority relations. The important point is that, given any
remotely realistic construction of the empirical possibilities,
to require simultaneous realization of all these commitments
would be to advocate an incoherent ideal for public policy.
Just such a charge plausibly can be directed at John Rawls's
doctrine of fair equality of opportunity, when it is combined
with his comments about the family. Rawls's principle of fair
equality of opportunity includes the principle of merit as we
have been conceiving it, the requirement of what he calls
"careers open to talents." But it goes much farther, approxi-
mating equality of life chances in his formula that: "Positions
are to be not only open in a formal sense, but all should have a
 OPTIONS FOR LIBERAL THEORY 155

fair chance to attain them.'' 5 In part this means that "those

with similar abilities and skills should have similar life
chances."6 The issue turns, of course, on the conditions for
developing abilities and skills. On this issue, he arrives at a
formula for equalizing life chances regardless of income
classes, for giving equal prospects to those who do similarly
well in the lottery of natural assets:
More specifically, assuming that there is a distribution of
natural assets, those who are at the same level of talent
and ability, and have the same willingness to use them,
should have the same prospects of success regardless of
their initial place in the social system, that is, irrespective
of the income class into which they are born/
Rawls's formula yields equality of life chances regardless of
income class provided that there are equal chances in all strata
to develop skills and motivations. Rawls takes this step as a
matter of principle but then shrinks from applying it to the
family:
Chances to acquire cultural knowledge and skills should
not depend upon one's class position, and so the school
system, whether public or private, should be designed to
even out class barriers [emphasis added]. 8
Rawls admits, however, that school systems are likely to be
ineffective by themselves at equalizing developmental condi-
tions, regardless of class barriers. Because of the crucial role of
the family, Rawls therefore concludes that his principle "can
be only imperfectly carried out":
The principle of fair opportunity can be only imperfectly
carried out, at least so long as the institution of the family
exists. The extent to which natural capacities develop and

5. Rawls, Theory of Justice, p. 73.

6. Ibtd.
7. Ibtd.
8. Ibtd.
156 0 PT I 0 N S F0 R li B ER A L T H E0 R Y

reach fruition is affected by all kinds of social conditions

and class attitudes. Even the willingness to make an effort,
to try and so to be deserving in the ordinary sense is itself
dependent upon happy family and social circumstances
[emphasis addedj.9
Here is an explicit acknowledgement of the conflict between
the family as we know it and the equalization of developmen-
tal conditions that would be required for equal life chances
once merit lor "careers open to talents") is assumed. Rawls's
strategy of "adding" to careers open to talents a "fair chance to
attain" the positions allocated by merit leads him down essen-
tially the same road toward equal developmental conditions
that we travelled in chapters two and three above. He stops
short of intruding upon the family, however, and concludes
instead that his principle can be "only imperfectly carried out."
It is unclear why fair equality of opportunity, a principle that
has lexical priority over all other principles except liberty, is to
be left imperfectly realized because of the family. 10 Perhaps
Rawls would accept an argument like the one offered here that
protecting the family should be interpreted as a matter of
liberty, and hence, should fall under a principle of comparable,
if not greater, priority.
As there is no hint of such an argument in the book, there is
no need for us to speculate here on reformulations of Rawls's
theory. We only need note that by permitting the family to
stand as a barrier against the full realization of fair equality of
opportunity, Rawls is endowing the remaining element of our
trilemma with an independent normative status. He is grant-
ing the family sufficient weight to withstand the pressure of
fair equality of opportunity, that is, to prevent its full realiza-
tion.
As noted before, Rawls commits himself clearly to merit
since "careers open to talents" is an explicit part of his fair
opportunity principle. And since he states that "chances to
9. Ib1d., p. 74.
10. Ibid., sections 8 and II.
 OPTIONS FOR LIBERAL THEORY 157

acquire cultural knowledge and skill should not depend upon

one's class position," and that "those with similar abilities and
skills should have similar life chances," something approach-
ing equality of life chances regardless of class position is
clearly his aim.
Yet as Rawls acknowledges, the logic of this position leads
directly to conflicts with the family. It is mystifying why a
principle that must be fully realized before other components
of the theory of justice can come into place, must also remain
forever imperfectly realized "in practice" because of conflicts
with the family. Since fair equality of opportunity is treated
elsewhere as a principle that can and must be fully realized,
one can only conclude that the conflicts embodied in the
trilemma have not been faced squarely. By aspiring to approxi-
mate both equal life chances and merit !with background
inequalities applying), 11 while also granting the normative
appropriateness of leaving the family unaffected, Rawls grants
all the conditions necessary for our trilemma without ventur-
ing to offer a solution. We simply cannot realistically expect to
realize merit, approximate equal life chances and, at the same
time, leave the family unaffected, given the background condi-
tions of inequality that Rawls also grants. By simultaneously
affirming all of these commitments, Rawls has failed to offer
us a coherent scenario for "fair equality of opportunity," even
under the ideal conditions presumed by the argument.
I cite both Rawls's book and Johnson's speech only to show
11. Maximin should not be confused w1th equality. Dependmg on the
operation of mcentive effects, It may greatly increase mequahty, prov1ded that
benefits to the worst-off are accompanied by sufficiently great benefits to other
strata. Wh1le Rawls hopes that the msutuuonalizatwn of fair equahty of
opportunity will, Itself, tend to limit inequahues, th1s empirical tendency (1f 1t
were to apply) would surely not produce equahzat10n to the point of undermm-
ing our assumption that there are background mequahues. For a discussiOn of
the substantial inequahues that Rawls's theory would legltlmate see Brian
Barry, The Liberal Theory of fustlce (Oxford: Clarendon Press, 1973), p. 156. A
crucial factor is Rawls's explicit prov1sion that the liberty pnnc1ple does not
require equalization of income and wealth. One may have equal hberty despite
the unequal "worth" of hberty. See Norman Damels, "Equal L1berty and
Unequal Worth of L1berty," in Damels, ed., Readmg Rawls.
158 OPTIONS FOR LIBERAL THEORY

how the best of recent American public ideology has been

committed to the ingredients required for our trilemma. My
point, of course, is not to discredit liberal theory but only to
emphasize that a more adequate version of liberalism must
face certain hard choices that have, heretofore, been glossed
over. This book is an effort to chart the terrain for such a
reformulation.
I believe, however, that the equal opportunity problem I
have explored here is far from unique. It is but a specific
version of a more general issue facing liberalism. The essential
core of contemporary liberalism can be thought of as a series of
process equalities-institutionalized practices designed to
guarantee equal consideration of everyone's preferences or
interests. The doctrine of equal opportunity explored here is
but one process equality; both in public debate and in theoreti·
cal work, liberals commonly advocate others that are equally
essential. Political equality, equality before the law and, more
controversially, equality of consideration for one's interests by
the health care system offer good examples. I agree with other
recent theorists-Rawls, Dworkin, Walzer, and Ackerman
among others-who identify this notion of equal concern and
respect for everyone as the "nerve" of liberalism. 12 And the
most essential part of this doctrine, in both theory and prac·
tice, has been the notion that this equal concern and respect
should be institutionalized in certain publicly sanctioned
practices that affect especially important areas of life.

12. The origmal positiOn IS, of course, a way of formahzmg equal concern
and respect for everyone. For general comments on hbcrahsm, sec Ronald
Dworkm, "L1bcrahsm" in Stuart Hampshire, cd., Public and Private Mnrahtr
jCambndgc· Cambndgc Umvcmty Press, 1978), pp. 113-43. In particular, see
his comment on p. liS· "a certain conception of cquahty, 1s the nerve of
liberalism." Ackerman's conception of "neutrality" in h1s Social Juwce rests
on a s1m1lar interpretation of liberalism. Also sec Michael Walzer's defense of
cquahty, centrally concerned with "the abohuon of the power of money
outside its sphere." Th1s concern IS "liberal" m the sense developed here (see
my discussiOn of process equalities), although Walzer would term It "demo-
cratic soc1ahst" ja position I would wish to mclude Within liberalism m any
case). Sec Michael Walzer, Radical PrmCiples. Ref]ectwns of an Unrecon-
structed Democrat (New York: Basic Books, 1980), p. 240.
 0 PTI 0 N S FOR LIBERAL THEORY 159

The same fundamental issue we confronted in the trilemma

of equal opportunity recurs for the other process equalities just
mentioned. Because these process equalities cover only certain
specially designated spheres of life, they coexist, in modern
complex industrial societies, with background inequalities,
both social and economic, of the sort we presumed in our equal
opportunity argument. The difficulty is that these background
inequalities spill over, in each case, so as to undermine the
process equality. Just as advantaged strata receive more than
equal developmental conditions thereby undermining equal
opportunity, they also easily obtain more than equal political
influence (both direct and indirect), and more than equal con-
sideration of their interests from the legal and health care
systems. In each case, this spill-over effect undermines a
process equality that has been held to be at the core of a
defensible liberalism.
The parallel can be stated more precisely. Recall that our
trilemma arose from a forced choice among three qualitatively
distinct kinds of considerations: (a) the procedural fairness
embodied in the principle of merit, (b) the background fairness
embodied in the equal life chances principle, and (c) the claim
of liberty defined by the autonomy of families principle. Each
of the other process equalities mentioned is susceptible to a
parallel construction. Each involves a formal competition for
the consideration of claims in which an attempt is made to
insulate the decision-making process from irrelevant factors.
Just as the principle of merit defines one such formal competi-
tion, formal political equality, equality before the law, and the
equal allocation of services and benefits in a nationalized
health care system can all be construed in terms of such formal
competitions-designed to institutionalize equal consider-
ation for everyone's relevant claims. In each case, then, there is
a recognizable issue of procedural fairness.
Once a formal competitiOn is institutionalized in a context
of social and economic inequalities, however, an issue of the
second kind, of background fairness, also arises. The general
point is that persons from the higher strata may have unequal
160 OPTIONS FOR LIBERAL THEORY

opportunities to acquire the characteristics or gather the re-

sources and assistance that make for success in the formal
competition. When persons from the higher strata achieve this
kind of disproportionate success, the second element, the issue
of background fairness, arises.
The third element, liberty, comes into play, because dispro-
portionate success in the formal competition (defined, for
example, by merit, political equality, or equality before the
law) could be prevented by the restriction of certain liberties.
In the political sphere these liberties include free speech and
association, and in the legal and medical care spheres they
include the liberties presently defining lawyer-client and doc-
tor-patient relations. These liberties parallel the liberties at
issue with the family for the equal opportunity case because
they affect the informal acquisition of characteristics or re-
sources that claimants are able to bring to bear upon the more
formal process. Given social and economic inequalities, the
general problem is that procedural and background fairness in
these process equalities can only both be maintained at a cost
in liberty. How the sacrifice in liberty compares to the value of
fully implementing each of these process equalities remains an
open question that must be faced for each process equality in
turn. I have focused on the equal opportunity case, in part,
because the liberty in question is a central one whose implica-
tions for ideal theory have not been explicitly faced. If some of
the other liberties at stake in parallel cases are less important,
then those particular process equalities could be fully realized
at a less substantial cost in liberty.
Nearly an entire book was needed to sort through the com-
plexities in this trilemma in a single issue area, equal opportu-
nity. In this concluding chapter, I can only suggest how parallel
cases might apply to other issue areas. Let us pursue this
comparison in more detail for the central liberal principle of
political equality.
A variety of electoral systems can be devised that give each
citizen's vote equal weight in a formal sense. Proportional
representation offers o~e ideal. One-man, one-vote, in equal
 OPTIONS FOR LIBERAL THEORY 161

population districts offers another familiar if less demanding

ideal-subscribed to for the House, but not for the Senate, in
this country. 1a Yet no matter how perfect the degree of formal
equality, as this might be ascertained from the structure of the
electoral system without any reference to the way persons
with differing preferences are distributed within it, effective
political equality requires something more.
It requires some effort to equalize capacities or opportunities
to influence the informal processes that determine the sub-
stance of politics-the taking of positions, the setting of agen-
das, the gathering of political resources, the formation of
public opinion. All of these informal processes may be subject
to systematic inequalities of influence even when the design of
the electoral system is as close to formal political equality as
might be imagined.t4
Consider Lindblom's "circularity" thesis, that major corpo-
rations, and persons in the higher strata who share their
interests, have a "disproportionate influence" on both deci-
sion-making processes and the formation of public opinion in
American society. Most importantly with respect to public
opinion, Lindblom raises
the ominous specific possibility that popular control in
both market and government is in any case circular. It may
be that people are indoctrinated to demand-to buy and to
vote for-nothing other than what a decision-making elite
is already disposed to grant them. The volitions that are to
guide leaders are formed by the same leaders. 15

13. For a systemauc review of cntena for formal political equahty see
Jonathan Sull, "Pohucal Equahty and Electlon Systems," m Fishkm, ed.,
Theory and PractJce of Representation.
14. Note that this formula does not commit ttself to fully equalizing
influence. A theory that prescnbed equal opportumtJes for mfluence jregard-
less of melevant factors such as mcome or class) would be more defensible. See
my objection below that "those wtth Similar mterests and competence who
differ greatly in their resources will also differ greatly m theu opportumties to
influence each of these processes."
15. Lmdblom, Politics and Markets, p. 202.
H12 OPTIONS FOR LIBERAL THEORY

Needless to say, political equality in any meaningful sense

would be undermined to the extent that such a "circularity"
thesis applied. The degree of the problem in America is a
continuing controversy. 16 Whatever its precise dimensions,
few would deny that there are significant spill-over effects
from socioeconomic inequalities onto each of the informal
political processes we have mentioned-the taking of posi-
tions, the setting of agendas, the gathering of political re-
sources, and the formation of public opinion. What is disturb-
ing in terms of political equality is that socioeconomic
position in general, and money in particular, serve to magnify
the influence of some persons on each of these informal
processes. Citizens do not have equal opportunities to influ-
ence each of these processes because those with similar inter-
ests and competence, who differ greatly in their resources, will
also differ greatly in their opportunities to influence each of
these processes. Some speak with magnified voices because
"money talks"-in campaign contributions to independent
groups, independent expenditures, and ownership of, and influ-
ence on, the mediaY This magnification has an influence not
only on specific decisions but also upon the formation of
public opinion, on the definition of ground rules for public
debate, and on the kinds of questions and positions that can
reasonably expect to get a hearing. All of these informal
processes raise the issue of background fairness no matter how
strictly formal political equality is maintained.
Of course one might attempt, as we have in some post-
Watergate reforms establishing political contribution and ex-
penditure limitations, to insulate some of these political proc-
esses from background inequalities.tM Yet the Supreme Court,

16. For a shnll counterattack on Lindblom's thes1s see Eugene Bardach,

"Pluralism Rev1sed," Commentary 66 (August 1978), pp. 68-70.
17. For a recent rev1ew of the role of money m these processes see M1chael J.
Malbm, ed, Parties. lntere.~t Group.~ and Campaign Fmance Laws (Washmg-
ton, D.C.. Amen can Entcrpnsc lnstttutc, 191!01.
II!. for a comprehensive assessment, sec An Analy.m of the Impact of the
Federal Electwn Campa1gn Act. 1972-78, (Cambndge: The Institute of Poh-
tics, John F. Kennedy School of Government, Harvard Umversity, 1979)
 0 P T I 0 N S F 0 R LIB E R A L T H E 0 R Y 1(,3

in Buckley v. Valeo, in upholding some and overturning some

of these efforts, noted rightly that all of them required, in
varying degrees, sacrifices in liberty. Sacrifices in both liberty
of speech and association would be required by any thorough
effort to insulate the political process from background in-
equalities.
Consider the connection between money and speech. As the
Court emphasized:
Virtually every means of communicating ideas in today's
mass society requires the expenditure of money. The dis-
tribution of the humblest handbill or leaflet entails print-
ing, paper and circulation costs. Speeches and rallies gen-
erally necessitate hiring a hall and publicizing the event.
The electorate's increasing dependence on television, ra-
dio and other mass media for news and information has
made these expensive modes of communication indispen-
sable instruments of effective political speech. 19
Because of the pervasive role of money, restrictions on con-
tributions or expenditures would have the effect of restricting
speech. As the Court noted with respect to expenditures:
Being free to engage in unlimited political expression
subject to a ceiling on expenditures is like being free to
drive an automobile as far and as often as one desires on a
single tank of gasoline. 20
Even the contribution limitations which the Court upheld
were considered an important sacrifice of liberty, one that the
Court argued, however, was outweighed by other even more
important factors. 21
In addition to freedom of expression, freedom of association
is also clearly at stake.
Contribution and expenditure limitations also impinge on
protected associational freedoms. Making a contribution,
19. 424 u.s. 1, 1911976).
20. Ibid., at 19, n. 18.
21. Ibid., at 24--29.
164 OPTIONS FOR LIBERAL THEORY

like joining a political party, serves to affiliate a person

with a candidate. In addition, it enables like-minded per-
sons to pool their resources in furtherance of common
political goals. 22
The McCarthy presidential campaign of 1968 was cited to
show how the freedom of association involved in political
contributions may be a significant one in that it serves to
increase the diversity of candidates and causes in the political
spectrum. As Heard notes more generally of campaign contri-
butions, "the traditional fat cats are not all of one species,
allied against common adversaries. Big givers show up impor-
tantly in both parties and on behalf of many opposing candi-
dates."23 Stewart Matt and Martin Peretz have been as promi-
nent in their support of liberal causes as W. Clement Stone and
Walter Annen berg have been in their support of more conserva-
tive ones. And large initial contributions may be crucial in
launching campaigns that might otherwise die stillborn. Julian
Bond's campaign manager was cited in the Buckley briefs
because he blamed the new contribution restrictions for his
inability to gather enough "seed money" for a presidential
campaign. 24 And the same brief also employed the McCarthy
example:
The 1968 presidential campaign of appellant Eugene Mc-
Carthy is a paradigm case of why contributions to a
candidate are protected by the First Amendment. Individ-
uals who contributed to the McCarthy campaign in order
to further the cause of terminating the war in Southeast
Asia were able to pool their resources in the most effective
way available at that time. If each had been required to act
only as a volunteer, or to contribute only to "issue" organi-

22. Ibtd., at 23.

23. Alexander Heard, The Costs of Democracy !Chapel Htll· Umvers1ty of
North Carolma Press, 1960), p. 6; also ctted m Appellant's Brief, Buckley v.
Valeo, p. 62.
24. Appellant's Bnef, Buckley v. Valeo, pp. 61-62.
 OPTIONS FOR LIBERAL THEORY 165

zations not supporting a candidate ... contributors as a

result would have been far less able to further the cause in
which they believed so strongly. Campaign contributions
were thus a means of exercising the freedom of association
as well as the freedom of expression. 25
Despite these arguments, the Court upheld contribution
limitations, believing that the cost in liberty was worth the
other goals served (greater equality of influence, reduction in
the incidence and appearance of corruption, and so forth). 26 Yet
these examples show that the liberty to make political contri-
butions that was thus restricted may also have an impact on
the variety of causes and candidates supported in the political
system. It is not only X's liberty to join in a cause with Y that
may be restricted; the liberty of another who is unaffiliated,
such as Z, to hear the political messages resulting from such a
campaign (and perhaps to join) may also be affected.
Of course, any effort to insulate public opinion formation
from background inequalities would also have to intrude upon
contributions, expenditures, and acts of expression by other
individuals than those directly involved in campaigns and on
other issues than those directly relevant to particular cam-
paigns. The Court struck down limits on independent expendi-
tures as a violation of the First Amendment. And in First
National Bank v. Belotti, the Court held that in "the realm of
protected speech, the legislature is constitutionally disquali-
fied from dictating the subjects about which persons may
speak and the speakers who may address a public issue." Under
this broad conception, the Court significantly extended First
Amendment protections to corporationsY
Whether or not corporations (rather than individuals within
them) should be given such First Amendment protection, it
should be clear that any thorough attempt to insulate the

25. Ib1d. at 49.

26. 424 US at 24-29.
27. 435 US (1978) at 784-85.
166 OPTIONS FOR LIBERAL THEORY

process of public opinion formation from economic inequali-

ties would require a systematic sacrifice in liberties of speech
and association applying to individuals. These liberties can be
placed within the framework of the harm principle employed
earlier. Persons or groups X and Y who wish to engage volun-
tarily in expression, to combine resources, or to form groups to
promote causes should be free to do so, according to this notion
of liberty, as long as they do not directly and significantly harm
others. Now perhaps these liberties of expression and associa-
tion need to be restricted or controlled in the interests of
political equality but it is important to realize that the sacri-
fice in liberty is a real one. If the influence of economic
inequalities on politicians-an influence that is indirect as
well as direct-were ever to be eliminated, the required coer-
cive restrictions on free expression and association would be
quite enormous. Campaign contributions are only the most
obvious and most direct influence. Even with strict public
financing, one might imagine a Lindblomian process of circu-
larity maintaining itself through unequal influences on opin-
ion formation. But any more thorough effort to eliminate
private money from political expression, broadly construed,
would require a systematic sacrifice of first amendment free-
doms of the sort that has never been seriously contemplated in
this society. Such a thorough going effort would have to
squarely face questions of media ownership, as well. As the
appellant's brief in the Buckley case noted:
It is totally irrational to permit William Loeb, publisher of
the Manchester Union Leader, broad freedom to attack a
candidate while restricting the right of that candidate's
supporters to employ economic resources in defense. The
concern of the court below that the "wealthy few" not
have "a stronger political voice than the unwealthy many
... " is a concern which strikes as directly at the press and
electronic media as at campaign contributions. 28
28. Appellant's Bnef, Buckley v. Valeo, p. 126.
 OPTIONS FOR LIBERAL THEORY 167

Even if campaign contributions and expenditures were

strictly limited, and even if Political Action Committees and
issue advertising were eliminated entirely, unequal influences
resulting from media ownership would have to be confronted.
In a technologically complex society, where significant speech
is seldom defined as the isolated man on a soap box, but rather
as the newspaper article or editorial, or the television program
or advertisement, eliminating the spill-over effects of unequal
resources onto the formation of public opinion would require a
host of restrictions on liberty of expression and association. Far
from advocating such systematic restrictions, I mean to show
that the liberties at stake in this form of insulation are compa-
rable in importance to the liberties at stake in the autonomy of
the family. Realization of political equality in more than a
formal sense, like realization of equal opportunity in more
than a formal sense (as specified by the merit principle), would
require systematic confrontation with the spill-over effects
from background inequalities. These can only be controlled at
some considerable cost in liberty, either from insulation or, as
we saw earlier in section 4.4, from leveling so as to eliminate
the background inequalities directly.
The same analogy can be developed with other process
equalities such as equality of consideration before the law or
from the health care system. 29 In each of these cases spill-over
effects depend on the freedom of persons from the higher strata
to freely contract for superior services with legal or health care
professionals whose superior skills and resources are attracted
by the resulting market incentives. While the Legal Services
Corporation and Medicaid exemplify the recent progress we
have made in leveling up efforts for the poor in both fields, by

29. For an mterpretauon of Amencan constitutional theory m terms of the

nght to "treatment as an equal," see Ronald Dworkm, Takmg Rzghts Serwusly
(Cambndge: Harvard University Press, 1977), espec1ally chapter 12. For a
Rawls1an approach to health care, see Norman Damels, "Health Care Needs
and D1stnbut1ve Justice," Philosophy and Public Affairs 10, no. 2 (Spring
1981): 146-79.
168 0 P T1 0 f\i S F 0 R ll 8 ERA l THE 0 R Y

any realistic appraisal, such efforts could never aspire to full

equalization-unless they were accompanied by systematic
restrictions on the freedom of the non-poor to contract for
better services. Such restrictions have never been palatable;
even the National Health Service in Britain has had to tolerate
coexistence with a system of private clinics for those who wish
to pay for them.
In each of these cases the basic liberal idea of equal consider-
ation can only be imperfectly institutionalized for a given
process equality because of spill-over effects from socioeco-
nomic inequalities. Insulation from these background inequal-
ities could only be accomplished through restrictions on lib-
erty. We must either accept the costs in liberty or accept
compromise in the procedural or background fairness em-
bodied in the process equality. These are the three general
options in the trilemma. The only remaining possibility is the
alternative sacrifice in liberty (discussed in section 4.4) that
would result from leveling, from eliminating background in-
equalities directly. This latter possibility only demonstrates
the truly radical character of each of the core process equalities
when taken seriously.
Could a defensible version of liberalism accept compromise
of equal opportunity, political equality, or equality before the
law as a matter of ideal theory? Even though one's initial
reaction to such a question is likely to be negative, the sacrifice
in liberty in at least some of these cases is equally daunting.
The required restrictions on the family or on freedom of
expression and association (for equal opportunity and political
equality, respectively) look far from inspiring for any adequate
embodiment of the liberal ideal.
To what extent should we expect a solution to these recur-
ring conflicts between liberty and equality? If the notion of
equality were equality of outcomes rather than merely the
process-related equalities explored here, then the problem
posed by our general trilemma would not be surprising. Liber-
tarians have made the conflict between liberty and equality of
 OPTIONS FOR LIBERAL THEORY 16'1

outcomes a commonplace of recent political theory. It is, I

believe, more surprising and disturbing that the apparently
more modest process equalities considered here, if taken seri-
ously, pose conflicts with liberty that are equally stark and
unavoidable. Let us now turn to the kinds of solutions to these
recurring conflicts that might be expected from further efforts
to develop liberal political theory. In particular, let us examine
some recent attempts to perfect the liberal notion of a process
equality-the liberal notion of a procedure granting everyone's
claims equal consideration-in hypothetical versions designed
to choose first principles for organizing society. Perhaps these
hypothetical versions can surmount the difficulties of fully
institutionalizing such process equalities, given background
inequalities. However, while such hypothetical process equali-
ties represent a valuable methodological strategy, I will argue
that they should not be counted on for any new and systematic
solution to the problems discussed here.

5.2 Living WJth

Inconclusiveness W should
hat kinds of solutions
we expect to these
recurring conflicts? This question takes us into a kind of issue
quite different from those we have focused on in most of this
essay. Thus far, we have concentrated on the normative ethics
of social choice-substantive questions of what policies ought
to be chosen by social institutions and political actors. We are
now entering a different area usually distinguished from nor-
mative ethics under the heading of meta-ethics.·~0 What is the
basis for ethical judgments of a given kind and what are their
general defining features? These are meta-ethical questions
that can be discussed independently from the particular sub-
stantive content of the moral judgments under discussion.
30. The d1stmcuon between meta-ethics and normative cth1cs IS a standard
one. See, for example, Richard B. Brandt, Ethical Theory (Englewood Chffs,
N.J .. Prentice-Hall, 1959), pp. 4-10, and William K. Frankena, Ethics. 2nd ed.
(Englewood Cliffs, N.J.: Prenucc-Hall, 1973), pp. 4-5.
170 0 PT I 0 N S f 0 R LI BER A L T H E0 R Y

In this concluding section of a work that is primarily norma-

tive rather than meta-ethical, I cannot deal systematically
with this vast subject. I believe, however, that I can briefly
introduce certain distinctions that will place the rest of the
discussion within a useful framework.
Let me distinguish three meta-ethical claims that are com-
monly made on behalf of moral or political principles:~ 1 I will
then argue that the possible consistent positions on these
claims reduce to four. The resultant scheme of four ethical (or
meta-ethical) positions provides a framework for evaluating
the kinds of answers we might expect to the conflicts explored
here.
The strongest claim commonly made on behalf of a moral
position is:
1. The claim that one's principles are absolute, in other
words, that their inviolable character is rationally un-
questionable.
By "inviolable" I mean principles that it would always be
wrong to violate. 32 One is never morally justified, in other
words, in overruling such absolute principles. In addition to
being inviolable, an absolute principle is also "rationally un-
questionable." By this I mean that it is not open to reasonable
disagreement. Perhaps it is a necessary truth, if such a thing is
possible in ethical matters. Or if not a necessary truth, it has a
kind of apodeictic basis that renders further skepticism inap-

31 Wh1le this scheme will be applied to pohucal pnnc1ples, or ethical

critena for social choice, It can be apphed just as easily to moral pnnciples for
mdividual choice .
•~2. V10latmg such a strong pnnc1ple, even when It conflicts With another, IS
morally prohibited. Conflicts among pnnc1ples of th1s kmd may, of course,
create blind-alley SituatiOns, where nothmg one could do would be nght or
morally acceptable. Such Situations define a kind of tragedy. See Bernard
Williams, Problems of the Self (Cambndge: Cambndge Umvers~ty Press,
1973), especially pp. 172-74 for an account of cases of extreme moral confhct
m wh1ch the requirements of the overridden pnnc1ples must sull be faced m
some way. Even 1f one cannot follow a pnnc1ple, because Jt conflicts with
another, one ought to make amends, expenence regret, and so forth.
 OPTIONS FOR LIBERAL THEORY 171

propriate. Of course, any particular position that laid claim to

this characteristic would have to include a further account of
the meaning of "rational" or "reasonable" and a further ac-
count of the basis for the principle's immunity from reasonable
questioning. But these details of particular positions need not
concern us at the moment.
A second, less demanding claim would weaken the absolut-
ist character of the basis attributed to one's principles. I will
identify it as:
2. The claim that one's principles are inviolable, or that it
would be objectively wrong ever to violate (permit excep-
tions to) them.
According to the second claim, one may attribute an "objec-
tive" basis to principles that hold without exception. This
objective basis, that we still need to clarify, however, falls
short of the first absolutist claim of being rationally unques-
tionable.
A third, even less demanding claim may be identified as:
3. The claim that one's principles are objectively valid, in
other words, that they are supported by considerations
that anyone should accept, were they to view the problem
from what is contended to be the appropriate moral
perspective.
This claim no longer includes the inviolability requirement.
Hence, such principles do not purport to be immune from
justified exceptions or overriding. For example, in cases of
conflict between principles of this kind, it may be right to
override one principle for the sake of fulfilling another. 33

33. Weak or pnma facte prmctples may justifiably be overnden in cases of

conflict. A conception of an action's bemg right "on balance" can be employed
for such conflicts. One might sull be required to make amends, to the extent
possible, for the overnden principles, but pnnctples of thts kmd would not
define tragic or bhnd-alley sttuauons (m whtch nothmg one could do would be
right).
172 OPTIONS FOR LIBERAL THEORY

But such principles, even though not inviolable, fall under a

claim of objective validity las do principles conforming to the
second claim). The proponent of such a principle claims that
from the appropriate moral perspective, one that he believes
should have jurisdiction over anyone else's choice of principle
lone that, in other words, he believes to be valid for anyone),
there is sufficient support for the adoption of his principle.-14
This notion of the "appropriate moral perspective" may be
formalized in a moral decision procedure such as Rawls's
original position or the perfectly sympathetic spectator of the
utilitarians. I shall explore such decision procedures below. Or
it may be simply the informal appeal to impartiality familiar
from the Golden Rule or from appeals that one should look at a
situation from the perspective of the others affected. As
Thomas Nagel characterizes it, "the general form of moral
reasoning is to put yourself in other people's shoes."-15 However
formal or informal such notions of the moral point of view, the
idea of such an appropriate moral perspective provides the
basis for a claim that a given principle has a kind of objective
validity. Of course, proponents of different principles may have
quite different notions of the appropriate moral perspective
from which principles ought to be derived or chosen; but that
is only another way of saying that the objective validity claim
13, above) is far weaker than the absolutist claim (1, above). It is
weaker in that there is no claim that the basis for the principle
is beyond reasonable question. Unlike the first position, no
immunity from rational disagreement is claimed for the cru-
cial assumptions, the "appropriate moral perspective" from
which the principle can be supported:16
As defined, the properties mentioned in these claims stand
in certain logical relations. An "absolute" principle must also
34. Even though parucular accounts of the appropnate moral perspective
may lead uniquely to some particular principle, the claim of any such principle
is mconclus1ve because of the }unsd1ction and foreseeability problems investi-
gated below
35. Thomas Nagel, Mortal Questwns ICambndge: Cambridge University
Press, 1979), p. 126.
36. See the discussiOn of reflective eqmhbrium below.
 OPTIONS FOR LIBERAL THEORY 173

be "inviolable" and "objectively valid." In other words, a prin-

ciple satisfying claim 1 must also satisfy claims 2 and 3.
Similarly, a principle satisfying claim 2 must also satisfy claim
3. The idea is that a principle satisfying any of these claims
must also satisfy those following it, but not those preceding it.
On the other hand, if a principle fails to satisfy a given claim,
it must also fail to satisfy those that precede it, but not those
following it. If a principle is not objectively valid, then it
cannot be either inviolable or absolute in our sense. Or if it is
not inviolable, it cannot be absolute. Failure to satisfy claim 3
implies failure to satisfy claims 1 and 2. Failure to satisfy 2
implies failure to satisfy 1. In general, satisfaction of a claim
implies satisfaction of those following it and nonsatisfaction of
a claim implies nonsatisfaction of those preceding it.
Either of these logical patterns would reduce the consistent
possibilities for combining these claims to the four positions
depicted below:J7
II III IV

Absolutism R1gonsm Intultlomsm Subjectivism

1. Absolutist
claim +
2. Inviolability
claim + +
3. Objective va-
lidity claim + + +

37. Consider the first of these logical patterns for consistent positions-that
conformity to a given claim requues conformity to those followmg it. Let us
number the three cla1ms, C 1, C2, C3.
a. cl is either + or -. If cl IS +I then by the hypothesized pattern, c2 and c3
must also be +. Hence, the pattern + + + (position I).
b. If, on the other hand, cl is -I then c2 may be either + or -.If c2 IS +I then
by the hypothesized pattern, C3 must also be +. Hence, the pattern - + +
(positiOn II).
c. If, on the other hand, c2 is -I then c3 may be either + or -. If c3 IS +I then
the resultant pattern is - - + (posltlon III).
d. If, On the other hand, C3 IS - 1 then the result IS the pattern - - - (pOSltlOn
IV).
174 OPTIONS FOR LIBERAL THEORY

Accepting all three claims produces a position we can call

"Absolutism" (position I)-an assertion of rationally unques-
tionable principles that hold without exceptions or overrid-
ings. Kant offers a good example of such a demanding posi-
tion:18 Rejecting claim 1 but accepting the remaining two
produces a position I call"Rigorism" (position II) because of its
claim to apply objective principles rigorously, or without any
exceptions or overridings. Rawls in the modern era and Ben-
tham in an earlier one offer good examples of such a position
applying objective principles rigorously.3 9 If we reject both

Of course, this scheme of four positions emphasizes certam distmctions to

the exclusion of others. Consider, for example, a position laymg claim to
rationally unquestionable conclusions about pnma facie pnnciples. Such a
posmon does not fall under enher absolutism or ngorism because Its pnnc1ples
are not mvwlable. It would have to be classified at position III, intuitiomsm
While this would be appropnate m some ways, the pos1t1on asserts more than
the defimng claims of posltlon III. Instead of principles that are merely
objectively vahd, 1t asserts prmciples that are rationally unquestionable. In
other words, whtle these classtflcauons exhaust the conststent possible post-
ttnns on the claims defined here, they also may conf]ate under one classifica-
tion positions that we might wtsh to dtstmgutsh, at least for some purposes.
These conflatlon problems do not affect the use I w1ll put the scheme to here .
.~8. Our "synthetic a prwr,-· knowledge of morality IS "categoncal"; it is
"uncondttloned" and "apodetctlc." Its pnnciples hold "let the consequences be
what they may" See Immanuel Kant, Groundwork of the Metaphys1c of
Morals. trans. H. J. Paton (New York: Harper and Row, 1964), pp. 79-84. As he
explams the exceptlonless character of a prion moral judgments elsewhere,
they hold "wtth strict universality, that 1s, m such a manner that no exception
is allowed as possible," Crlllque of Pure Rea.~on, trans. Norman Kemp Smith
(London: Macmillan, 1929), p. 44 .
.~9. I take the term "ngonsm" from Marcus Singer who means by it "that
certain moral values hold absolutely or m all circumstances." Marcus George
Singer, Generalzzatwn m Eth1cs. An Es.~ay on the Logic of Eth1cs (New York:
Atheneum, 1971), p. 228. As for the classification of Rawls and Bentham at this
posmon, sec Rawls, Theory of fustice, pp. 21 and 51 for h1s rejection of an
absolutist basis. For his aspuauon to satisfy the mv10lability claim see his
discussion of the prionty problem in section 8. Jeremy Bentham's utilitanan-
ism displays comparable smgle mmdedness; see h1s An Introductwn to the
Prmc1ple.~ of Morals and Leg1.~latwn. (Oxford: Clarendon Press, 1907; first
pubhshcd 1789). For a new mterprctation cla1mmg that Bentham embraced
different standards for soc1al and individual choice, see Dav1d Lyons, In the
Interests of the Governed· A Study in Bentham's Phllosophy of Uulzty and
Law (Oxford: Clarendon Press, 1973). Classifying h1m at position II for social
chmce, w1ll be sufficient for my purposes.
 OPTIONS FOR LIBERAL THEORY 175

claims 1 and 2 but accept claim 3 we find ourselves with one of

the positions Rawls is most concerned to argue against, "Intui-
tionism" (position III). Lacking a single inviolable principle (or
group of inviolable principles in lexical order), one may, never-
theless, lay claim to objective principles that are weak or prima
facie, that is, principles that are capable of being overridden or
traded-off, one for another. This position is often called Intui-
tionism because it requires a careful weighing of moral factors
in each particular case. 40 Isaiah Berlin's advocacy of a "plural-
ity" of irreconcilably conflicting ultimate principles offers a
good recent example of this position. 41 Lastly, if we reject all
three claims, we find ourselves in the category of "subjectiv-
ism" (position IV). The subjectivist does not assert objective
principles in any sense. He makes no claim that his principles
are supportable from some appropriately impartial moral per-
spective. Typically, he views his principles, if he has any, as
arbitrary personal tastes or private whims that are not suscep-
tible to justification. Since I have defined this category as a
residual one, I might have defined further subdivisions within
it. 42 However, these will not be necessary for our purposes
here. Edward Westermarck in his Ethical Relativity and Sartre
in his early existentialist period offer good examples of this
position. 43
40. Sec Rawls, Theory of {us lice, pp. 34·40.
41. Berhn, Four Essays, pp. 167-72. Th1s plurahsm IS really at the center of
Berlin's political theory. See Bernard Williams's "Introduction" to Isaiah
Berlm, Concepts and Categories (London· The Hogarth Press, 1978).
42. Subjective umversalists might be d1stmguished from relativists who
judge others accordmg to values that they accept (or that the relevant groups
accept). The amorahst who refrains from all moral judgments defines another
subdiviston. Smce subjectivism (position IV) was defined by the rejectton of all
three claims, any of these pos1t10ns would be compatible with the classifica-
tion defined th1s broadly.
43. See Edward Westermarck, Ethical Relativity (1932; repnnt ed., West-
port, Conn .. Greenwood Press, 1970) Sartre's subjectivism is clearly expressed
in "Existenuahsm 1s a Humanism," m Walter Kaufman, ed., Ex1stentwhsm:
From Dostoevsky to Sartre (New York: Meridan, 1956). A more recent argu-
ment for one variant of subjectivism can be found in G1lbert Harman, "Moral
Relativ1sm Defended," The Philosophical Rev1ew 74(1975) and The Nat me of
Morality (New York Oxford University Press, 1977).
176 OPTIONS FOR LIBERAL THEORY

What are the methodological resources within liberal theory

that might provide a solution in any of these senses (short of
position IV) to the moral conflicts explor;!d here? A direct
appeal to intuition or conscientious moral opinion cannot be
expected to resolve the conflicts in the trilemma of equal
opportunity, or, indeed, those in the more general trilemma
applying to other process equalities. The ingredients for these
trilemmas all have a prominent place in common moral opin-
ion and in public ideology. A direct appeal to those opinions
provided the basis for my analysis. Some other methodological
strategy would be required to resolve or overcome the norma-
tive conflicts revealed here.
In an earlier era, there may have been sufficient consensus
on ultimate religious and metaphysical matters to support an
absolutist public ideology. For example, the appeals to "self-
evident truths" in the Declaration of Independence were not
merely a rhetorical flourish; they depended crucially on reli-
gious assumptions ("Men are endowed by their creator ... ").
Morton White explains Jefferson's conception of "moral laws
of nature":
They were thought to be decreed by God; they were
regarded as precepts for the direction of the voluntary acts
of reasonable agents; and some of them were thought to be
discoverable by intuitive reason. 44
Hamilton was even more explicit. He described natural rights
as "written, as with a sun beam ... by the hand of the divinity
itself; and [theyJ can never be erased or obscured by mortal
power."45
But the operation of religious and ethical diversity in a
pluralistic and liberal society has long since undermined the
religious and metaphysical consensus required to support such

44. Morton White, The Philosophy of the Amen can Revoluuon !New York:
Oxford, 1978), p. 157.
45. Ibid., p. 80.
 OPTIONS FOR LIBERAL THEORY 177

eighteenth century appeals to absolutism. Hence that consen-

sus is no longer available as an unquestionable moral basis for
a public ideology. It is also arguable that a truly liberal political
theory could never rest upon such an explicitly religious foun-
dation because some degree of neutrality by the state between
such doctrines would be necessary for freedom of religion,
presumably an important component in any recognizably lib-
eral theory. 46
At any rate, the direct appeal to conscientious moral opinion
cannot be expected now to resolve our trilemma. Let us turn to
the distinctive methodological innovation of recent liberal
theory-the development of moral decision procedures for the
choice of political principles. Such decision procedures can be
viewed as a further development of process equalities that
embody equal concern and respect, but in a perfected, hypo-
thetical form in which all possible biasing contingencies are
eliminated. Instead of merely focusing on the institutionaliza-
tion of actual practices conforming to this idea of equal con-
cern and respect in particular areas of life, the new strategy in
recent liberal theory has been to construct a hypothetical arena
of decision embodying the moral point of view where every-
one's claims can be more perfectly adjudicated and where a
uniquely adequate solution to problems of distributive justice
can be reached.
This notion of a hypothetical decision procedure is what
unites Rawls's theory of justice with modern utilitarianism, as
developed by such theorists as Harsanyi and Singer, 47 and with

46. For a limned account of neutrality with respect to religiOn, see Phihp B.
Kurland, Rellgwn and the Law. Of Church and State and the Supreme Court
!Chicago: Aldme, 1961). For a more ambitious account of neutrality with
respect to all theories of the good, mcluding rehg10n, see Ackerman, Soc10l
fustice m the Liberal State. For a cntique of Ackerman's central argument, see
my forthcommg article, "Can there be a Neutral Theory of Justice?" Ethics
47. See John C. Harsanyi, Essays on Ethics, Social Behavwr and Scientific
Explanation lDordrecht, Holland: D Reidel, 1976), part A; and Peter Smger,
Practical Ethics lNew York: Cambndge Umvemty Press, 1979), especially
chapter 2.
178 OPTIONS FOR LIBERAL THEORY

a modern utopian-egalitarian such as Bruce Ackerman (in his

innovative book Social Justice in the Liberal State). 48 The
general strategy can be identified as the construction of a
hypothetical arena of decision for the choice of principles of
distributive justice according to an impartial consideration of
everyone's interests. This strategy offers the possibility of
systematic theory that might move beyond the analysis of
intuitions or commonly shared assumptions. Should we expect
the development of these hypothetical decision procedures to
produce a breakthrough resolving the conflicts in our tri-
lemma? Can they be expected to settle the problem of distribu-
tive justice, at least within the confines of ideal theory?
However productive such decision procedures may be for
some purposes, there are some fundamental reasons to expect
their results to be inconclusive. We should not expect position
I resolutions from them. But this should not be a cause for
despair if the range of alternatives has been correctly identified
here. Lacking position I solutions, we can aspire to the more
limited claims of positions II and III without being forced into
the arbitrary Subjectivism of position IV. I will focus on two
sources of inconclusiveness that can be labeled the jurisdiction
problem and the foreseeability problem. Afterward, I will
return to the scheme of ethical positions just outlined.
The jurisdiction problem becomes apparent from the ques-
tion: why should our ethical disputes be resolved by one
particular hypothetical arena of decision rather than compara-
ble others that can also claim to consider everyone's interests
impartially?
Consider some of the possibilities. Rawls proposes an "ori-
ginal position" from which each of us can imagine ourselves
choosing principles of justice out of rational self-interest while
knowing nothing in particular about ourselves or about our
place in society. A "veil of ignorance" shields each of us from

48. Ackerman, Socwltusuce m the Liberal State (New Haven: Yale Umver-
sity Press, 1980). See the dtscusslOn of ncutrahty below.
 0 PT I 0 N S F 0 R L I B ER A L T H E0 R Y 179

the knowledge that might permit us to bias the construction of

principles toward our own particular self-interest. Rawls be-
lieves that this construction of the fair hypothetical choice
situation, when combined with certain other assumptions,
provides an argument for his particular proposed principles of
justice. I won't enter the controversy here as to whether or not
these assumptions actually yield Rawls's proposed principles. 49
I will only note that Rawls's argument, like other moral
decision procedures in liberal theory, depends on both li) a
particular interpretation of impartiality (the veil of ignorance
and other related conditions of the original position), and (ii) a
particular interpretation of everyone's interests (Rawls's the-
ory of "primary goods" that, he contends, it is in the interests
of everyone to want regardless of whatever else they want).
The important point is that even if one were committed to
the basic strategy of choosing principles of justice from a
hypothetical arena designed to perfectly consider everyone's
interests impartially, one would have to resolve the problem of
why one particular interpretation of impartiality rather than
another, and why one particular interpretation of interests
rather than another, should be adopted for the design of the
hypothetical decision procedure. Consider a slight variation.
Rawls admits that a minor adjustment in his own proposed
original position-an adjustment permitting probabilistic cal-
culations-would yield the rival principle he is most deter-
mined to argue against, the principle of average utility. If one is
permitted to assume that one has an equal chance of turning
out to be anyone, then one's expected value of payoffs in life
would be maximized by a society maximizing average utility. 5°
This can be considered a modification in the interpretation of
49. See my "Just1ce and Rauonahty" for an extended d1scusswn of th1s
issue. In Rawls's most recent work, there lS an acknowledgment of the morally
controvers1al character of the argument from the origmal position in the claim
that lt depends upon a particular "model concepuon of a moral person." See
John Rawls, "Kanuan Construct1v1sm in Moral Theory: The Dewey Lectures
1980," The Tournai of Philosophy 77 no. 9 (September 1980): 515-72.
50. Rawls, Theory of fustice, p. 165.
1110 OPTIONS FOR LIBERAL THEORY

impartiality (interpreted now so as to require an equal proba-

bility of being anyone) and a modification in the interpretation
of interests (interpreted now to mean satisfaction of prefer·
ences). Whether equal probabilities rather than mere uncer-
tainty (as in Rawls's proposal) is the more appropriate interpre·
tation of impartiality or whether utility rather than primary
goods is the more appropriate interpretation of interests, are
not issues that can be decided by the decision procedure itself.
Rather, construction of the decision procedure requires some
independent basis for these decisions.
Ackerman provides us with still another interpretation of
impartiality and still another interpretation of interests. His
hypothetical choice situation is a dialogue for occupants of a
spaceship about to start life anew on an uninhabited planet.
They must determine the appropriate distribution of "manna,"
a substance that is "infinitely divisible and malleable, capable
of transformation into any physical object a person may de-
sire."'1 But the crucial point about Ackerman's mental experi-
ment is that the occupants must justify a particular distribu-
tion of manna by providing an argument that passes certain
constraints, the key one being an account of moral impartiality
that Ackerman dubs "neutrality," as follows:
Neutrality. No reason is a good reason if it requires the
power holder to assert:
(a) that his conception of the good is better than that
asserted by any of his fellow citizens, or
(b) that regardless of his conception of the good, he is
intrinsically superior to one or more of his fellow
citizcns.' 2
Participants in this hypothetical dialogue soon discover that
the only argument that passes this neutrality constraint can be
summed up as "I'm at least as good as you are, therefore I

51. Ackerman, Sncwl/u.~llce. p..~I.

52. lb1d., p. 11.
 OPTIONS FOR LIRERAL THEORY IRI

should get at least as much."'" And this argument is employed

to justify equal shares of manna on arrival at the planet. Then,
in a more complex consideration of the distribution problem
over time, in real societies lacking the imaginary substance
manna, Ackerman discusses "second" and "third best" approx-
imations to the hypothetical ideal of the spaceship dialogue. In
this way he is led to prescriptions for education, genetics, and
initial material endowments that are analogous to the equal
manna solution. The results are ingenious in their mainte-
nance of diversity amidst a basic framework of strict initial
equality. While the argument, like Rawls's, has proven contro-
versial, the central point is that when a quite different but
appealing conception of impartiality (neutrality) is combined
with a quite different but appealing conception of individual
interests (receiving as much manna as possible), a radically
different principle of justice emerges.
A fourth hypothetical decision procedure offering a familiar
account of how everyone's interests might be considered
equally is the perfectly sympathetic and impartial "spectator"
of the classical utilitarians. Going back to Adam Smith, this
idea would have us imagine a spectator who reproduced in
himself every pain and every pleasure in the world. Perfectly
reproducing in himself every sensation of disutility and of
utility experienced by anyone, he would prefer those states of
the world that maximized the over-all balance of pleasure over
pain. In this procedure, another account of impartiality (per-
fectly sympathizing with everyone's pain and pleasure in the
same way) is combined with an account of everyone's interests
(utility) so as to yield the familiar principle of aggregate util-
ity. 54

53. Ibid., section 14.

54 Rawls, Theory of fustlce. pp. 183-92. Sec also Adam Smith, The Theory
of Moral Sentiments (Indianapolis· Liberty Classics, 1969), pp. 22, 31, 33, 35,
36, 38, 41, 71, 161-62, 211, 228, 247-49, 352, 371, 422; and Rodcnck Firth,
"Ethical Absolutism and the Ideal Observer," Philosophy and PhenomenologJ·
cal Research 12, no. 3 (March 1952): 317-45.
182 OPTIONS FOR LIBERAL THEORY

These are not, of course, the only attempts to construct

moral decision procedures that might plausibly claim to con-
sider everyone's interests impartially. Elsewhere, I have argued
that procedures for cutting a cake fairly provide an analogy that
can be developed into a moral decision procedure for distribu-
tive justice in a society. The intuitive idea is that just as the
cake-cutter ought to determine the slices without knowing
which piece will be left for him (after the others are permitted
to choose theirs), a person might be imagined to determine
social allocations without knowing which ones will be left for
him after the others choose theirs. While this procedure would
yield maximin Ithe principle that the minimum share should
be maximized) under certain assumptions about the "cake,"
under other plausible assumptions about the interests at stake
it can be interpreted to yield its own distinctive principle.''
Here are five distinct moral decision procedures: Rawls's
original position, the average utility version of the original
position, Ackerman's theory of neutral dialogue, the perfectly
sympathetic spectator of the classical utilitarians, and my
adaptation of the cake-cutter's procedure. Each of them em-
bodies a distinct interpretation of what it might mean to
consider everyone's interests equally or impartially; each of
them rests on a distinctive interpretation of the interests to be
assessed. Even slight modifications in either of these basic
components can produce extraordinary differences in the re-
sulting principle-as the debate between Rawls and the propo-
nents of average utility attests.
Of course, these five procedures do not begin to exhaust the
field. There is no limit in principle to the proposals that might
be developed. And there are certainly other interesting ones
that have been proposed.s6 Not knowing how many appealing
55. See my Tyranny and Legitimacy, chapter 12.
56. See Douglas Rae's "Court of Allocatmn" m "Maximm Justice and an
Alternative Pnnc1ple of General Advantage," American PolztJcal Science
Review, 69, no. 2 lJune 1975): 630-47. See also Thomas Nagel, The Possibility
of Altruism (Oxford: Oxford University Press, 1970), p. 141 for a proposal in
which the chooser IS to imagme himself hvmg all the lives affected, seriatim.
 OPTIONS FOR LIBERAL THEORY 183

accounts of impartiality might be presented, nor how they

relate to the many controversial conceptions of individual
interests (see section 2.1 above), we are forced to compare
decision procedures informally. Even if we were absolutely
sure of our judgment about each procedure in turn, we could
never know whether, after committing ourselves to a particu-
lar procedure and its resulting principle, some better procedure
with different results might not be devised. I will return to this
issue below when I discuss the foreseeability problem.
The jurisdiction problem yields inconclusiveness in the fol-
lowing way. Any proponent of any of the five principles just
mentioned, or any proponent of any other principle who bases
his argument on this kind of decision procedure, is open to
challenge at the point of commitment to his proposed account
of impartiality and of his proposed account of interests. Those
who would champion different principles can always challenge
whether a given procedure should have jurisdiction or whether
another is more appropriate. And because each procedure can
be connected to a different substantive principle (about which
proponents in an ethical dispute will have differing convic-
tions), adoption of one particular procedure can always be
challenged on the ground that it is biased toward X's convic-
tions rather than Y's. The result is always open to reasonable
disagreement because good faith disputes over substantive
principles can always find their parallel in good faith disputes
over moral decision procedures, in other words, over the appro-
priate reasonable basis for resolving disputes. The issue of
which procedure to adopt cannot be settled by the procedure
itself. And given the variety of procedures, each one supporting
a different substantive outcome, the mere invocation of a
moral decision procedure supporting one particular proposal is
not enough to settle a moral disagreement.

Yet another proposal is developed by Lawrence Kohlberg in h1s "Justice as

Reversibility" m Laslett and F1shkm, eds., Ph1losophy, PobtJcs and Society,
F1fth Series.
184 0 P T l 0 N S F 0 R Ll BE R A L THE 0 R Y

Just as different arbitration panels can be expected to pro-

duce different results in, say, a labor management dispute,
different moral decision procedures clearly yield different prin-
ciples of justice. But in a labor management dispute, the
jurisdiction problem can be solved either by mutual consent of
the parties involved or by recourse to the mutually acknowl-
edged authority of a court order. But for our moral decision
procedures there is no basis for actual consent, and no mutu-
ally acknowledged source of authority, tying us to one particu-
lar moral decision procedure rather than another. There are
only further moral arguments about what we ought to agree to,
or about what our actual notions of morality would commit us
to, if we thought about them as the theorist advocates.
Since Rawls has addressed this issue more explicitly than
the others, I will pause for a moment to consider his response.
There is a revealing ambiguity in his account of why we should
accept his particular conditions proposed in the original posi-
tion. In the last paragraph in the book, he says that "we do in
fact accept" the conditions, or that if we do not, these are
conditions that we still "can be persuaded to" accept:
Finally, we may remind ourselves that the hypothetical
nature of the original position invites the question: why
should we take any interest in it, moral or otherwise?
Recall the answer: the conditions embodied in the descrip-
tion of this situation are ones that we do in fact accept. Or
if we do not, then we can be persuaded to do so by
philosophical considerations of the sort occasionally in-
troduced. 57
And if we are not or cannot actually be "persuaded," what is
the basis for binding us to the results of the procedure? Rawls
has an easy answer for the "egoist" who does not commit
himself to the procedure since we knew all along that egoism
"is incompatible with what we intuitively regard as the moral

57. Rawls, Theory of fusllce, p. 587.

 OPTIONS FOR LIBERAL THEORY 185

point of view." 58 But if we do not accept the procedure because

we would rather accept some alternative account of the moral
point of view, some alternative decision procedure that offers
what we regard as a more appealing account of impartiality or a
more plausible conception of interests, then we have a jurisdic-
tional challenge that cannot be handled as easily. At this point,
Rawls can only invoke the claim of "reflective equilibrium," a
claim that only postpones the issue another step.
"In searching for the most favored description of this situa-
tion [the decision procedure], we work from both ends," we are
told. We are to construct the choice situation and carefully
examine the results, adding one weak condition to anotheri
But presumably there will be discrepancies. In this case we
have a choice. We can either modify the account of the
initial situation or we can revise our existing judgments,
for even the judgments we take provisionally as fixed
points are liable to revision. By going back and forth,
sometimes altering the conditions of the contractual cir-
cumstances, at others withdrawing our judgments and
conforming them to principle, I assume that eventually
we shall find a description of the initial situation that both
expresses reasonable conditi0ns and yields principles
which match our considered judgments duly pruned and
adjusted. This state of affairs I refer to as reflective equilib-
rium. 59
Yet if each of us conscientiously goes through this process
and arrives at a different "reflective equilibrium," as presum-
ably sincere proponents of the five moral decision procedures
described above already have done, then reflective equilibrium
becomes a framework for reasonable moral disagreement. It is
a position with different substantive implications for different
practitioners of the process. We have only to replace the

58. Ibtd., p. 136.

59. Ibtd., p. 20.
18(, OPTIONS FOR liBERAL THEORY

pronoun "we" with the pronoun "I" and envision the resulting
dissension among conscientious proponents of these five pro·
cedures, and others that might be devised.
Of course, to claim that the development of such moral
decision procedures leaves the result inconclusive-in the
sense of leaving it open to reasonable disagreement-is not to
claim that the result is entirely arbitrary. A contested reason·
ableness is quite different from an arbitrary whim. In other
words, our principles can fail to fulfill the requirements for
position I without providing any basis for concluding that they
must, therefore, be subjective, in the sense of position IV.
There are two remaining positions in between (positions II and
III). Within this middle ground, a controversial family of rea-
sonable solutions (whether inviolable or not) can be distin·
guished from the claim at position IV that every arbitrary taste
must be as good as every other.
I will return to this issue below. First, however, I will turn to
a related source of inconclusiveness, one that I have dubbed
the foreseeability problem. For purposes of argument, let us
make the heroic assumption that perfectly unbiased and com·
petent moral judges are available who can correctly judge each
particular moral situation presented them. We might imagine a
panel of judges operating with perfect consensus; or perhaps
we might imagine ourselves, endowed with such powers, as
members of this idealized panel. They prescribe particular
actions or courses of conduct from the alternatives available in
each situation; they do not, however, go so far as to offer
general principles that might prescribe solutions beyond each
isolated case at hand. Even with this limitation it is an ex·
tremely optimistic assumption. My argument is that even
were such perfect moral judges available to evaluate particular
cases, we would still have to regard any general principles as
inconclusive and open to being overridden or drastically re·
vised.
This argument is designed to focus on the possibility of a
method based on concrete intuitions or clear moral convic·
 OPTIONS FOR LIBERAL THEORY 187

tions about particular cases. The jurisdiction problem dis-

cussed above challenges a contrasting strategy, that of positing
a procedure as the fundamental commitment and then deduc-
ing general principles directly from the procedure (or from the
procedure and the assumptions it permits, as in Ackerman's
dialogue). As we have seen, this strategy leaves the result open
to reasonable disagreement since the adoption of one set of
initial assumptions rather than another is open to contested
interpretations of the two crucial component notions of the
procedure-impartiality and interests.
Nevertheless, one might imagine a quite different strategy,
one based on generalization from particular cases, from partic-
ular intuitions, or conscientious moral judgments about what
ought to be done in specific situations. Just as the strategy of
positing a decision procedure and deriving its implications is
essentially deductive, the strategy of positing particular cases
and generalizing from them is essentially inductive. Rawls, as
we saw above, hopes that the two strategies will coincide in
reflective equilibrium. But a realistic appraisal of the varieties
of conscientious moral opinion and the varieties of decision
procedure apparatuses that might be devised to support them,
leads to the conclusion that reflective equilibriums are likely
to reflect a wide range of moral disagreements.
Furthermore, this inductive strategy faces a special limita-
tion of its own. Any moral principle that might be generalized
from a finite list of particular cases must be vulnerable to being
overturned in the future, even if our judgments of particular
cases came from a panel of perfect moral judges. A general
moral principle, as opposed to a particular judgment (or a finite
list of particular judgments) must apply to an open-ended class
of possible cases. 60 Whatever descriptive dimensions are em-
ployed to identify alternatives favored by such a general princi-
ple, they must amount to an incomplete description of each

60. See R. M. Hare, Freedom and Reason (Oxford: Oxford Umvemty Press,
1963), chapter 3.
188 OPTIONS FOR LIBERAL THEORY

particular state of affairs or course of conduct prescribed. It is

always possible that other aspects of the partially described
states of affairs lor courses of conduct) will be of sufficient
importance to override the principles. This vulnerability ob-
tains even if the principle conforms perfectly to all particular
decisions taken up to now by our hypothesized perfect moral
judges. 61
The difficulty can be dramatized with a story I will borrow
from Lawrence Tribe. An English couple acquire a magic talis-
man, a monkey's paw, that will grant them three wishes:
The couple first wish for £200; shortly thereafter a mes-
senger arrives to inform them that their son has been
killed in a factory accident and that his employer has
offered £200 out of sympathy. Their second wish is that
their son return; it is answered by a strange knocking at
the door that the parents somehow know to be their son-
but not in the flesh. The tale ends with the couple's third
wish, that the ghost go away.~ 2

The couple's three wishes, like the prescriptions in any

principle of moral or political choice, pick out certain dimen-
sions of the desired alternatives, dimensions that must incom-
pletely describe the states of affairs obtaining when the desired
choices are realized. Just as the couple had each of its wishes
fulfilled but with accompanying factors so terrible that they
were fulfilled to its extreme regret in those particular cases, so
may any general principle be fulfilled but with accompanying
61. I have adapted the nouon of a superhuman moral1udgc from Dworkm
Dworkm's "Hercules," however, has powers extcndmg to the faultless con-
struction of general pnnctples. See Dworkin, Taking R1ghts Serwusly. pp.
105-30. The argument also depends on famthar hm1tauons of mducuvc
mference made famous by Hume and Popper. Sec Karl Popper, The Logic of
Scientific Discovery (London: Hutchmson, 1959), especially pp. 29, .~69-70,
and Conjectures and Refutations (London: Routledge and Kcgan Paul, 196.1),
espec1ally pp. 42-48.
62. Laurence H. Tnbe, "Pohcy Sc1cncc· Analys1s or Ideology," Philosophy
and Public Af{al1S 2, no. 1 (Fall 1972). The story appears on pp. 102-m and
Tribe employs 1t to show the hmtts of "hteral-minded dcv1ccs." He borrowed 1t
from Norbert Wtener who used 1t to show the hmits of technology.
 0 P T I 0 N S F 0 R L I 8 E R A L THE 0 R Y 189

factors so terrible that its proponents would view its fulfill-

ment with extreme regret in particular cases.
In another work, I explored counterexamples that can be
arrayed against most of the principles currently prominent in
liberal theory-procedural principles such as majority rule and
unanimity; structural principles such as equality, maximin
justice, and utilitarianism; and absolute rights principles such
as Nozick's side constraint theory. These counterexamples
reflect the power of the monkey's paw in the parable just cited.
They are each compatible with complete fulfillment of the
states of affairs or courses of conduct specified by the principles
under discussion. Yet in each case, other elements of the
incompletely described situation or policy are so terrible that
reasonable proponents of the principles under discussion
would surely wish the prescribed result to be avoided in those
cases. 63
Of course, if we could somehow completely enumerate or
foresee in advance all of the possible negative contingencies
that might accompany the incompletely described states of
affairs or policies prescribed by our general principles, then we
could reformulate our principles to include exceptions or qual-
ifications for those particular overriding contingencies. The
difficulty is that we lack any theoretical basis for ever presum-
ing that we have completely enumerated all the contingencies.
Because any such identification must always be open to revi-
sion, any principles generalized from the particular cases al-
ready decided must always remain vulnerable to being over-
turned by the unforeseen overriding factors presented by some
new case. Hence there is a crucial source of inconclusiveness
applying to such inductive strategies just as there was a crucial
source of inconclusiveness applying to the deductive strategies
considered earlier.
It should be obvious that principles generated by either the
deductive or inductive strategies just considered, like princi-

63. I have made this more general argument m Tyranny and Legitimacy,
part 2.
190 OPTIONS FOR LIBERAL THEORY

ples generated by more direct appeals to intuition, will clearly

fall short of the kind of conclusive solution required by posi-
tion I. These strategies only provide a basis for inconclusive
principles. Such principles, even if they are formulated to hold
inviolably, must only lay claim to a tentative and controversial
kind of validity-a kind compatible with positions II and III,
but not with position 1. 64 Such principles should either be
subject to reformulation whenever contravening cases arise, or
they should be formulated at the outset to hold only prima
facie, to be traded off within a framework of conflicting princi-
ples.
Should this kind of inconclusiveness be a disappointment?
There is a common tendency to believe that anything short of
the absolutist position (I) must be entirely "relative" or "arbi-
trary," falling into some version of the subjectivist position
(IV). The psychological literature on moral reasoning shows
that this tendency is common among ordinary moral reason-
ers.65 This tendency also reflects a long philosophical tradition.
Kant, for example, believed that the only possible alternative
to his absolutist account of the moral law was the acceptance
that morality must be "merely a Chimerical Idea without
truth." 66 And it has not been uncommon for more recent
theorists to believe that the dichotomy between absolutism
and subjectivism (sometimes also treated under the heading of
relativism) must be exhaustive of the possibilitiesY Leo

64. I assume that absolutist or rationally unquestionable pnnciples presume

to settle the questwns to which they apply conclusively.
65. See Lawrence Kohlberg, "Continuities in Childhood and Adult Moral
Development Revisited," m Paul B. Baltes and K. Warner Scha1e, eds., Life·
Span Development Psychology (New York: Academic Press, 1973), pp.
179-204; Elliot Thriel, "Conflict and Transition in Adolescent Moral Develop-
ment," Child Development 45 (1974): 14-29; and my "Relativism, Liberalism
and Moral Development," in Richard W. Wilson and Gordon J. Schochet, eds.,
Moral Development and Polwcs (New York: Praeger, 1980), pp. 85-106.
66. Kant, Groundwork of the Metaphysic of Morals, p. 112.
67. The overly s1mphfi.ed dichotomy between absolutism and relatiVISm IS
cnucised m Thomas London Thorson, The Logic of Democracy (New York:
Holt, Rmehart and Wmston, 1962), chapters 1-5, and in Shia Moser, Absolut-
 OPTIONS FOR LIBERAL THEORY 191

Strauss offers an especially influential example. He attacked

Isaiah Berlin's assertion of conflicting controversial principles
as "a characteristic document in the crisis of liberalism-of a
crisis due to the fact that liberalism has abandoned its absolut-
ist basis and is trying to become entirely relativistic."68 Strauss
chided Berlin for strayin~ into the "impossible middle ground
between relativism and absolutism." Berlin would have been
more honest, Strauss implied, if he had admitted that his
rejection of absolutism committed him to nothing more than
subjectivism or relativism. 69
But if our principles fall short of absolutism, by itself, that
fact is insufficient to trap us in subjectivism. Between the
rejection of position I (absolutism) and the acceptance of posi-
tion IV (subjectivism) there stand two other meta-ethical possi-
bilities-positions within which all of us may struggle with
our inconclusive and controversial convictions that are sup-
portable from various objective perspectives. The possibility of
reasonable disagreement requires that there be a role for rea-
son, even if it is one that is insufficient to settle these ques-
tions in any conclusive way.
Unless the possibility of this middle ground is explicitly
affirmed, liberal theory is rendered vulnerable to a kind of
moral self-destruction-vulnerable to a delegitimation of itself
as a moral ideology. The religious and metaphysical consensus
that once might have supported an absolutist version of liber-

1sm and Relativism m Eth1cs (Springfield, Ill.: Charles C Thomas, 1968). Both
offer exhaustive documentation of the prevalence of the distinction.
68. Leo Strauss, "Relativism," m Helmut Schoek and James W. W1ggms,
eds., Relativism and the Study of Man (Princeton: D. Van Nostrand, 1961), p.
140.
69. Ib1d. Strauss does not exphc1tly commit himself to the cla1m that the
middle ground IS "1mposs1ble." H1s pos1t1on on the m1ddle ground can be
mferred, however, from the fact that Berlin is criticized for abandoning
absolutism without thereby embracmg relauvism, and from the fact that the
"cris1s of liberalism" is diagnosed (m the passage c1ted) as "due to the fact that
hberalism has abandoned Its absolutist bas1s and 1s trying to become entirely
relativistic" (emphasis added). The thrust of his posmon IS that Berhn has not
faced the relativistic implications of his reJection of absolutism.
1'12 0 P TJ 0 !'\: S F 0 R 1.18 ERA L THE 0 R Y

alism has evaporated in the modern era. And the methodologi-

cal resources available to contemporary liberal theory clearly
fall short of any plausible absolutist claim. Hence, unless the
middle ground positions between absolutism and sheer subjec-
tivism arc affirmed, a quite reasonable skepticism about posi-
tion I would provide the basis for trapping us into position IV.
The denial of conclusive principles would limit us to entirely
arbitrary ones. Subjectivism denies any objective, impartial, or
appropriate moral perspective from which our principles are
rationally supportable. It reduces us to a position in which any
moral perspective must be as good as any other, in which
private preferences about justice or equality become indistin-
guishable in kind (although perhaps not in degree or intensity)
from private preferences about flavors of ice cream.
A moral ideology that asserts such subjectivism must dele-
gitimate itself by affirming the arbitrariness, the sheer subjec-
tivity of its prescriptions. Modern and defensible versions of
liberal ideology, by inevitably falling short of absolutism, will
find themselves trapped in subjectivism-when they are also
committed to, or when their proponents grant, absolutist
expectatiOns that require that something like position I be
satisfied in order for what we have called position IV, subjectiv-
ism, to be avoided.
The affirmation and creative development of the middle
ground, positions II and III in my scheme, are the best line of
defense for liberal theory. Even if one were to arrive at no more
of a solution to the substantive trilemmas than an acceptance
of conflicting prima facie principles, to be traded off in particu-
lar cases, such acceptance would still provide the basis for a
defensible liberalism. Once this interpretation of the meta-
ethical alternatives is adopted, a possibility is granted within
which liberalism can thrive as a defensible moral ideology.
In fact, without some further breakthrough, permitting a
tentatively plausible rigorism (position II), we are left with no
more than an intuitionism of conflicting principles, to be
traded off in particular cases, a version of what I have been
 OPTIONS FOR LIBERAL THEORY I'l3

calling position III. Admittedly, such a position is at an ideo-

logical disadvantage compared to many of its competitors. It
affirms a plurality of conflicting principles without adding up
to a single ideal or unified vision for the organization of
institutions in a just society. It lacks a unified program because
it lacks any general priority relations among its conflicting
parts. How its principles are to be balanced remains an open
question, to be faced in particular cases as they present them-
selves. It leaves unclear how much one conflicting principle is
to be emphasized, compared to another. It does not define a
general direction for public policy since it offers us no single
vision of social justice in clear focus. Rather than aspiring to
some new and systematic solution (some version, in other
words, of positions I or II), it embraces the conflicting princi-
ples in our trilemmas, to be balanced against each other in
particular cases.
Accepting such a plurality of values, to be weighed prima
facie against each other, even under the best conditions, would
represent a decisive lowering of expectations by the standards
of contemporary theory. Such a version of position III would
then be received by many as a disappointment or as a non-
theory/0 for it fails to offer us a systematic solution, a vision of
the just society in clear focus to be gradually approached as
conditions permit.
This limited liberalism, offering us ideals without an ideal, a
plurality of principles without a unified vision, may be the
most honest response to the true difficulties of distributive
justice. A more rigorous and conclusive solution would only
glibly paper over dilemmas, or trilemmas, that will present
themselves in any case. It is in the spirit of an effort to clear
ground-where such an honest but limited liberalism may
thrive-that this book is offered.

70. As Rawls says "An mtumomst conception of JUStice is, one might say,
but half a conception." Theory of fustJce. p. 41. For an overview of liberal
rejections of mtmtlonism see section Hof that work.
 INDEX

Absolute principles, 170-71, 172-75, cations, 51-55, 62-63, 64-67; and

189 equal hfe chances, 50-64, 104; and
Absolutism, 174-75, 176-77, 178, equal opportunity, 35-43, and lib-
186, 190, 191-93 erty/equahtyconflict, 145-46, pub-
Ackerman, Bruce, 158, 178, 180-81, he policy re, 147-51; sacnfice of, in
182 equal opportumty trilemma, 50-82
Advantaged strata, 7, 55, 64, 66, 135,
159-60; and intervention strate· Background mequaht1es, 1-2, 3-4, 5,
gies, 66, 68, 70, 71-72, 80, 82, and 6-9, 22, 84, 106-07, 129, 131, 157;
unequal talent development, cost to hberty in ehmination of,
51-55, 61, 66. See also Mmority 132, 144-46; and equal life chances
groups option, 55, 61, 62, 64, 67, m pohtl·
Affirmative act1on, 85, 95-96, 98. See cal process, 162-63, 165, 167; and
also Compensation, Preferential preferential treatment, 89-90, 93,
treatment, Reverse d1scnmmat1on 95-97, 105; process equaht1es and,
Aggregative pnnc1ples, 15, 16, 159-60, 167-68, 169; in tnlemma
122-23, 130; m group compensa· of equal opportumty, 47-51; and
tion, 124, 125 unequal life chances, 51-55, 104.
Annenberg, Walter, 164 See also Arbitrary native character-
Anti-Defamation League, 98-99 IStiCS; Preferential treatment
"Appropriate moral perspective," 172 Bakke case, 21, 85, 87, 90, 95, 96, 97,
Arbitrary native charactenstlcs, 4, 5, 109n2, 149-50
32-35, 87; defined, 28, 32, and pref- Barry, Brian, 16n9, 22
erential treatment, 93, 94, 95-96, Bentham, Jeremy, 174
97, 105; and problem of assign· Berhn, Isaiah, 39, 175, 191
ment, 24, 25-30; as qualificatmns, Bickel, Alexander, 88
108-{)9 Blackmun, Harry, 147n51
Assignment, problem of, 11, 13, Blacks, 18, 47, 86, 95, 148; and equal
16-19, 45, 55-56, 67, 10~9, 132; opportumty, 128-29, 152-53, as so-
equal opportumty and, 19-30; in Cial group, 114-18, 122, 126
lottery system, 59, 61-62; preferen- Bond, Juhan, 164
tial treatment in, 85-87. See also Bronfenbrenner, Urie, 72-74, 75
Merit, principle of Buckley v. Valeo, 8, 163-64, 166
Autonomy of the farmly, 5-6, 7-8,
33n27, 84, 89, 106-07, 108, 110, Carens, Joseph, 137-43
132, 159; and background mequah- "Cucularity" thes1s ILmdblom),
ties, 67-82; defined, 5, 35, 58; and 161-62
development of talents and qualifi- Coerc1ve interference, 36, 38-42, 58,

195
196 INDEX

Cocrc1vc mterfercnce jcontmued) Edelman, Manan Wnght, 88-89

63, 64-65, 68, 79, 132, 166, and Educatwn, 40-42, 46, 68-75, 96, 148,
autonomy pnnc1ple, 81; m freedom !55, prcceptoral, 140-41
of property, contract, and employ- Efficiency, 21, 137, 138, 144, and lot-
ment, 140, 144, JUStification of, tery system, Ill, 112-13, and
39-40, 65-66, 77-78, m stnct equality of hfe chances, 55-56, 61,
equality, 135, m preference forma- 62-63
non, 141-42 Egalitanamsm, 3, 135, 137-39, 141,
Coleman, James, 68-69, 71 144-45
Compensation, 30, 46, 47, 67, 82, Equal concern and respect for every-
96-105, ISO; group, 113-31; m- one, 158-60, 168, 169, 180-85; hy-
tergcnera tiona!, II 7-18 pothetical verswns of, 169, 177-82
CompetitiOn, 107, 159-60, and lot- Equalopportumty, 1-10, 14&-47, 158,
tenes, 110-13, and compensation, 168; autonomy of the family and,
113-31 See also Fau competition 35-43; blacks and, 128-30, cost to
assumptiOn liberty of, 107, 132, 145-46, 160,
Consensual relauons, 35, 3&-38, 42, defined, 128, equality of Ide
43n45, 58, 135-36, 146; justifica- chances and, 30-35; problem of as-
tion of mterference m, 65-66; m Signment m, 19-30, rad1cahsm of,
marketplace, 139-40 106-07, 146; Rawls's doctnne of,
154-58, strong doctnne of, 20-21,
DeFums, Marco, 90, 98 145-46 See also Tnlemma of equal
DeFums case, 21, 85, 87-89, 96, 149 opportumty
Developmental condltlons· differen- Equahty, 13-14, 16, 17, 106-51,
tial, SJ-54, 110; equalization of, 6, 158n 12, 189; confl1ct w1th hberty,
32-33,9&-97, 152-57, 159,parcntal 2-4, 7-10, 132-46, 168-69; lack of,
control of, 51-55, 62-63, 64-67, m modern society, 47-50, and the
and preferential treatment, 89-90, law, 6-8, 158, 159, 160, 167-68,
104-05. See Background mequah- levelmg approach to, 14, 167, 168,
ties; Quahficatwns, for positions and max1mm, 157nl1
Disadvantaged strata, 8, 89-90, 92, Equahty of hfe chances, 4-6, 7, 18-19,
93, 94, 96, 99, 102-05, 119, 120, 20, 22, 30-35, 47, 84, 95, 106-07,
121,124, 149,moralclalmo f, 13n6, 145; causal mechamsm for unequal
14-16, soc1al expenmentatwn for, life chances, 53-54; and compensa-
147-49; and strategies of mterven- tiOn for past d1scnminauon, 9&-97,
twn, 67-68, 71-72, 73, 77, 105; 103, 104-05, defined, 32; and defi-
targetmg of benefits to, 149-50 mtwn of quahficatwns, 108; and
D1stnbuuon of goods and welfare, 1, equalization of developmental con-
122, 129-30 dltlons, I 53, 154-5 7, fauness and,
D1stribuuvc jusuce, 1, 2, 45-46, 107, 108, 109, 159; m group com-
133-35; chmce of pnnc1ples m, pensauon, 120-21, and mequahues
177-82, 184, 193; three problems between groups, 128-31; lottery
of, 11-19. See also Reahst1c budget system and, 111-12; pubhc pohcy
constramt rc, 147-51; and sacnfice of fam1ly
D1stnbutive pnnc1ples, 12, 15-16, 19 autonomy, 5!-55, 64-67, 68, and
Dorn, Edwin: Rules and Rac10l sacnfice of ment pnnciple, 55-64,
Equality, 128-29, 130 89, 108, 109
Dworkm, Ronald, 158 Equality of outcomes, I, 3, 6, 48,
 INDEX 197

50-51, 106, 107, 168-69, cost to cive mterference

hberty m, 131-46 Great Society program, 147, 152-53
Ethical positiOns, 170-75, 190-93 Greenawalt, Kent, 91
Groups, 24, 30, competition amo11~
Fau competltlon assumpuon, I, 5-6, and compensauon, 113-31· geneuc
19-20, 55-56, 84, 104-05, 153, £au- inequahues between,33n27, proxy,
ness m, 22-23, 24-25, 26, 34, and for actual disadvantaged, 93-94. See
group compensation, 113-14, also Mmonty groups
116-22, 125-31; and lottery sys-
tem, 62, 110-13; and ment, Haley, Alex: Roots, 100-02
107-10; and preferential treatment, Hamilton, Alexander, 176
89-90, 96, 114-15, 122-24, 127; Harm pnnciple, 14-15,39-42, 43n45,
pubhc pohcy imphcations of, 140, 146, 166; m family mterven-
146-51; and unequal positions uon strategies, 37n32, 76-77; m
131-46 I
government mtervention, 36n28,
Fauness, 5, 21, 34, 60; m assessment 57-58; m sacnfice of merit pnnci-
of quahfications, 28-30; back- ple, 135-37
ground, 22, 107, 109-10, 152, Harsanyi, John C., 177
159-60, 162-63, 168; cost to, m Hayek, F. A., 137
equal hfe chances opuon, 55-56, Health care needs: equal consider-
61, 62; and ment, 107-10; proce- ation of, 7-8, 45-46, 159 160
167-68 I I
dural, 4, 22-23, 24-25, 31, 85-86,
107, 108, 109-10, 111, 112, 159, Heard, Alexander, 164
168. See also Equality of hfe Herrnstem, R. J., 33n27
chances; Ment, pnnciple of Houlgate, Laurence D., 37n32
Family, the, 1-2, 3-4, 33, 35, 63, 75, Huxley, Aldous: Brave New World, 65
79, 168, defined, 36-37; and educa-
tiOnal environment, 69-70 1 and Idealtheory,6,8,9,30,44-47,53,66,
equabzation of developmental con- 178, comprom1se m, 150--51 154
dmons, 51-55, 62-63, 64-67, 153, 160, 168-691 conflict of equ~l op~
154-57; equal opportunuy and, 11- portumty and hberty m, 146-47;
43. See also Advantaged strata, Au- departure from, m tnlemma of
tonomy of the family equal opportumty, 82-84, 104-05,
Fust Natwnal Bank v. Belotti, 165 106; lack of coherent, 10, 104 106
154, 193 I I
Fiss, Owen, 114-15, 116, 117, 120,
122, 123, 124, 127-28, 130 Inconclusiveness, 10, 169-93
Foreseeabliity problem, 172n34 I 78 Indoctrmat10n, 38n33, 45n1, 80-81,
186-90 I I 141, 142, 143, 144, 161
Freedom of property, contract, and Intimate relauons, 42-43. See also
employment, 63-67, 107, 111-13, Consensual relauons
134, 135, 136, 138-40, 144, 145 Intuitionism, 174-75, 176, 178 186
190, 191-93 I I

Inviolable pnnciples, 171, 173-75

Gill, Gerald, 148 Israel, 65n37, 124
Government and social instituuons:
mtervention by, 3, 57-59, 66-82, Jefferson, Thomas, 176
148-49; lack of ideal for, 193; moral Jencks, Chnstopher, 52-54
demands on, 45-46. See also Coer- Jensen, Arthur R., 33n27
198 INDEX

Jews, 98, 12.~-24 Lmdblom, Charles E.: "cuculanty"

Johnson, Lyndon B., 152-54, 157 theSIS, 161-62, 166
Jurisdiction problem, 172n:J4, 178-86 Lottery system, 57-58,61-64, 110-13
Justice, 20, 193; compensatory, 46,
4 7, mtergenerauonal, I 00-02,
McCarthy, Eugene, 164-65
117-18; retnbutlve, 46; strict com-
Maximm jusuce, 13, 14-16, 17,
pliance, 44, 46-47, 52, 82, 147. See
157n11, 182, 189
also Distnbutlve JUStice; Maximm
Merit, prmciple of, 4-6, 19-30, 51-52,
JUstice, "Patterned" principles
68 84 106--07, 132, 145, 159, 167;
deflned, 22, 107; and equabty of life
Kant, Immanuel, 174, 190
chances, 30-35, 55-64; and equali·
Kurland, Philip, 88
zauon of developmental condt-
tions, 153, 154-57, 160; fauness
Legal Serv~ces Corporation, 94, 167
and, 107-10; and famlly autonomy,
Leveling down strategies, 66, 67-68,
37, hbertyofemploymentand, 112;
69-70,71-72,80,82
and preferential treatment, 86-87;
Leveling up strategies, 66, 67-68,
pubhc pohcy re, 147-51; sacnfice
78-80, 82, 154, 167-68
of, 29-30, 89-90, 93, 94, 103,
Liberalism (liberal theory), I, .~o. 43,
104-05, 107-11, 114; sacrifice of
5~51, 84, 104, 168-69; conflict be·
liberty in, 136-37; and unequal life
tween hberty and equahty at corr chances, 53-54
of, 50-51, 132; in contemporary
Meta-ethics, 169-70; and moral pnn-
pubhc policy and political theory, ciplcs, 170-75, 192-93
152-69; dcparture(s) from, m re· Mill, John Stuart· On Liberty, 39-40
sponse to tnlemma of equal oppor- Mmority groups, 85-86, 89, 113-14;
tunity, 109, 110, 112-13, I:t~-.~4, advantaged, 92-93, 95-96, 97-98,
I46-47; equahty m, 2-4, f.--7, R-IO; 102, I22-23, 130-3I, and preferen-
hm1ted versiOn of, 9-I 0, I 58, tial treatment, 9~95, 114-15, 122.
19I-93; negative hberty m, J9, op- See also Blacks; Preferenual treat-
tions for, I52-9.~; problem of as-
ment
signment m, II, I9, 20; solutions to Mistargcting of benefits, 103-04,
moral conflict in, I ?f.--9.~
105n93, 150; m group compensa-
Libertanans, .~. LU, 1.~7, 168-69 tiOn, 121-22, 124, 130-31
Liberty, 40, 50-51, 156, IS7nll, Moderate scarcity, 44-45, 46, 47, 66,
I 59-60; conflict w1th equality, 2-4,
82, 147
7-10, 1.32-46, 168-69, conflict Moral confhct, 170n32, 171-72, 175,
with, in lottery system, 6.3-64; cost 176, 178, 183-86, 191
to, m ehmmation of mequahues, Moral dectsion procedures, 172,
7-8, 9; cost to, m equahty of out- 174-75, 177-90
comes, 131-46; cost to, m equal Moral pnnctples: generahzation of,
opportunity, 107, 132, 145-46, 160; 187-90; meta-ethical clatms re,
cost to, m pohucal cqualuy, 16.3, 170-75, 176, 191-93
165, I66; cost to, m process equah-
ues, 167-68; ethical positions re
conflicts m, 170-75; negauve, Nagel, Thomas, 172
.W-40, 42, 43, 84, 135-.17, 140, 146, Neutrahty (concepti, 180-81, 182
private sphere of, 4, 39-40, 42-43, Nonexistence, 41, 118-19. See also
47, 61, 62, 63, 65, 66, 107, Ill, 145 Persons, potenual
 INDEX 199

Normative ethics (of social chmcc), thetlcal vcrstons of, 169, 177-82,
169, 170, 176 radical character of, 168
Noz1ck, Robert, I3, 125, 136n36, 137, Pubhc pohcy, 9, 83, 106, 146-Sl;
189; and stnct equality, 133-.%, 153-54, 169, 193
139-40
Quahficauons, for postttons, 4, 5,
ObJectively valid pnnc1plcs, 171-72, 23-30, 52, 54, 62, 107-09, 112; de-
173-75 fined, 108, development of, 3-4, 5,
Ongmal position (Rawls), 59-60, 172, 20, 31-33, 61, 108, ISO; cquahza-
178-80, 181' 182, 184-85 uon of development of, 6, 67-82,
96-97; and equal hfc chances,
Parfit, Derek, 100n91 34-35, 55-57, fatr assessment of, 3,
PaternalistiC mferences, 12, 37n.l2, 23-30; and fatr compctttton,
65-67, 72 136-37. See also Developmental
"Patterned" pnnciples, 133-36 condttlons
Persons, potential, 59-60; moral
cla1ms of, 100-02 Race. See Arb1trary native character-
Plato: Republic, 65 IStics; Blacks; Minonty groups;
Political Action Committees, 167 Preferential treatment
Political equality, fr-8, 158, 160-67, Rae, Douglas, 16n10, 182n56
169 Rawls, John, 125, 143, 158, 174, 175,
Political mfluence, 7-8, 159, 161-67 I77-78, 193n70; doctnne of fatr
Pohucal pnnciples: meta-ethical equality of opportumty/famtly au-
cla1ms re, 170-75, 17fr-77; moral tonomy, 154-58; max1mm justtce,
deciswn procedures for chmce of, 13, 14-16, moderate scarcity,
177-90 44-45; ongmal posltlon, 59-60,
Pos1t1ons, soc1al, 12, 13, 16, 17-18, 19, 172, 178-80, 181, 182, 184-85,
22-23, 55; rankmg of, 32; and lot- strict compliance theory, 46; strong
tery system, 111-12; unequal, I,:~. doctnnc of equal opportunity, 20,
104, 107, 109, 131-46. See also Fau 2I, theory of "primary goods," II,
competition assumption 12n.3, 28,32
Posner, Rtchard, 92 Reahsttc budget constramt, 4S-46,
Powell, Lewts, 109n2 66,68, 77,78
Preference formatwn, 139, 141-43 Rcflecuvc cqutltbnum, 185-86, 187
Preferential treatment, 21, 82-105, Relauvism, 175n42, 190, 191
108, 149; for the dtrectly disadvan- Reverse discnmmatwn, 55-56, 61,
taged, I31, I49-50; m group compe- 85,86-89,92-94
tition, 1I3, 114-15, 120, 122-23, Rtgonsm, 174-75, 17H, I H6, 190,
127-28, 130-31. See also Affirma- 191-93
tive action; Compensation; Disad-
vantaged strata; Reverse dtscnmi- Sacrifice of pnnciples: mimmum con-
natlon dition for, 84-85, 89, 93, 94, 10.~-05,
Primary goods (theory), 11, 12n3, 28, 149
32, 179; mcome as, 143 Sandalow, Terence, 90, 91
Pnvate-property·market (PPM) sys- Sartre, Jean-Paul, 175
tem, 137-38, 139, 140, 141, 143 Self-regardmg sphere of actton, 40-42
Process equahtles, 6-10, 158-60; con- Separateness of persons, 2, 19, I07,
flict wtth hberty, 9, 167-69; hypo- 121-27, 1.l1
200 INDEX

Smdler, Allan P, 149 Amencan pubhc tdeology and poht-

Smger, Marcus, 174n39 tcal theory, 152-SR; basts of,
Smger, Peter, 177 159-60; general verswn, 9-10, 176,
Slavery, 60n34, 122 192, 193,opuonsre,SI-67, 106-07,
Smtth, Adam, 181 154, 158; responses to, 10f1-SI;
Soctal duty, 44n1, 137-38, 139, strategtes of mterventwn m,
141-42, 143, 144 67-IOS
Soc1al group, 120; blacks as, 114-18,
122, 126; preferential treatment for, Umverstty of Cahforma at Davts, 90,
127-28, 130. See also Groups; Mt- 94n85, 149-50
nonty groups Umverstty of Washmgton, 87-89
Soctahzauon, 138, 141-4.~, 144. See Uuhtanamsm, 14-1S, 16, 17, 121-22,
also Indoctnnauon 125, 172, 177-78, 189; aggregate
Solzhemtsyn, Aleksandr I , 126 pnnctple, 13, 14-16, 17-18, 181;
Sovtet Umon, 59n30, 126 average pnnctple, 60nn.B, .~4, 179,
Sowell, Thomas, 93 182, sympathetic "spectator" m,
Statistical dtscnmmauon, 24, 25 181, 182
Strauss, Leo, 190-91
Structural pnnctples, 11, 12-19, 189
SubjectlVlsm, 174-75, 176, 178, 186, Value, problem of, 11-12, 19,32
190-93 Values, plurahty of, 19.~
Voucher plan, 70-72
Thoday, J. M., 33
Thought control, 141, 145. See also Walzer, Mtchael, 158
Indoctnnauon, Soctahzauon Westermarck, Edward: Eth1ca/ Re/a-
Tnbe, Lawrence, 188 llvlty, 175
Tnlemma of equal opportumty, 5-6, Whtte, Morton, 176
7-10, 20, 33n27, 35, 44-105; m Wtlhams, Bernard, 30