From Aristotle to Cicero

From Aristotle to Cicero

Essays on Ancient Philosophy

GISELA STRIKER

1
 3
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Gisela Striker 2022
The moral rights of the author have been asserted
First Edition published in 2022
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2021945150
ISBN 978–0–19–886838–5
DOI: 10.1093/oso/9780198868385.001.0001
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
 Acknowledgements

Chapters 1–6, 8–10 and 12–15 were previously published as follows:

Chapter 1 (orig. in German as ‘Aristoteles über Syllogismen aufgrund einer
Hypothese’) in Hermes 107, 1979, 33–50.
Chapter 2 (orig. in German as ‘Notwendigkeit mit Lücken’) in neue hefte für
philosophie 24/25, 1985, 146–64.
Chapter 3 in Ancient Philosophy 14, 1994, 39–51.
Chapter 4 in M. Frede/G. Striker (ed.), Rationality in Greek Thought, Oxford:
Clarendon Press, 1996, 203–20.
Chapter 5 in J. Gentzler (ed.), Method in Ancient Philosophy, Oxford:
Clarendon Press, 1998, 209–26.
Chapter 6 in D. Gicheva-Gocheva (ed.), The Challenge Aristotle, University
Printing House St Kliment Ohridski, Sofia (Bulgaria), 2018, 99–110.
Chapter 8 in B. Morison/K. Ierodiakonou (ed.), Episteme, etc., Oxford
University Press, 2011, 141–50.
Chapter 9 in A. O. Rorty (ed.), Essays on Aristotle’s Rhetoric, California
University Press, 1996, 286–302.
Chapter 10 in B. Reis/S. Haffmans (ed.), The Virtuous Life in Greek Ethics,
Cambridge University Press, 2006, 127–41.
Chapter 12 in B. Inwood/J. Mansfeld (ed.), Assent and Argument, Leiden: Brill,
1997, 257–76.
Chapter 13 in Archiv für Geschichte der Philosophie 83, 2001, 113–29.
Chapter 14 in Ph. Mitsis (ed.) The Oxford Handbook on Epicurus and
Epicureanism, Oxford University Press, 2020, 41–58.
Chapter 15 in Fiona Leigh (ed.), Themes in Plato, Aristotle, and Hellenistic
Philosophy, University of London Press, 2021, 121–30.
I am grateful to the publishers for the permission to reprint and translate.
Chapters 7, 11, and 16 have not been published before.
 Introduction

This collection brings together different lines of research that I have been pursuing
throughout my academic career: Aristotle’s logic and ethics, and Hellenistic
epistemology and ethics.
At the University of Göttingen, where I got most of my philosophical educa-
tion, a doctoral degree in philosophy required the addition of two secondary fields,
for which I chose the two classical languages, Greek and Latin, which I had already
learned in high school. After receiving my doctoral degree with a dissertation on
Plato’s Philebus, I continued to work on ancient philosophy, but turned to the
Hellenistic period, a field less studied by scholars than the great classics, although
it had an enormous influence on later European philosophy. Now, since teaching
is not limited to one’s specialty, during the first fifteen years, while still in
Göttingen, I was also teaching courses in contemporary philosophy—epistemology,
ethics, philosophy of mind—that kept me in touch with recent developments. Since
I moved to the United States, I have been mostly responsible for teaching ancient
philosophy, but I have continued to follow contemporary philosophy in the fields
I was studying in the ancient authors. This seemed to me the best way to engage with
historical texts, not limiting oneself to exegesis, but engaging in discussion with
authors who often had a different perspective, not treating them merely as prede-
cessors, but also to learn from them. So, for example, Greek ethics saw morality in
the wider framework of a good human life, and the ancient Sceptics were not
fascinated by questions about the existence of an external world, questions that
seem to have been prevalent ever since the time of Descartes.
A few years into my time in Göttingen, I was invited to produce a new German
translation and commentary on Aristotle’s Analytics. Since I had long been
interested in Aristotle’s logic, I accepted—vastly underestimating the difficulties
of the task and the time it would take me to complete it. In the meantime,
Hellenistic philosophy was having a kind of renaissance among philosophers
interested in the history of their subject, who had until then more or less limited
themselves to the study of the Presocratics and the two great classical philo-
sophers, Plato and Aristotle. This led to a lively exchange of ideas between
scholars, both philosophers and classicists, now working in this field. So
I continued my research in both areas, slowly making my way through
Aristotle’s Prior Analytics, first in German, then in English, while also publishing
papers on Hellenistic epistemology and ethics.

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.001.0001
x 

The present collection includes a number of essays on Aristotle’s logic that deal
with questions in more detail than would have been appropriate for a commen-
tary, including an, as yet unpublished, paper on the title ‘Analytica’ (chapter 7),
and a short piece outlining the development of Aristotle’s theory of argument that
might serve as a background to the paper on analysis (chapter 6). It was presented
at a conference to celebrate Aristotle’s 2400th birthday in Sofia, Bulgaria. The first
two chapters were originally published in German. I am very grateful for the
meticulous translations of Joshua Mendelsohn.
The four articles on Aristotle that follow those chapters are mainly the result of
teaching these subjects for many years. The second part of the collection contains
later studies in Hellenistic epistemology and ethics. After a lot of discussion and
new contributions by other historians of philosophy who had begun to investigate
Hellenistic philosophy in the preceding decades, I have been going beyond the
earlier generation of Stoics and Epicureans and Sceptics, and also sometimes
trying to correct my earlier views. The series ends with a, yet to be published,
study of a work by Panaetius, a Stoic of the second century , whose book on
appropriate action Cicero used as a model for his last philosophical work, the De
Officiis. At a time when Cicero was still being read as a philosopher in his own
right, that book, dedicated to his son, became a vademecum for young gentlemen
from the Renaissance until at least the end of the eighteenth century. It was clearly
read as a general book of morality and manners, not as a work of Stoic philosophy,
and I was interested in the changes, if any, in Stoicism that made this book
enormously influential, quite apart from the influence of Stoic philosophy in
other fields.
Special thanks are due to Peter Momtchiloff for giving me the opportunity to
bring these papers together.

Hamburg, March 2021

 1
Aristotle on Syllogisms ‘from a Hypothesis’
For Professor Jürgen Mau on the occasion of his 60th birthday,
26 September 1976

In chapters 23 and 44 of Prior Analytics I, Aristotle discusses arguments ‘from a

hypothesis’ and claims that these include the reductio-arguments. His descriptions in
both chapters show that what direct and indirect arguments of this group have in
common is that the ‘hypothesis’ is not a part of the argument, but either an agreement
to the effect that a certain proposition will be accepted as proven if another
proposition has been proven, or—in cases of reductio ad impossibile—a logical
principle for which no explicit agreement is needed.

In chapters A 23 and A 44 of the Prior Analytics, Aristotle deals with a type of

arguments that he calls syllogisms ἐξ ὑποθέσεως.¹ On neither occasion is his aim to

¹ Without wishing to enter the fray of the long-contested question as to what an Aristotelian
syllogism actually is, I should like to clarify, so that I am not misunderstood, that by an Aristotelian
syllogism I mean a valid argument or a deduction. T. Smiley (‘What is a Syllogism?’, J. Phil. Logic 2,
1973, 138) draws attention to the fact that one cannot identify syllogisms with arguments because the
same syllogism may be used in various arguments. It seems to me doubtful, however, that Aristotle
would have noticed this distinction—the definition of the syllogism in A 1.24b18–20, in any case, seems
only to refer to deductions that occur as complete arguments. On this, see M. Frede, ‘Stoic vs.
Aristotelian Syllogistic’, Archiv. f. Gesch. Phil. 56, 1974, 16ff.
In the two chapters I am considering here, Aristotle uses the word συλλογισμός in a broad sense that
seems to include all valid arguments, as well as in a narrower sense in which it only refers to an
argument in one of the syllogistic moods. The context makes clear which sense he intends in each case,
so I have not indicated this. For the sake of brevity, I will refer to syllogisms ἐξ ὑποθέσεως as
‘hypothetical arguments’ in what follows. This should not be taken to imply that they refer to what
were later called hypothetical syllogisms.
The moods of the Aristotelian syllogistic will be referred to in what follows using their traditional
Latin names (Barbara, Celarent, etc.). These names provide information about the form of the premises
as well as the conclusion of the syllogism in question. They each have three syllables whose vowels
indicate the quantity and quality of a statement: a stands for a universal affirmative statement of the
kind ‘A belongs to all B’ or ‘all B are A’, e for a universal negative (‘A belongs to no B’), i for particular
affirmative (‘A belongs to some B’) and o for particular negative (‘A does not belong to some B’).
Aristotle divides syllogistic moods into three figures (σχήματα) that differ in their placement of the
middle term: In the first figure the middle term appears once as subject and once as predicate; in the
second figure both times as predicate, in the third figure both times as subject. If one knows the figure of
a mood, one can construct the corresponding syllogism using its name. The premises of the mood
Barbara in the first figure, for instance, have the form ‘A belongs to all B’ and ‘B belongs to all C’, and

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0001
2    

characterize these arguments or to demonstrate their validity. In A 23, hypothet-

ical syllogisms are treated in the context of completeness considerations.
Following on from A 7, Aristotle attempts to show that every proof and every
valid argument can be reduced to the universal moods of the first figure (Barbara
and Celarent). In A 44 his object is to argue, seemingly in opposition to what he
tries to prove in A 23, that hypothetical arguments cannot be put into syllogistic
form. The apparent contradiction between the theses of the two chapters will not
interest us further here—it is probably best solved by restricting the thesis of A 23
to the effect that every valid argument must contain at least one syllogism, which
can then be reduced to Barbara or Celarent. Aristotle himself seems to have
intended this restriction, since even in A 23 he describes hypothetical arguments
as consisting of two parts, where only one of the parts is a syllogism.
The question that will occupy us here is rather the following: What did Aristotle
actually mean by a syllogism ἐξ ὑποθέσεως? The expression, which also shows
up in the Topics (A 18,108b12), is explained neither here nor in any other
place. This allows us to infer that it was presumed to be familiar, at least in the
Academy. As Shorey has emphasized, Aristotle is probably referring to the
procedure that is introduced by Plato in the Meno as investigation ἐξ ὑποθέσεως
(Meno 86Eff.).² So Aristotle is here speaking of a kind of argument with which the
reader or listener is already familiar, and which therefore need not be character-
ized more closely.
Aristotle also counts reductio ad impossibile as a hypothetical argument (40b25).
It is not entirely certain whether this classification was also already in place, since,
first, Aristotle does mention it himself and, second, it seems to serve to completely
classify all valid arguments under two headings.
Aristotle begins his discussion of completeness with the claim (40b23) that every
proof and every syllogism must demonstrate that something belongs or does not
belong, either universally or particularly; and—the significant part for our
purposes—either in the ostensive way (δεικτικῶς) or from a hypothesis (ἐξ
ὑποθέσεως). The ostensive method of proof had been contrasted with proof διὰ
τοῦ ἀδυνάτου in A 7 already (29a32, cf. also A 29,45b8, b14; B 14.62b29, b38); and
since reductio ad impossible is one type of hypothetical argument, one might at first
assume that it is the contrast between direct and indirect arguments that Aristotle
intends. Aristotle does, in fact, use the expressions δεικτικὴ and δεικτικῶς in the
sense of ‘direct’ or ‘directly’ in chapter B 14 (62b29ff.); he describes the difference

the conclusion ‘A belongs to all C’. For further details, see G. Patzig, Die aristotelische Syllogistik, 3rd
ed., Göttingen 1969, 11–13; W. and M. Kneale, The Development of Logic, Oxford 1962, 67ff. See 232f.
for a more precise explanation of the names.
² Cf. P. Shorey, ‘ΣΥΛΛΟΓΙΣΜΟΙ ΕΞ ΥΠΟΘΕΣΕΩΣ in Aristotle’, Amer. Journ. Philol. X, 1889,
460–2. On the Meno see R. Robinson, Plato’s Earlier Dialectic, 2nd ed., Oxford 1953, 114ff.
    ‘    ’ 3

between ostensive proof and reductio ad impossibile as follows: ‘A demonstration

that leads to an impossibility differs from an ostensive demonstration in that it puts
as a premise what it wants to reject, leading away into an agreed falsehood, while an
ostensive demonstration begins from agreed propositions.’ One could then explain
why direct proof is characterized in this passage as ostensive, that is, as ‘showing
something’, by saying that a direct proof presents states of affairs that make it
necessary that the thing to be proven is the case, whereas indirect arguments only
show why the opposite cannot be the case. However, first, ostensive proof is only
contrasted with reductio ad impossibile here, not with hypothetical proof in general,
and second, the hypothetical arguments that are treated in A 23 and 44 are not all
indirect; this actually holds only for reductio arguments.
The word ‘hypothesis’ might perhaps suggest the contrast between scientific
proof from necessary premises and arguments from ‘merely’ hypothetical prem-
ises, which need not actually be true. However, this cannot be what Aristotle
means here, for in this case, the distinction between ostensive and hypothetical
arguments would have to coincide with the distinction between proofs in the strict
sense (that is, syllogisms with necessary premises) and dialectical syllogisms,
which Aristotle mentions briefly in A 1 (24a30–b15). But as we will see in what
follows, Aristotle counts all arguments that are syllogistic in form as ostensive
arguments, and we must therefore assume that Aristotle means to specify a formal
property of certain arguments when he describes them as ἐξ ὑποθέσεως.
In view of the ensuing completeness argument, it is certainly correct, as
Bochenski³ and others have noted, that Aristotle classifies all non-syllogistic
arguments as συλλογισμοὶ ἐξ ὑποθέσεως. This label is, however, clearly not a
purely negative characterization, and so there remains the further question of
what formal commonality of non-syllogistic arguments Aristotle had in mind
when he used this expression.
This question is complicated by the fact that Aristotle distinguishes in both
chapters between hypothetical arguments through the impossible and other
hypothetical arguments.⁴ An adequate interpretation needs, therefore, not only
to show what formal difference there is between ostensive and hypothetical
arguments; it also needs to show what the various types of hypothetical arguments
that Aristotle distinguishes have in common and how they differ. In the best case,
such an interpretation would put us in a position to determine what domain of
arguments Aristotle’s completeness claim refers to and to what extent it is possible
to accept it in light of this.

³ I.M. Bochenski, La Logique de Théophraste, Fribourg (Suisse) 1947, 105. Cf. also F. Solmsen, Die
Entwicklung der aristotelischen Logik und Rhetorik, Berlin 1929, 64ff.
⁴ The fact that Aristotle has no special term for the ‘other’ cases seems to me to indicate that
originally only these arguments were called (συλλογισμοὶ) ἐξ ὑποθέσεως. The Platonic argument in the
Meno must have belonged to this class, and outside of the An.Pr. it is in any case only arguments of this
sort that are called hypothetical.
4    

II

Studying the word ὑπόθεσις seems in this case not to be of any further help.
Aristotle normally uses the term ‘hypothesis’ to refer to various kinds of under-
lying assumptions of arguments.⁵ The fact that an argument contains a hypothesis
is certainly not sufficient to render it a hypothetical argument, since hypotheses
can also occur in syllogisms. Our topic here is not merely arguments ‘with’ a
hypothesis, but rather arguments ‘on the basis of ’ or ‘from’ (ἐξ) a hypothesis—
that is, the expression refers not only to the fact that a hypothesis is used, but
rather to the role of the hypothesis in the argument. Whereas it is reasonably clear
what the hypothesis is for the ‘remaining’ hypothetical arguments (which Aristotle
discusses second in A 23 and first in A 44), the inclusion of reductio ad impossibile
arguments means that we still need to ask the question: What actually is the
hypothesis on account of which all of these arguments are to count as hypothetical?
To begin with the simpler case, direct hypothetical arguments: If one follows
the example of A 44 (50a19–26), a typical argument of this form appears to be as
follows. It begins with a hypothesis or agreement (ὁμολογία) that takes the form of
an if–then sentence, for example: ‘If there is not a single power of contraries, then
there is also not a single knowledge.’ There follows a syllogistic proof for the
antecedent of the conditional—or in any case Aristotle says so. The proof that he
alludes to in the text is, however, as Alexander rightly notes (in An.Pr. 387, 5–11)
not a syllogism. Aristotle writes: ‘ . . . that not every power is a power of contraries:
not, for example, of the healthy and the sick, for then the same thing would be
healthy and sick at the same time.’ The proof sketched here seems to consist of a
modus tollens argument and a syllogism (cf. Ross’s commentary ad loc.). Aristotle,
however, evidently assumes that it can be put in syllogistic form. From the
analogous example in B 26 (69b13ff.) one can infer, as the commentators explain,
that he had in mind a syllogism in Felapton: ‘The healthy and the sick are not
based on a single capacity; the sick and the healthy are opposites; therefore, not all
opposites are based on a single capacity.’ This, however, does not fully capture the
argument given in the text, and thus Alexander (387.1–5) and Philoponus (in An.
Pr. 358,27–31) add another syllogism as a proof of the first premise.⁶
After establishing the antecedent of the conditional, one can proceed to the
consequent, the demonstrandum, ‘on the basis of the hypothesis’. The transition is,
as Aristotle emphasizes (50a24–26), necessary, but not syllogistic.

⁵ Cf. H. Bonitz, Index Aristotelicus, Berlin 1870, repr. Graz 1955, s. v. ὑπόθεσις, 796b59: ‘logice
ὑποθέσεις eae sunt propositiones, sive demonstratae sive non demonstratae, quibus positis aliquid
demonstratur.’
⁶ The example makes clear that Aristotle’s remark about this type of hypothetical argument
represents only a rough generalization. He had evidently not verified whether the analysis he points
to—that such arguments consist of a syllogism and a hypothetical step—was really applicable in all
cases. It is thus hardly surprising that he does not consider the possibility of proving the antecedent of
the hypothesis other than by a syllogism. On this, see also pp. xxf. below.
    ‘    ’ 5

What Aristotle describes here seems to be, to put it anachronistically, an

argument in modus ponendo ponens supplemented with a syllogism for the second
premise. Aristotle initially leaves it open whether this is the only possible form this
type of argument can have. It certainly seems that the hypothesis describes the
conditional premise, and it is thus natural to follow the usage of later logicians in
viewing the use of a conditional premise as the reason for characterizing an
argument as ἐξ ὑποθέσεως.⁷ This, however, does not fully explain Aristotle’s
text, as one can see from the case of the reductio.
As described in B 11 (61a27–31), an Aristotelian syllogism per impossibile
proceeds as follows: One begins with a premise that is accepted as true and the
assumption of the contradictory of the demonstrandum. From these two things a
syllogistic conclusion is derived that is evidently false. Since the first premise was
assumed to be true, the negation of the demonstrandum is shown to be false by the
‘impossible’ conclusion. Its contradictory is thus true by the law of excluded
middle (B 11,62a13–15).⁸
The apagogical proofs of chapter A 5 and 6 (27a36–b1; 28b17–20) seem at first
to present a different procedure. Here Aristotle begins with the premises of the
mood to be established and then uses one of the premises together with the
negation of the conclusion to derive the negation of the other premise. In this
way he shows that the negation of the conclusion is incompatible with the
premises and thus that the conclusion follows from the premises. The impossi-
bility of the proposition derived arises in this case from the fact that its contra-
dictory is taken as a premise. These proofs thus proceed according to the law of
propositional logic:

ðp&qÞ & ½ðq&sÞ > p > s

In the syllogisms per impossibile that are treated in B 11–14, by contrast, the
negation of the syllogistically derived conclusion does not seem to be one of the
premises (61a24).⁹ That the derived proposition is ‘impossible’ seems to arise
simply from the fact that its contradictory is evidently true (61a25; b14). Aristotle,
however, adds in B 14 (62b33–37) that the proof through reductio ad impossibile
proceeds from two propositions that are granted, namely from one of the premises
of the syllogism and the negation of the derived conclusion. Accordingly, the
difference seems to be a matter of where the respective premises are introduced. In

⁷ As does M. Kneale in The Development of Logic, Oxford 1962, 98 (or so it seems): ‘Here the
subject-matter seems to demand an explicit discussion of the varieties of argument with one condi-
tional premise.’
⁸ Strictly speaking, of course, all that can be inferred from the ‘impossible’ conclusion is that one of
the two premises must be false. From this, and the presupposition that one of the two premises is true, it
then follows that the other is false. Aristotle makes clear at the outset which premise is to be given up by
characterizing it as the ‘hypothesis’.
⁹ Smiley, op. cit., 153, n. 2.
6    

A 5 and 6 Aristotle carries out a deduction in order to show that a certain type of
conclusion follows from certain premises, which therefore must naturally be
supplied at the outset. In B 11, however, Aristotle is talking about how a particular
type of conclusion can be proven. Now, a proof by reductio will be expedient to use
only if the falsity of the derived statement is evident (cf. Top. Θ 2.157b34–158a2;
An.Po. A 26,87a14–17) and its negation thus need not itself be assumed in advance.
That, of course, does not mean that it is not one of the presuppositions from which
the proposition to be proven follows.
The proof of the incommensurability of the diagonal, mentioned as an example
in A 24 and 44, points to a different type of reductio, of which Aristotle nowhere
offers a detailed analysis (and which could hardly be put into syllogistic form).
Here the procedure by all appearances seems to consist in deriving an explicit
contradiction from an assumption, together with premises that have been proven
or are known to be true. Accordingly, this form of reductio is based on the law of
propositional logic [p > (q&–q)] > –p.
That Aristotle uses this example, even though his own apagogical proofs
normally proceed otherwise, seems to me to be evidence that he had not distin-
guished these two kinds of reductio proof. The incommensurability proof simply
provides an especially clear example of a conclusion that is evidently impossible.¹⁰
I will thus assume in what follows that when Aristotle speaks of arguments
through the impossible generally, he is referring to the standard case he initially
describes.
No conditional premise seems to appear in an argument of this sort. One could,
of course, formulate the syllogistic part of the argument as an if–then premise, as
the formula of propositional logic presented above suggests. However, if the
expression ἐξ ὑποθέσεως referred to the use of a conditional premise, then
Aristotle himself would be characterizing the syllogism as a hypothesis, and this
is plainly not what he does. That rules out the possibility of explaining Aristotle’s
terminology by speaking of hypothetical syllogisms in the way that later became
common.
In view of the type of arguments just described, three different interpretations
have been proposed. These all have something going for them, but they also all
face problems.

¹⁰ A further type, which Corcoran (‘Aristotle’s Natural Deduction System’, in J. Corcoran (ed.),
Ancient Logic and its Modern Interpretations, Dordrecht 1974, 115) claims to find in A 29,45b1–3 and
refers to as ‘abnormal reductio’, seems to me to be the result of a misunderstanding. Aristotle does not
claim that one can transform a direct proof into an indirect one by taking the opposite of the conclusion
as a hypothesis at the outset, deriving the conclusion and finally recognizing that one has assumed the
opposite of the conclusion. He says only that if one already has an ostensive syllogism, then in every
case, one can start from the contradictory opposite of the conclusion together with one of the premises
and derive a false statement (namely the negation of the other premise) and so prove the conclusion per
impossibile. It is evidently presupposed here that the truth of the ostensive premises is known, due to
the use of the tables of premises.
    ‘    ’ 7

(A) On the oldest interpretation, due to Alexander of Aphrodisias,¹¹ the

hypothesis on account of which reductio ad impossibile is classified as a hypo-
thetical argument is the contradictory of the demonstrandum. The fact that
Aristotle undeniably characterizes this proposition in many places expressly as a
ὑπόθεσις (41a32, cf. especially B 11–14, B 17 passim) initially favours this inter-
pretation. This use of language evidently corresponds to the fact that a proposition
is assumed without one thereby wishing to commit oneself to its truth. To this
extent, this interpretation seems to provide the most natural explanation. The
characterization of reductio as a hypothetical syllogism can then be understood as
follows: The proof of the demonstrandum relies on a hypothesis insofar as its
‘recognized falsity’ is derived, and on this basis one can then infer the contradict-
ory of the hypothesis. Difficulties, however, arise for this interpretation when one
attempts to describe the formal characteristic that allows such arguments to be
treated together with the remaining hypothetical arguments. In the case of
reductio, the hypothesis is, on this interpretation, a categorical premise that is
assumed merely argumenti causa. In the remaining cases, however, it is one of the
accepted premises of the argument, a premise that also has a different form. It is
not easy to see why Aristotle would have wished to bring these two types of
argument together under a single heading. The most conspicuous commonality,
to which the characterization as hypothetical should at least be related in some
way, is, of course, the occurrence of a non-syllogistic argumentative step. This is
also what Aristotle seems to emphasize when, in both chapters, he describes
hypothetical arguments as consisting in two steps (41a23–41b1 passim;
50a23–28, 30–32): They consist, 1. of a syllogism, and 2. of a step ἐξ ὑποθέσεως,
which is indeed necessitated, but not syllogistic. According to the interpretation
discussed here, however, the hypothesis in the reductio is precisely a part of the
syllogism. One should, therefore, perhaps look for a different explanation, one on
which the hypothesis is that which permits the transition to the demonstrandum.
The difficulty that arises from the fact that Aristotle would be speaking of two
different hypotheses within the reductio is perhaps not so severe when one
considers that the classification of reductio arguments as hypothetical was perhaps
introduced only in view of the complete classification of all valid arguments.
Aristotle might thus have first used the term ‘hypothesis’ only for premises
assumed argumenti causa in a reductio and then later have had the idea of
bringing reductio arguments together with those which he had earlier called
hypothetical arguments, which would then have led to the somewhat unfortunate
double use of ‘hypothesis’. This brings us to the two other interpretations.

¹¹ In An.Pr. 256,18–25; cf. also N.M. Thiel, ‘Die Bedeutung des Wortes Hypothesis bei Aristoteles’,
(Diss.), Freiburg 1919, 26ff.; H. Gomperz, Die deutsche Literatur über die . . . aristotelische Philosophie
1899 und 1900, Archiv. f. Gesch. Phil 16, 1903, 274f.; K. Ebbinghaus, Ein formales Modell der Syllogistik
des Aristoteles, Hypomnemata 9, Göttingen 1964, 41ff.
8    

(B) According to Sigwart, Shorey, and, more recently, Mignucci,¹² the hypothesis
is the negation of the ‘impossible’ conclusion. The assumption that this con-
clusion is false enables us to infer the opposite of the first hypothesis. This
assumption is not explicitly characterized as (the) hypothesis by Aristotle, but
he does say in B 14 (62b35–37) that one must assume the falsity of the
conclusion. In addition to this, one might also appeal to the remark in A 44
(50a35–38), according to which one needs no prior agreement for reductio,
‘because the falsehood’ (scil. of the derived conclusion) ‘is evident’. With
hypothetical arguments of the first type, one requires agreement as (or to) the
hypothesis, but not with a reductio—thus the ‘evident falsehood’ must be that
which is here characterized as a hypothesis.
This interpretation seems quite plausible in view of the two-part structure of
hypothetical arguments. Difficulties, however, arise again when we ask what the
two types of hypothetical arguments have in common. Mignucci seems to assume
that it is the unproven assumption that is characterized as a hypothesis in each
case—the negation of the syllogistic conclusion in the case of reductio and the
conditional premise in the other case, since a syllogistic proof is given for its
antecedent, which forms the second premise of the modus ponens argument. We
noted at the outset, however, that the characterization ἐξ ὑποθέσεως could not
refer merely to the occurrence of an unjustified premise, since these also occur
in ostensive syllogisms insofar as they are not required to count as proofs.
Furthermore, it is questionable whether Aristotle, as Mignucci assumes, thought
of the syllogistic part of the reductio as a proof of a conditional statement. What
is ‘shown’, that is, derived in this specific case, is according to Aristotle the
impossible conclusion, not, say, the sentence ‘if the diagonal is commensurable,
then even and uneven numbers are equal’ (cf. 41a24, 28, 32f.; 50a31). And
according to Aristotle’s own description in B 14 (62b32–35), a reductio presup-
poses two premises that are ‘admitted’ or recognized as true, namely one
premise of the syllogism and the opposite of the conclusion—both of these
are, of course, unproven within the argument. One thus arrives, at best, at the
resigned judgement of Martha Kneale: ‘ “hypothesis” seems here to be his
general name for a statement which plays an essential rôle in an argument
but not as any part of a syllogism nor yet as the conditional formulation of a
syllogism.’¹³
(C) On the third interpretation, which was first proposed by Pacius and has
been defended more recently by Maier and Ross,¹⁴ the hypothesis is not one of the

¹² Ch. Sigwart, Beiträge zur Lehre vom hypothetischen Urteil, Tübingen 1871; Shorey, op. cit., 461;
M. Mignucci, Aristotele, Gli Analitici Primi a cura di M.M., Naples 1969, 424.
¹³ Op. cit., 99.
¹⁴ Cf. I. Pacius, In Porphyrii Isagogen et Aristotelis Organum Commentarius Analyticus, Frankfurt
1597 (reprint Hildesheim 1966), 153; H. Maier, Die Syllogistik des Aristoteles, Bd. II 1, Tübingen
1900, 236ff.; W.D. Ross, commentary on ‘Aristotle’s Prior and Posterior Analytics’, Oxford 1949, 371f.
    ‘    ’ 9

premises, but rather the logical law that allows one to proceed from ascertaining
the conclusion’s falsity to accepting the demonstrandum. This interpretation also
has textual support: Aristotle says twice in A 23 that the initial thesis is inferred
from a hypothesis ‘when’ (ὅταν) or ‘since’ (ἐπεί) its contradictory gives rise to
something impossible. Ross understands this as an—admittedly inexact—
formulation of the hypothesis, which, spelled out more fully, would say: ‘If from
a proposition something impossible follows, this proposition is false and its
contradictory true.’¹⁵ One need not necessarily take this formulation as an explicit
description of the hypothesis—Ebbinghaus¹⁶ draws attention to the ὅταν, which
seems to suggest that an inference ἐξ ὑποθέσεως is always possible when some-
thing impossible follows, without taking the hypothesis itself to be identified.
These passages are, therefore, not conclusive evidence for this position. Whether,
like Ross, one ought to rely on A 44, according to which a reductio needs no
agreement—for the reason, according to Ross, that the hypothesis is a logical
law—is questionable, for what is there characterized as evident is, of course, not
the logical law, but rather the falsity of the conclusion. On the one hand it does seem
plausible to ground the difference between the two types of hypothetical arguments
in the fact that the transition to the demonstrandum is in one case made possible by a
logical law and in the other by a conditional premise that requires agreement; we
can, however, not be sure that Aristotle meant to claim this. Furthermore, on this
interpretation we run into the difficulty, mentioned also by Ebbinghaus, that the
hypothesis in the reductio is not a part of the argument, but rather a rule which
justifies one of the steps of the proof, whereas in the other case it seems to be a
premise.
This last point brings us back to the arguments of the first type. So far we have
been assuming that these are arguments in modus ponendo ponens. This is
certainly true in view of the example given by Aristotle, but that does not yet
show that Aristotle himself would have described the argument in this way. Now,
it is well known that Aristotle neither explicitly recognized the conditional as a
form of statement nor specifically investigated it. On the contrary: in the first
chapter of the An.Pr. he seems to assume that all premises must be categorical

¹⁵ G. Patzig (Die aristotelische Syllogistik, 3. Aufl., Göttingen 1969, 158, 165) proposes that the law of
excluded middle is the hypothesis. He does so by viewing the syllogism that leads to the impossible as a
proof of the falsity of the first hypothesis (on this see e.g. B 11.62a14–15: δειχθέντος ὅτι οὐχ ἡ ἀπόφασις
ἀνάνγκη τὴν κατάφασιν ἀληθεύεσθαι). Aristotle himself does, in fact, say (An.Po. A 11.77a22; cf. An.Pr.
B 11.62a13) that the proof ad impossibile assumes (λαμβάνει) this law. The syllogism can, however,
prove the falsity of the first hypothesis only if the falsity of the conclusion is assumed—that is, the step
to the negation of the first hypothesis is, strictly speaking, no longer a step of the syllogistic part of the
proof. However, according to Aristotle’s own description, what is ‘shown’ by the syllogism is not the
falsity of the hypothesis but rather the false conclusion. At least in these two chapters the view of Ross
and Maier thus seems to me to be more plausible. Since Aristotle nowhere explicitly formulates the
hypothesis, it seems to me equally likely that he had not consciously distinguished between these two
views.
¹⁶ Op. cit., 43, n. 1.
10    

statements. At least as conclusions of proofs, conditional statements seem not

to be at issue if, as Aristotle holds, ‘every proof and every syllogism must show
something as belonging or not belonging’ (40b23–24). If, however, in
Aristotle’s view both the premises and the conclusions of all valid arguments
must be categorical statements, one needs to ask whether he viewed the
hypothesis of a hypothetical argument as a premise at all. In fact, some things
do indicate that he thought of the hypothesis differently: namely as an
agreement according to which one of two statements will count as proven
when the other is.¹⁷ Those hypotheses are indeed usually formulated as
conditional statements; in A 44, however, we also encounter the formulation
‘if it is proven that . . . , then . . . ’ (50a34–35; cf. also Top. A 18,108b16–17). The
expression τὸ μεταλαμβανόμενον in A 23 suggests something similar. The
μεταλαμβανόμενον is something that is substituted for something else—in the
case of hypothetical arguments, a statement that takes the place of the demon-
strandum. Underlying this seems to be the view that the proof of the substi-
tuted statement takes the place of the proof for the demonstrandum—in
contrast to the contemporary view according to which one would have to
say that the demonstrandum is derived or proven from the conditional premise
together with the ‘substituted’ statement.¹⁸

III

If one does not regard the hypothesis as a premise, then it can at least plausibly be
explained why Aristotle treats the two types of hypothetical syllogisms as a single

¹⁷ Frede (op. cit. 29f.) also arrives at this result when considering the question why Aristotle seems
by all appearances to have thought that only syllogistic arguments in the strict sense fall under his
definition of the syllogism. Cf. also I. Mueller, ‘Greek Mathematics and Greek Logic’, in J. Corcoran
(ed.), Ancient Logic . . . , op. cit. 56f.
¹⁸ In the view of some commentators, the μεταλαμβάνομενον in A 23 is a term that is substituted for
another, not a statement. The word μεταλαμβανόμενα in A 29.45b18 does indeed doubtless refer to the
added terms of the statement; in A 23, however, the juxtaposition of τὸ μεταλαμβανόμενον with τὸ ἐξ
ἀρχῆς seems to indicate that Alexander is correct when he assumes that we are here dealing with the
substituted statement. In any case, this makes no important difference to the interpretation of this
passage. A further difficulty seems to me to stem from the commentators themselves—including
Alexander: They assume that the words ἐν ἅπασι γὰρ . . . κτλ. (41a38) refer to all hypothetical argu-
ments, thus also to reductio, and then they attempt to show how the syllogism of an apagogical
argument also relates to a substitute. It is, however, equally natural to read ἐν ἅπασι as referring only
to the ‘remaining’ hypothetical arguments. To support this reading further one may appeal to the place
in A 29 mentioned above. There Aristotle characterizes direct hypothetical arguments as κατὰ
μετάληψιν ἢ κατὰ ποιότητα, and if one follows the explanation of Alexander (324.22–24), μετάληψις
refers to the substitution of one statement for another. This seems essential to the ‘remaining’
arguments, since Aristotle adds that one must search for the premises of these arguments in view of
the substituted terms. The fact that Aristotle describes both the statement as well as the terms as
μεταλαμβανόμενα is hardly surprising, since additional terms are, of course, introduced with an
additional statement. In a reductio ad impossibile, on the other hand, one requires no
μεταλαμβανόμενα, because the three syllogistic terms are all one needs.
    ‘    ’ 11

class: The formal commonality consists in the fact that, in both the apagogical
arguments as well as in the others, a non-syllogistic rule of deduction is employed.
That is, there is a step that is justified neither by conversion nor by one of the
syllogistic moods. The distinction between the two types, on the other hand,
consists in the fact that the hypothesis corresponds to a logical rule in the case
of the reductio whereas it is a more or less ad hoc, non-logical rule that is agreed
upon in the case of the other hypothetical syllogisms—which was then rightly
treated as a premise by Aristotle’s successors. It is for just this reason that the
second type requires an explicit agreement.
This interpretation also explains the curious fact that Aristotle seems not to
take pure modus ponens arguments into account at all, but rather tacitly assumes
that a syllogism for the second premise—and only for this one—must be con-
structed. If one does not view the conditional hypothesis as a premise, then a pure
modus ponens argument has only a single premise, and therefore, as Aristotle once
again emphatically claims in A 23, cannot represent a complete argument. A proof
of the conditional premise, as suggested by Alexander and Philoponus, can, of
course, for this very reason not be at issue, since a conditional cannot appear as the
conclusion of an argument.
If Aristotle classified reductio ad impossibile as a hypothetical argument for the
reason just given, then what was important for him was less the logical status of
the hypothesis and more the type of logical step that it justifies. From the
standpoint of formal logic, it would otherwise be more natural to separate
formally valid methods of deduction—ostensive syllogisms and reductio ad
impossibile—from formally invalid ones and then, within the group of formally
valid arguments, to distinguish between direct and indirect deductions. For
ostensive deductions, which, in addition to the arguments of the syllogistic
moods, also include the direct conversion proofs of chapters A 4–6 (cf. A
7,29a30–34), what seems to be essential is that each step can be justified by
recourse to the underlying relations of belonging that hold among the terms of
the premises. For the evident validity of the perfect moods, Aristotle appeals to the
meaning of the expressions κατὰ παντός and κατὰ μηδενός respectively (cf. A
4.25b39–40; 26a24, 27). For the imperfect moods, the necessity is made evident
(cf. 24b24 and 27a17) by interposing statements between the premises and
conclusion that are ‘necessary because of the terms laid down’ (24a25; cf. 28a6).
This means, as one can see from the proofs of chapters A 4–6, that they are derived
using the conversion rules. Aristotle adds for completeness in A 5,28a6–7 that the
apagogical proofs use an additional ‘hypothesis’. These proofs are, however, not
ostensive. For ostensive proofs it must, therefore, be the case that they only
contain such statements as intermediate steps that are necessary ‘because of the
terms’. Now, the conversion rules explicitly refer to specific relations among
terms, and the perfect moods are also based on these relations, as we have seen.
A difficulty for this view seems, however, to arise from the fact that some of the
12    

imperfect moods—Baroco in the second figure and Bocardo in the third—cannot

be proven by means of conversion. Both are proven by Aristotle using reductio ad
impossibile. In these cases one cannot say that the deduction uses only sentences
that are ‘necessary because of the terms’. To put it another way, the proofs of these
moods do show that the conclusion follows from the premises, but they do not
show that the conclusion follows simply ‘on the basis of the terms assumed’.
Strictly speaking, one should not therefore, it seems, consider these moods to be
ostensive. Aristotle, however, clearly treats all syllogisms in the narrow sense as
ostensive arguments.
Alexander (24, 12–18) and Philoponus (37, 16–38, 29), who both address this
question, try to make do with the observation that the additional hypothesis only
serves to derive the impossibility, not the actual conclusion. But this hardly helps,
since it shows precisely that the conclusion itself is, in these cases, derived ἐξ
ὑποθέσεως. Elsewhere (318, 17ff.), Alexander entertains the possibility that, after
one has proven the validity of a mood by reductio, one can then use this mood as
an ostensive argument: ἢ καὶ ἐπὶ τούτων εἴη ἂν μετὰ τὴν πίστιν τὴν διὰ τῆς εἰς
ἀδύνατον ἀπαγωγῆς τοῦ συλλογιστικὴν εἶναι τὴν συμπλοκὴν καὶ ἐκ τῶν κειμένων
προτάσεων δεικτικῶς δεικνύμενον τὸ προκείμενον—that is, Aristotle could have
understood δεικτικῶς in the sense of ‘direct’. As noted at the beginning, Aristotle
in fact uses the expressions δεικτικῶς, δεικτικός etc. in some places to contrast
with διὰ τοῦ αδυνάτου, and thus presumably in the sense of ‘direct’, while in other
places he uses it as a contrast to ἐξ ὑποθέσεως. It is, therefore, possible that
syllogisms might be viewed as ostensive in the first instance because they are
direct arguments. That, however, would not justify Aristotle in viewing them all as
ostensive rather than hypothetical, for if the validity of some moods can only be
shown through a hypothetical argument, this would mean that the conclusion in
these cases could not be derived solely by recourse to the relations assumed to hold
between the terms. Aristotle does not address this problem himself, but he does
mention in connection with the apagogical proof of Bocardo (A 6.28b20) that one
can also carry out the proof without (!) reductio ad impossibile—namely by
ἔκθεσις. The same holds of Baroco, although Aristotle does not actually say so.
If one counts ἔκθεσις as an ostensive deductive step, then there is indeed an
ostensive proof for every syllogistic mood within Aristotle’s system, and
Aristotle is justified in treating every syllogism in the strict sense as an ostensive
argument.
That ἔκθεσις is an ostensive deductive step in the relevant sense is fairly clear,
since the ‘setting out’ is grounded in the relation between terms contained in one
of the premises. It makes no difference whether the thing ‘set out’ is to be
understood as a term or as an individual constant.¹⁹

¹⁹ On ἔκθεσις see Patzig, op. cit. 166ff. and W. Wieland, ‘Review of Patzig, Die aristotelische
Syllogistik’, Philos. Rundschau 14, 1966, 24f.
    ‘    ’ 13

One can also not object that in this procedure a term is added ‘from outside’
(ἔξωθεν, 24b21), since what we have here is a term or individual that is subordin-
ated to one of the given terms. The fact that ἔκθεσις can also appear in an indirect
proof (as in the proof for the conversion of universal negative premises,
A 2,25a14–17) also does not provide evidence against this view, since by ἔκθεσις
Aristotle, of course, does not mean a procedure by means of which one might
prove a given proposition, but rather—just as in the case of conversion—merely
an operation that is applied to a particular premise. Neither conversion nor
ἔκθεσις, therefore, count as deductive procedures on the same level as the syllo-
gistic moods or reductio—on their own, they cannot be the basis for a complete
argument. It is thus, incidentally, also entirely correct for Aristotle not to mention
ἔκθεσις in his considerations regarding completeness in A 23.
While, therefore, each step in an ostensive deduction can be justified by
recourse to the relations among terms, it is characteristic of the hypothetical
part of the reductio as well as of the remaining hypothetical arguments that the
internal structure of the premises plays no role.
The principle that a statement is false if something absurd follows from it
clearly does not refer to any particular type of statement, and a conditional may
contain arbitrary statements as its substatements. This suggests that Aristotle
speaks of a hypothesis, rather than a syllogism in the stricter sense, precisely
when a deductive step needs to be justified by recourse to the relations of
implication between statements (of any sort) rather than by taking into account
the term relations at hand. The combining of reductio with direct hypothetical
arguments seems to show that the decisive point for characterizing a step of a
proof as hypothetical is that the implication is independent of the relations among
the terms. One could thus describe hypothetical arguments as those that employ a
rule that is not grounded in the relations among its terms.
This is not to say that Aristotle intended to characterize this kind of non-
syllogistic rule explicitly with the word ὑπόθεσις. In order to claim this, we would
need to assume that he had consciously distinguished between premises and rules
of deduction, and that seems rather doubtful.²⁰ Nevertheless, it can, I think, be
shown that Aristotle does not treat the hypothesis as a premise, and in actual fact
the hypothesis is best treated as a rule.
We pointed out above that Aristotle probably took over the expression ἐξ
ὑποθέσεως from the Academy. This label could hardly have originally referred
to what Aristotle views as the relevant formal trait, for the simple reason that
Aristotelian formal syllogistic did not yet exist. The later way of talking about
hypothetical syllogisms shows that the characterization was taken to refer to the
if–then form of the hypothesis. This is not a possible interpretation of Aristotle

²⁰ See, for example, An.Po. A 10,76b10–22 with the commentary of J. Barnes (Aristotle’s Posterior
Analytics, Oxford 1975), 135, 138f.
14    

since his reductio arguments, at least, do not contain a conditional premise. Now,
rules of inference are often formulated as if–then sentences, but this also applies to
rules by which an inference is made in an ostensive syllogism, and therefore this
cannot be viewed as the defining characteristic of hypothetical arguments either. It
seems that what is important for Aristotle is that one goes from one proposition to
another ‘on the basis of a hypothesis’ without taking into account the internal
structure of these propositions. The status of this kind of convention is not
something that Aristotle seems to have investigated.

IV

In view of the completeness claim of chapter A 23, we still need to inquire into the
extent to which Aristotle’s formal analysis is applicable to the sorts of arguments
that he wants it to encompass. For reductio arguments it is easy to see that
Aristotle’s description only applies to a certain type—namely syllogisms διὰ τοῦ
ἀδυνάτου as described in B 11–14 and the indirect proofs of A 5 and 6. Even if we
were to grant that every step of the proof of the incommensurability of the
diagonal, which Aristotle uses as an example, could be put in syllogistic form,
this argument, which seems to lead to an explicit contradiction, would still need to
consist of at least two syllogisms and a hypothetical step. Aristotle, however,
formulates the conclusion so imprecisely—‘odd numbers will be equal to even
numbers’—that one is rather inclined to conjecture that he did not even try to
apply his analysis to the mathematician’s incommensurability proof.²¹
Direct hypothetical arguments are, according to the passages discussed here,
characterized not only by the fact that they make use of an ‘agreement’, but also by
the fact that within them one proposition is substituted for another and proven, so
to speak, in its place. The expression μεταλαμβανόμενον suggests that the substi-
tuted proposition is meant to be equivalent to the proposition whose place it takes,
which would imply that it could also be refuted in place of the original thesis. This
is precisely what seems to be the case with the example in Top. Γ 6, 119b35, and
also in A 44, where the hypothesis is given once in the form ‘if not p, then not q’
and a second time in the form ‘if p, then q’ (cf. 50a19–20 and 34–35 respectively).
However, this does not seem to hold true in all cases, for, as Aristotle himself notes
(Top. Γ 6.119b21–24), arguments ἀπὸ τοῦ ἧττον (‘from the less so’) can be used
‘only for proving, not for refuting’—that is, the implication here holds only in one
direction.²²

²¹ Cf. Mueller, op. cit., 56.

²² Alexander of Aphrodisias claims that arguments ἀπὸ τοῦ μᾶλλον καὶ τοῦ ἧττον καὶ ὁμοίου are
hypothetical (265.30ff., 324.19f.). He identifies them with the arguments κατὰ ποιότητα in A 29.45b17.
I see no reason to doubt Alexander’s interpretation, especially since Aristotle himself describes
arguments ἀπὸ τοῦ ὁμοίου as ἐξ ὑποθέσεως in Top. A 18,108b12ff.
    ‘    ’ 15

The few remaining examples of this type of argument, which Aristotle explicitly
characterizes as hypothetical (Top. A 18.108b12ff.; Γ 6,119b35ff.; An.Po. B 6,92a7ff.,
20ff.), suffice to show that the formal analysis indicated represents a simplification,
stemming primarily from the fact that Aristotle in each case only seems to be taking
into consideration one part of the complete argument. This was noted above regard-
ing the example from A 44. The other four cases can be characterized generally as
follows: In each case Aristotle proceeds from a universal premise about a particular
domain—for example, the domain of similar things, or pairs of opposites, or even of
the class of objects to which the subject of the thesis in question belongs.²³ By using
this premise the conditional proposition is then derived, which seems to play the role
of a hypothesis. Now, if the antecedent of this conditional is proven, then the
consequent must also count as proven. Clearly, Aristotle’s analysis can only be
applied to the part of the argument that begins with the conditional ‘hypothesis’. In
no case is a proof offered for its antecedent—according to A 23 and 44 this would
have to be a syllogism, but this is, as the example of A 44 has already shown, highly
dubious. According to the description of Top. A 18, an argument about ‘similar
things’ would need to look something like this:

(What holds of one of two similar things, holds of the other. a and b are similar to
each other)
If a is F, then b is also F. (hypothesis)
P₁
P₂
a is F (μεταλαμβανόμενον; proven using a syllogism from P₁, P₂)
(Conclusion) b is F (from the hypothesis)

While the text of this passage seems to emphasize that the hypothesis is the
conditional derived from the universal premise, Aristotle speaks in other cases
as if the universal premise were itself the hypothesis. In An.Po. B 6,92a20ff., the
part of the argument bracketed above is explicitly expounded; this argument
proceeds as follows:

(H) If A, B and C, D are opposites and A is defined by C, then B is defined by D.

1) a, b, and c, d are opposites
2) a is defined by c
(C) b is defined by d.

²³ Cf. Top. A 18.108b13: ὥς ποτε ἐφ’ ἐνὸς τῶν ὁμοίων ἒχει, οὕτως καὶ ἐπὶ τῶν λοιπῶν; Top. Γ
6.119b35: ὁμοίως ἀξιώσαντα, εἰ ἑνί, καὶ πᾶσιν ὑπάρχειν ἢ μὴ ὑπάρχειν; An.Po. B 6.92a7: λαβόντα τὸ μὲν
τί ἧν εἶναι τό ἐκ τῶν ἐν τῷ τί ἐστιν ἴδιον; 92a21: τὸ δ’ ἐναντίῳ τὸ τῷ ἐναντίῳ <ἐναντίῳ> εἶναι.
16    

If one tries to apply the schema from A 44 to this argument, then one must
presumably follow Maier²⁴ in viewing 2) as the μεταλαμβανόμενον. The hypoth-
esis, which here, however, does not appear explicitly, would need to be: ‘If a is
defined by c, then b is defined by d’. Aristotle seems, however, in this case to view
the universal premise itself as the hypothesis.
It is possible that Aristotle did not view the transition from the general to the
particular hypothesis as a step of the derivation, but rather as a sort of contraction
of the universal principle to the particular case, in the way he does for the ‘general
principles’ of sciences (cf. An.Po. A 10,76a37–b2; A 11,77a22–24). Thus both the
universal principle as well as the specific implicative premise can be characterized
as (the) hypothesis. This is how, for example, Alexander (324,24–31; cf. 325,3–8)
proceeds when he first gives the universal principle as what is assumed—‘it is
assumed (ὑποτίθεται) that if what would seem to be more sufficient for happiness
is not sufficient, then neither would what is less so than it be sufficient’ (εἰ ὃ μᾶλλον
ἂν δόξαι αὔταρκες εἶναι πρὸς εὐδαιμονίαν, τοῦτο μή ἐστιν αὔταρκες, οὐδὲ τὸ ἧττον
ἐκείνου εἴη ἂν αὔταρκες’), but then later the derived conditional ‘that wealth is not
sufficient for happiness, given that even health is not, has been assumed
(ὑπόκειται)’ (324.29–30: τὸ μὲν μὴ εἶναι τὸν πλοῦτον αὐτάρκη πρὸς εὐδαιμονίαν,
εἴ γε μηδὲ ἡ ὑγίεια, ὑπόκειται) without so much as indicating the difference.²⁵
Now, it is doubtless no coincidence that this example represents one of the
typical forms of argument in the Topics.²⁶ The typical procedure of refutation by
modus tollens, where a thesis is refuted by showing one of its consequences to be
false, seems at first not to be captured by this sort of argument. However, one can,
of course, transform a modus tollens argument into an argument in modus ponens
by contraposition of the conditional premise; that is, instead of arguing ‘if p, then
q; not q; therefore, not p’, one can start with the premise ‘if not q, then not p’. We
may thus assume that Aristotle meant to bring together the arguments of the
Topics under the rubric of syllogisms ἐξ ὑποθέσεως, which he evidently realized
could not be put into syllogistic form. Although his analysis of these arguments
represents, as we have seen, a simplification, it nevertheless at least appears to
capture an important formal characteristic.
Given the types of arguments Aristotle explicitly recognized, the completeness
thesis of chapter An.Pr. A 23 is, of course, not justified. That Aristotle seems not to
have noticed this may be due to the inexactness of his treatment, which, as he
himself notes, is provisional; however, the more significant factor seems to be that

²⁴ Op. cit., 260, n. 1.

²⁵ Theophrastus seems to have distinguished things more clearly here in his definition of τόπος
(apud Al.Aphr. in Top. 5.21–26). He defines a τόπος as a principle or element from which we extract the
principles (i.e. premises?) about each kind of thing (ἔστι γὰρ ὁ τόπος, ὡς λέγει Θεόφραστος, ἀρχή τις ἢ
στοιχεῖον, ἀφ’ οῦ λαμβάνομεν τὰς περὶ ἕκαστον ἀρχάς ἐπιστήσαντες τὴν διάνοιαν, τῇ περιγραφῇ μὲν
ὡρισμένος . . . τοῖς δὲ καθ’ ἕκαστα ἀόριστος). Incidentally, the definition also seems to show that
Theophrastus did not view the universal principle as a part of the argument.
²⁶ See J. Brunschwig, Introduction to: Aristote, Topiques, vol. I, Paris 1967, XLff.
    ‘    ’ 17

he did not recognize the conditional as a special sort of statement, and therefore
treated the ‘hypothesis’ as an ‘agreement’ outside of the actual course of the proof
rather than as one of its premises—or, to put it differently, he only took relations
among terms into account as logical constants, not propositional connectives. In
this way he could arrive at the claim at the end of A 23 that even hypothetical
arguments are (dependent on) syllogisms insofar as each of them must at least
contain a syllogism.²⁷

²⁷ I am grateful to Professors G. Patzig and W. Carl for criticism and discussion.

 2
Necessity with Gaps
Aristotle on the Contingency of Natural Processes

In chapter A 13 of the Prior Analytics, Aristotle describes statements about what

happens for the most part or by nature as contingency statements. He explains that
in those cases the necessity has gaps, because natural processes may be interrupted.
A few lines later he claims that there may be demonstrative knowledge of such cases,
unlike mere coincidences. This clearly conflicts with his claim in the Posterior
Analytics that only necessary truths can be known in the strict sense and that the
premises of demonstrations must also be necessary. The apparent contradiction
could be avoided if statements about natural events were understood as conditionals
with an antecedent of the form ‘there is no obstacle’. But the limited syntax of
Aristotle’s syllogistic did not allow him to see those statements as anything but
contingently true.

Aristotle is the first philosopher to have given a precise definition of the

contingent and thus to have introduced this as an explicit philosophical
concept. His famous argument against determinism (de int. 9), with which
he sought to defend the contingency of future events, has given rise to a flood
of discussion and commentary since antiquity. By contrast, his formal
theory of contingency statements in the Analytica Priora (An.Pr.) has garnered
less attention—unsurprisingly, given that it seems to be an obvious failure
and was rejected by Aristotle’s own students. The following contribution
makes no attempt to render Aristotle’s modal syllogistic consistent and
plausible—such an attempt could hardly succeed. It does, however, aim to
show why Aristotle ever embarked on this difficult and ultimately unsuccessful
undertaking.
According to one of Aristotle’s well-known doctrines, there can be scientific
knowledge in two domains: Of things that always behave in the same way, and of
things that ‘for the most part’ (ὡς ἐπὶ τὸ πολύ) or ‘by nature’ (φύσει) behave in
the same way. There can be no scientific knowledge of things that occur ‘by
chance’ (ἀπὸ τύχης) (Met. E 2,1027a19–28; cf. An.Pr. A 13,32b18–22; Phys. B
5,197a19–20). This is because science refers to the universal and seeks out causes

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0002
    19

and explanations. Something that happens by sheer chance, like the fact that
someone stumbles on treasure while digging in his garden (Met. Δ 30,1025a6), can
receive no explanation that refers to just these occurrences under these descrip-
tions. One could perhaps explain why someone is digging in the garden (e.g. to
loosen soil for the plants) and also why there is treasure in this spot (because
someone else buried it there years ago so as to hide it from plundering soldiers).
Both, however, offer no general explanation for the fact that people find treasure
while digging in the garden.
The objects of most natural sciences belong to the domain of things that behave for
the most part, but not always, in the same way. In particular, this is true of biology.
The reason organisms do not always behave in the same way is clearly that the
development of plants and animals depends on all sorts of external factors in such a
way that unforeseeable disruptions may occur. Nevertheless, such disruptions are
the exception and not the rule, and thus one can find explanations in this domain that
apply to most cases even if not to all. So far, Aristotle’s theory is plausible and
uncontroversial, or so it seems to me. In the context of Aristotle’s philosophy of
science, however, this theory faces a difficulty. Namely, this theory stands in contra-
diction with the official doctrine of the Analytica Posteriora (An.Po.). According to
the first chapters of the An.Po., the premises and theorems of each science must be
universal necessity statements describing the objects of the particular science (the
underlying genus) and their per se properties (that is, the properties they have on
account of their definitions). Statements about what occurs ‘for the most part’ are,
however, neither necessary nor universal in Aristotle’s view. How is it that they can
nevertheless serve as scientific premises? What, for instance, could be the scientific
status of the statement that grapevines lose their leaves in autumn?
It may be that the striking contradiction arose in the course of the development
of Aristotle’s theory. One may hold this whether, like Solmsen¹, one views the
inclusion of natural sciences in his theory as a revolutionary innovation by a
renegade Platonist or whether, like Barnes², one views them as the remnants of an
earlier, more ‘liberal’ conception of science. This, of course, does not imply that
Aristotle never noticed the problem. The conjecture of Joachim³ that statements
that hold merely for the most part are to be reckoned as merely provisional and
must be replaced by necessity statements in a ‘real’ science is not supported by
Aristotle’s explicit pronouncements on the topic. It is also not straightforwardly
compatible with the text of the An.Po. There, Aristotle expressly says in at least
one place that ‘some principles (ἀρχαί) are necessary and others contingent’

¹ F. Solmsen, Die Entwicklung der aristotelischen Logik und Rhetorik, Berlin 1929,138–41.
² Proof and the Syllogism, in: E. Berti (ed.), Aristotle on Science: The “Posterior Analytics”, Padova
1981, 50ff.
³ H.H. Joachim, Aristotle on Coming-to-be and Passing-away, text, introduction and commentary by
H.H.J., Oxford U.P. 1922, Introduction, xxviif.
20    

(88b7). This, of course, contradicts the claim of the earlier chapter, but that can
hardly be explained by supposing that Aristotle did not intend to recognize ‘for
the most part’ statements as genuine scientific statements.
The most plausible explanation for the remark about the different kinds of
scientific principles is no doubt that Aristotle is referring here to his statement
regarding the status of natural scientific premises in An.Pr. A 13⁴. According to
that chapter, statements about what occurs ‘for the most part’ or ‘by nature’ are to
be counted among possibility statements in the strict sense of contingency state-
ments. In that same passage, Aristotle also reiterates his claim that what occurs for
the most part or by nature does admit of knowledge and even of syllogistic
demonstration. This is also the only place where Aristotle speaks about the logical
status of natural scientific generalizations. For these reasons, it seems plausible to
begin with this passage in asking whether, and how, Aristotle meant to classify
natural sciences as sciences in the strict sense, given the broader context of his
philosophy of science.
The passage in question (An.Pr. A 13,32a16–b23) occurs in the introduction to
Aristotle’s treatment of syllogisms with possibility premises. It begins with the
famous definition of the possible in the strict sense: ‘By “being possible” and “the
possible” I mean that which, while not being necessary, will not lead to anything
impossible when it is assumed to belong.’ With this definition, Aristotle commits
himself to a ‘two-sided’ conception of possibility for the sequel: What is possible,
in this sense, is what is neither necessarily so nor necessarily not so. The ‘one-
sided’ conception of possibility, which only rules out impossibility but includes
necessity, it explicitly laid aside. (In the following, I will adopt the convention that
‘contingent’ always refers to two-sided possibility and I will use ‘possible’ for a
one-sided possibility, except for quotations). The formulation of this definition
shows that Aristotle is here thinking of the possible holding of predicates. He is
defining the usage of ‘possible’ and ‘possibly’ in syllogistic statements of the form
‘A possibly belongs to every/some/not all/no B.’ Aristotle goes on to prove that
there are special conversion rules for contingency statements thus defined, which
apply neither to necessity statements nor to assertoric ones: Statements of the
form ‘A possibly belongs to every B’ and ‘A possibly belongs to no B’ are, due to
the definition of possibility, equivalent, and likewise for statements of the form ‘A
possibly belongs to some B’ and ‘A possibly does not belong to some B’. Since
Ross⁵, these rules have been referred to as the rules of complementary conversion.
After the proof of complementary conversion, there then follows a remark
concerning the domains in which contingency statements are employed
(32b4–23):

⁴ Cf. J. Barnes, Aristotle’s Theory of Demonstration’, Phronesis 14, 1969, 134,n.51.

⁵ In his commentary in W. D. Ross, ‘Aristotle’s Prior and Posterior Analytics’, ed. maior w.
commentary, Oxford, Clarendon Press 1949.
    21

After these explanations, let us add that ‘being possible’ is said in two ways: in
one way of what happens for the most part, when the necessity has gaps, such as
that a man turns grey or grows or ages, or generally what belongs by nature. For
this has no continuous necessity because a man does not exist forever, but while a
man exists, it happens either of necessity or for the most part. In another way
‘being possible’ is said of what is indeterminate, that is, what is possible both this
way and not this way, such as that an animal walks or that an earthquake
happens while it walks, or, generally, what comes about by chance, for this is
by nature no more this way than the opposite way. Both these kinds of being
possible also convert with respect to opposite premises, but not in the same way.
Rather, what is so by nature converts because it does not belong of necessity (for
in this way it is possible for a man not to turn grey), while the indeterminate
converts because it is no more this way than that.
There is no knowledge or demonstrative syllogism of indeterminate things
because the middle term is irregular, but there is knowledge of things that happen
by nature, and by and large arguments and investigations are concerned with
what is possible in this way. For the other sort a syllogism may come about, but
one does not usually try to find one.
These things will be explained more precisely later.

As the last sentence shows, this is a provisional note; Aristotle promises to return
to the problem. The promise is, however, fulfilled neither in the Analytics nor in
any other passage of Aristotle’s preserved corpus. For the question of the logical
status of statements about what takes place ‘by nature’ or ‘for the most part’, we
are thus confined to the terse remarks of this passage.
Aristotle’s students and the ancient commentators evidently took Aristotle to
be drawing a distinction between two meanings of ‘possible’ (contingent) in these
clarificatory remarks, corresponding to the expressions ‘for the most part’ or ‘by
nature’ and ‘by chance’ respectively. Accordingly, statements in which these
expressions occur are contingency statements, and the same rules apply to them
as do to explicit contingency statements. On this interpretation, the subsequent
chapters, in which the theory of inferences with contingency premises is devel-
oped, are of scientific interest because they at once deliver a theory of inferences
with ‘for the most part’ premises, which are needed in the natural sciences.
This view has been defended more recently by Jonathan Barnes⁶; it appears,
however, as Barnes himself concedes, to misfire from the outset on account of an
obvious difficulty: Statements of the form ‘B’s are mostly A’ cannot in general be
converted using the rule of complementary conversion, since from ‘B’s are mostly

⁶ Sheep Have Four Legs, Proceedings of the World Congress on Aristotle vol. III, Athens 1982,
113–119.
22    

A’ it clearly does not follow that ‘B’s are mostly not A’. Theophrastus and
Eudemus are said to have rejected complementary conversion for this reason
(Ps.—Amm. in An.Pr. 45.42–46.2). This shows, on the one hand, that they thought
the syllogistic of contingency statements was meant to apply to premises of
natural sciences; on the other hand, it shows that they found the conception of
possibility developed by Aristotle unfit for this purpose. We need not concern
ourselves here further with the alternatives that Theophrastus and Eudemus
developed.
It is hard to believe that Aristotle himself would have overlooked such an
obvious objection, especially since he explicitly considers the question of comple-
mentary conversion. Nevertheless, as Becker (op. cit. 77) emphasized, the expres-
sions ‘for the most part’ and ‘by nature’ do not occur again in the following
chapters, while complementary conversion is used constantly. Becker inferred
from this observation that the remark about the premises of natural sciences
(32b18–22) presumably did not originate with Aristotle himself. Rather,
Aristotle’s preceding note concerning complementary conversion was intended
precisely to draw our attention to the fact that the syllogistic of contingency
statements is not suited to function as a logic of ‘for the most part’ statements.
This note could, however, have been inserted later, when Aristotle—perhaps
prompted by the objections of his students—noticed the discrepancy. A later
editor, according to Becker, failed to understand this and rewrote the text,
resulting in the version that we have received.
Needless to say, conjectures of inauthenticity are, particularly in the case of
difficult texts, practically always a last resort. In this case, such a conjecture does
not seem to me to be necessary. One can understand Aristotle’s thesis concerning
the different usages of ‘possibly’ in a different way, as Hintikka⁷ has shown—not
in the sense that ‘possibly’ has the same meaning as ‘for the most part’ or ‘by
chance’, but rather as saying that statements of contingency hold in two types of
case—when something happens ‘for the most part’ or ‘by nature’, and when
something occurs by sheer chance. A statement of the form ‘all B’s are possibly
A’ will then be true both when B’s are for the most part or by nature A, and when
they are coincidentally A, without needing to claim that such contingency state-
ments state the same thing as the statements ‘B’s are mostly A’ or ‘all B’s are
coincidentally A’. This interpretation is, in substance at least, more plausible than
the first one, and it is able to explain Aristotle’s remark about complementary
conversion in the passage in question (32b13–18): It is not being claimed there
that statements in which the expression ‘for the most part’ occurs can be con-
verted (which would obviously be untenable). Rather, all that is being claimed is
that, in both of the aforementioned sorts of case, ‘all B’s are possibly A’ as well as

⁷ J. Hintikka, ‘Aristotle’s Different Possibilities’, in: Hintikka, Time and Necessity, Oxford,
Clarendon Press, 1973, 34f.
    23

the equivalent ‘all B’s are possibly not-A’ holds, even if not for the same reason. In
the first case, if a predicate applies by nature or for the most part, the conversion
rule applies because it is not necessary that all B’s are A; in the second case, it
applies because B’s could just as well be A as not-A.
Even supposing this interpretation to be correct, there remains the question of
whether Aristotle considered his possibility-syllogistic to be also a formal theory of
scientific proof. Granting that Aristotle recognized expressions like ‘for the most
part’ and ‘by nature’ do not mean the same as ‘possibly’, he could still have
claimed that whenever we have ‘for the most part’ statements as premises, a
syllogistic deduction taking the form of a syllogism from contingency premises
can be constructed. This is, however, not a thesis with broad ramifications. It says
merely that a contingency conclusion, at least, can be derived from ‘for the most
part’ premises, and it remains unclear whether this conclusion can likewise be
treated as a ‘for the most part’ statement. One could equally say that from
necessity premises an assertoric conclusion can be deduced, because the necessary
belongs to the domain of things that are in fact the case. It requires a further
argument to establish the more important claim that a necessity follows as a
conclusion from necessity premises.
In the passage we are considering, however, it is precisely the possibility of a
syllogistic deduction that seems to be at issue for Aristotle. One requirement for
scientific knowledge—perhaps the most important one—is that one has a proof of
what one claims to know. According to Aristotle, this proof must take the form of
a syllogistic deduction. It is indeed a further condition that the premises must be
necessity statements. Yet if one has only ‘for the most part’ statements available, as
is plainly supposed to be the case in many natural sciences, it remains nevertheless
important to establish that a syllogistic derivation is at least possible. One may
then go on to ask whether from ‘for the most part’ premises ‘for the most part’
conclusions follow. In the end, Aristotle seems to suggest that in such cases there
perhaps is a kind of necessity after all, even if not a continuous one. There is,
rather, an ‘interrupted’ (32b6) necessity.⁸ I will return to this point below.
It seems to me that there are indeed indications that Aristotle was attempting to
demonstrate the possibility of syllogistic deductions in the domain of the contin-
gent, and thus also in the domain of natural science. That Aristotle upheld the
validity of syllogisms with contingency premises at all is at least as astonishing as
the thesis that the premises of natural sciences are contingency statements. In a
way, the former is actually much more surprising than the latter. Aristotle’s claim

⁸ Jonathan Barnes has drawn my attention to the fact that Aristotle’s expression can also be
understood in such a way that an ‘interrupted’ necessity is no necessity at all, just like a cancelled
race is not a race of a special kind, but rather no race. Two things, I think, speak against this: First, that
Aristotle immediately goes on to speak of necessity again with regard to grey hair and signs of age.
Second, that certain cases of natural necessity, which Aristotle himself characterizes in this way, can in
fact hardly be understood in any other way than as cases of necessity with gaps (see below).
24    

that natural science makes use of contingency statements as premises stands in

contradiction only with his own theory, a view which he might well have softened
in response to the counterexample of natural science. By contrast, the assumption
that one can syllogistically derive any conclusion at all from two contingency
premises is in substance difficult to justify and seems, on first sight, to be entirely
implausible. It seems almost easier to find counterexamples to the moods Aristotle
accepts as valid than it is to find counterexamples to the moods whose invalidity is
not immediately obvious⁹. And so McCall¹⁰ (88) laconically explains that Aristotle
overlooked the fact that when two statements are each individually possible, it
does not follow that their conjunction is also possible. This impression can hardly
be avoided so long as one sticks to examples that are also contingency statements
in the contemporary sense. If, however, one considers examples in which both
premises are plausible statements of the form ‘all B’s are by nature A’, then
counterexamples no longer seem to arise. Now, at the end of An.Pr. chapter
A 13, Aristotle introduces an interpretation of universal contingency statements
under which, as Becker (op.cit. 32–37) has shown¹¹, the moods of the first figure
recognized as valid do in fact come out as formally valid inferences. Aristotle
claims that one can understand a statement of the form ‘A possibly belongs to
each B’ in such a way that it is not only claimed that A holds possibly of everything
that is B, but rather that A holds possibly of everything that can be B (32b23–32).
The assumption that the predicate term (possibly) covers everything possibly
falling under the subject term may have to do with the fact that Aristotle was
thinking of cases in which the one applies to the other by nature, even if not
necessarily. That is, if B’s are taken to be A by nature, it is natural to assume that
nothing can be B unless it at least could be A. The difference in meaning between
‘possibly’ and ‘by nature’ nevertheless remains intact, since if, for example, it
should be the case that all B’s are possibly A (because they are A by nature), one
can nevertheless still say that all B’s are possibly not-A—not, however, that they
are not A by nature.
A similar interpretative postulate to the one at the end of chapter A 13 is found
in An.Pr. A 15 (34b7–18). There, Aristotle deals with a counterexample to the
mood¹² AaB, QBaC > QAaC: One might have thought that the clearly invalid
argument ‘everything that moves is human; every horse is possibly moving;
therefore, every horse is possibly human’ was an instance of this mood. As with

⁹ Cf. W. and M. Kneale, The Development of Logic, Oxford U.P. 1962, 87–9.
¹⁰ S. McCall, Aristotle’s Modal Syllogisms, Amsterdam 1963, 88.
¹¹ See also J. Hintikka, ‘Aristotle’s Different Possibilities’, in: Hintikka, Time and Necessity, Oxford,
Clarendon Press, 1973, 38–40.
¹² In what follows, I use the notation which has now become standard for syllogistic statements:
A and B stand for the predicate and subject term respectively, a, e, i, and o respectively for the
expressions ‘belongs to all’, ‘belongs to no’, ‘belongs to some’, and ‘does not belong to some’. AxB
stands for an arbitrary statement of one of these four forms. The modal operators are: N (necessarily),
Q (contingently) and M (possibly).
    25

many of Aristotle’s examples of terms, this argument contains a premise that is

not obviously true in a time-independent way, but which, rather, might hold at a
particular time—specifically, in a situation in which it is only humans that move
(34b12). Aristotle now postulates that the assertoric premises should not be ‘with
a limitation of time’, and thus successfully excludes this and similar counterexam-
ples. He also claims that one normally uses only temporally unrestricted assertoric
premises in syllogisms (‘for it is through premises of this sort that we construct the
syllogisms’, 34b9). This surely does not apply to the examples of terms that
Aristotle himself constantly uses in the Analytics; it is more comprehensible if
he is at this point thinking of a deduction in the context of a science, where
temporally limited statements are indeed not permitted.
In both A 13 and in A 15 one gets the impression that Aristotle has imposed a
rule of interpretation after the fact in order to secure his theory against natural
objections.¹³ In both cases, the added assumptions are intelligible if one supposes
that Aristotle is thinking of premises of natural scientific arguments.
Now, if, as we are assuming, Aristotle was clear that ‘for the most part’ and ‘by
nature’ do not have the same meaning as ‘possibly’, and if it turned out to be so
difficult to verify syllogisms with contingency premises as valid, why did he then
not attempt to construct a syllogistic of ‘for the most part’ or ‘by nature’ state-
ments? The answer to this question is, I think: Because these statements cannot, or
can only with great difficulty, be brought into the form of syllogistic premises.
Treating them as contingency statements probably seemed to Aristotle to be the
only possibility for reducing them to a system of modal syllogisms. This could, in
any case, be justified by the fact that they imply the corresponding contingency
statements.
The syllogistic requires statements in the four standard forms AaB, AeB, AiB,
and AoB, with or without additional modal operators; at least one universal
premise is required for each valid syllogism. Since Aristotle seemed in the first
instance to envision scientific syllogisms only in the mood of Barbara, we can here
restrict ourselves to the question of how one can formulate a ‘for the most part’ or
‘by nature’ statement as a universal affirmative statement of the form AaB. Here
I will treat the two expressions ‘for the most part’ and ‘by nature’, which are, of
course, not synonymous, separately. For the sake of brevity, I will use the letters Ω
and Φ for ὡς ἐπὶ τὸ πολύ and φύσει respectively.
It is not clear to what extent Aristotle himself took the difference of meaning
into account in the Analytics. At A 3, 24b14 the two expressions are joined with an

¹³ After the fact: It would lead us too far astray to do so here, but it can be shown that Aristotle does
not presuppose or take into account his postulates at the outset. This is especially clear, of course, in the
case of his examples of terms. As regards the interpretation of universal contingency statements, it can
be shown that they do indeed justify the moods treated in Α 14, but that other parts of the contingency
syllogistic must then be invalid if this interpretation is to be taken across the board. On this see Becker,
op. cit. 32–7, 57–9.
26    

‘and’ that is probably to be understood in an explicative sense; at A 13, 32b7

Aristotle summarizes his description of that which occurs for the most part with
the phrase ‘or generally what belongs by nature’ (ἢ ὅλως τὸ πεφυκὸς ὑπάρχειν).
This is not to say that Aristotle takes ‘for the most part’ and ‘by nature’ to be
synonyms; it rather seems to show that he is discussing ‘for the most part’
statements insofar as ‘by nature’ statements belong to this class. He must have
been primarily aiming to establish the latter such statements as premises of
scientific argument, with an eye to the natural sciences.
The expression ὡς ἐπὶ τὸ πολύ (for the most part) is, as with many of Aristotle’s
semi-technical expressions, used in various contexts that suggest different inter-
pretations¹⁴. In some places Ω is used as a kind of quantifier, like ‘all’ or ‘some’. It
is clear that this usage does not suffice for a syllogistic of such statements, since, on
this interpretation, one can at most obtain particular premises. Aristotle even
seems to assume in certain places that universal statements are logically ruled out
(Top. B 6,112b5–9; An.Po. II 12.96a10–11)¹⁵.
In many places Ω appears alongside ἀεί (always) as a quantifier over times;
these are above all the places mentioned at the opening of this essay, in which
Aristotle distinguishes what is amenable to science from pure chance, of which
there can be no scientific knowledge. This interpretation likewise fails to take
universal statements into account, since ‘B’s are for the most part A’ clearly does
not mean the same as ‘at most points in time, all B’s are A’s’, since the first
statement can be true even when there is no point in time at which all B’s are A,
and Aristotle seems even to assume himself that the B’s are never actually all A.¹⁶
Both quantitative usages occur together in the Rhetoric (A 2,1357a34–b1;
B 25,1402b13ff.) in order to clarify the concept of the probable (εἰκός): If most
B’s are, at most times, A, then it is likely that ‘the B’s’ are A. Even this, it seems to
me, does not deliver any kind of universal statement. It is telling that Aristotle
explains the concept of probability with reference to non-quantified statements of
the form ‘the B’s are A’. One can indeed assume that such statements are treated in
arguments like universal statements. Aristotle himself seems to confirm this when
he says that the probable is related to that with regard to which it is probable as the
universal is related to the particular (1357a36). Since, however, the quantitative
expressions ‘for the most part’ and ‘most’ are introduced as an explication of
‘probable’ and not vice-versa, we learn from this claim only that a probable

¹⁴ cf. Becker op.cit. 79–81; M.Mignucci, ὡς ἐπὶ τὸ πολύ et nécessaire dans la conception
Aristotélicienne de la science, in: E. Berti (ed.), Aristotle on Science: The ‘Posterior Analytics’, Padua
1981, 185ff.
¹⁵ On this point, cf. Becker, op.cit. 80; Barnes, Aristotle’s Posterior Analytics, translated with notes,
Clarendon Press, Oxford 1975, 229, Mignucci , op.cit.187–90, 196.
¹⁶ Mignucci, who has developed a formal model for these temporal expressions of Ω, does say that
such statements can include universal quantifiers but gives no examples. In passing, he indicates
himself that his model perhaps provides an analysis of the Aristotelian term εἰκός (‘probably’) rather
than φύσει. For natural science it is, of course, φύσει which is the central expression.
    27

universal statement must, strictly speaking, have the (non-syllogistic) form ‘most
B’s are at most times A’. For this reason, also, as Aristotle says, one cannot refute
such a probability premise simply by pointing to particular exceptions, but only by
attempting to show that either most B’s are not A, or that at most times some B’s
are not A, or both (i.e. that most B’s are at most times not A, 1402b35–1403a1). If,
therefore, one wished to introduce Ω as a temporal modal operator, one would once
again only obtain particular premises, which would not be sufficient for science.
The only thing that can be said about all B’s in these cases seems in fact to be
that they are possibly A—even if this does not, strictly speaking, follow, since it is,
of course, possible that most B’s are A’s at most times while some are necessarily
not A. Aristotle’s apparent failure to take this case into account may, once again,
be connected with the fact that he is assuming that the B’s are A by nature. If
something that is meant to be A by nature is in fact not A, then one might indeed
assume that it at least could be A, thus that it is not necessarily not-A.
Aristotle’s usage of the expression ‘by nature’ still needs to be considered. The
first thing to observe is that Φ does not mean the same as Ω, although Aristotle’s
frequent juxtaposition of the two expressions can give this impression. This may
be seen, first, in the fact that Aristotle’s own conception of those things that occur
by nature also includes some things that happen by necessity and always in the
same way (cf. Phys. Β 8.198b36, GC 333b5; EE Θ 2.1247a32–35; Rhet. Α
10.1369a34–b2). Second, it is not sufficient, and perhaps not even necessary,
that something occur in most cases for it to be the case by nature. From the fact
that most people are bad, for example, it does not follow that they are naturally
bad; and one could also claim without any obvious contradiction that people are
naturally good, even though most of them are bad. This, in any case, was the view
of the Stoics, although perhaps not Aristotle’s.¹⁷ Since, however, disruptions of the
normal course of events are rather uncommon in nature—or at least, Aristotle
takes them to be—one may assume that what occurs for the most part and in most
cases is in accordance with nature. This, and no more, is what Aristotle is claiming
when he says, for example, ‘in all cases where it is not impossible for things to be
otherwise than they generally are but where they may so happen, still what is the
general rule is what is according to nature’ (GA Δ 8.777a18–20). One thus need
not assume that Aristotle equates the concept of what occurs by nature with the
concept of what is the case for the most part, or that he wished to define the former
by the latter. What he says can also be understood as an argument: Since that which
happens by nature does not happen always, but rather only for the most part, that
which is in accordance with nature belongs to the realm of the contingent.
The expression ‘by nature’ seems better suited to be formulated in syllogistic
premises than ‘for the most part’, since it can straightforwardly occur in universal

¹⁷ For the example, see Top B 6,112b11–12; for Aristotle’s own view on this point see EN II 1,
1103a18–26.
28    

statements, as one can see by looking at the famous opening sentence of the
Metaphysics—‘all humans by nature desire to know’. One could thus try to
construe Φ as a modal operator, which is what Aristotle himself suggests when
he contrasts what occurs ‘for the most part and by nature’ with what is necessary.
But does this new operator fit into the system of Aristotelian modalities? That
universal Φ-statements are necessary seems ruled out by the fact that they hold
only for the most part. They must also be weaker than assertoric universals, since
AaB does not follow from Φ(AaB), but at most only AiB does. In the modal
syllogistic, Aristotle seems, however, to use ‘necessarily’ as a primitive expression
in order to define the concept of the possible.¹⁸ One can certainly assume that
what occurs by nature is not impossible; but if one tries to define Φ using
‘necessary’, then Aristotle’s conception of the contingent does seem to suggest
itself: neither necessary nor impossible. The alternative, to introduce Φ as a new,
undefined modal operator, is something that Aristotle probably would not have
considered. It also would have proven to be a dead end if one intended to try to
combine premises of different modalities with one another, as Aristotle does in the
modal syllogistic.
Now, An.Pr. A 13 itself seems to show that Aristotle was not entirely satisfied
with this solution, since he indicates that there is indeed a type of necessity in the
realm of natural processes, albeit not a ‘continuous’ one. He speaks without
hesitation about necessity in nature in his natural scientific works, but it evidently
disturbs him that this necessity arises precisely where things behave mostly, but
not always, in the same way. Needless to say, one would expect that necessary
states of affairs, if they exist, are to be described in necessity statements. It may
thus be worth pursuing the question that Aristotle does not: Could Φ-statements
perhaps still be represented as N-statements?
One can perhaps ascertain what Aristotle means when he says that this
necessity is interrupted or has gaps on the basis of the examples he gives in An.
Pr. A 13 (examples which he does not, however, treat in any detail). That (all) men
go grey or grow up or get old is not necessary, since some die as children, or in any
case before they have a chance to develop grey hair. On the other hand, one might
think that men do necessarily grow up, and later show signs of aging if they live to
be sufficiently old. The fact that they get grey hair is, however, not necessary—they
might instead become bald—but it is nevertheless the case for the most part. As
one can see from the description ‘either by necessity or for the most part’, Aristotle
is introducing two different kinds of example, both of which are, however, cases of
interrupted necessity.
In the examples of growing and aging, the necessity arises because a condition is
satisfied, namely the condition that men live long enough. With his third example,

¹⁸ I will assume that ‘impossible’ is understood to be equivalent to ‘necessarily not’, as Aristotle

frequently does. Cf., for example, Met. Δ 12.1019b23–24.
    29

Aristotle seems to take into account a further caveat in view of the fact that some
people do not develop grey hair even if they live to be old. In this case one could
perhaps say that there is at least the necessity that men will either go grey or go
bald. That the necessity has gaps can be explained, seemingly, by the fact that it
does not extend to all specimens of the human species—some people do not grow
old because they die too young. And the gaps arise because certain conditions
must be satisfied that are not satisfied in all cases. This suggests that the statements
in which these gappy necessities are described should be formulated as condition-
als: For example, ‘all humans get old if they live long enough’, or, in symbols, (x)
(Hx > (Bx > Ax)). The appearance of a certain condition in Aristotle’s examples
might, at first, prompt the thought that one could produce a ‘continuous’ necessity
simply through a qualification of the subject term: All older men get grey hair or
show other signs of old age. That, however, does not give us any generally
applicable formula for the gappy necessities of natural processes, since the
appearance of natural alterations or the actualization of natural capacities, like
sightedness, are not always connected with determinately specifiable conditions
like being of a certain age. While humans are sighted by nature, some are blind
from birth. Aristotle’s usage in a different place does, however, offer a more
general formulation of what might be meant with ‘by nature’ (Phys. B 8.199b26,
cf. 199a10, b18; Θ 4.255b7, 10–12, 19–21): what is the case by nature always
happens in the same way ‘if nothing prevents it’ (ἂν μή τι ἐμποδίσῃ). If we use S as a
symbolic abbreviation for the antecedent of a conditional, then universal state-
ments in which the expression ‘by nature’ occurs can be formally represented as
S > AaB (if nothing prevents it, then all B’s are A). One might well supply a
statement of this form with a necessity operator, since it evidently holds in all
cases. The gappy necessity of natural processes can thus, it seems, be explained as
the constant necessity of certain conditional statements.
As is well known, however, conditional statements are not syllogistic premises,
and Aristotle also does not recognize them as statements of any other logical form.
In the syllogistic he assumes that all statements that occur in any logically valid
proof have the standard form AxB (with or without a modal operator) or can be
put in this form. With regard to universal premises of natural scientific arguments,
he could therefore only ask about the status of statements of the form (X)AaB, and
these are, even when the corresponding implicative statement holds, still neither
necessary nor impossible, and thus to all appearances contingent. In order to
achieve syllogistic form, one could, however, use the expression N(S> . . . ) as a
definition of a new modal operator Φ, since S is presumably considered to be a
constant. In this way Φ would be connected with the necessity operator.¹⁹ This
sort of introduction of Φ again, however, presupposes that conditional expressions

¹⁹ Jonathan Barnes made me aware of this possibility.

30    

are available, and in doing so goes beyond the formal resources of the Aristotelian
syllogistic. If one assumes the formal representation of Φ-statements that I have
suggested, then it is easy to show that Aristotle’s optimism regarding logical
derivation with these sorts of premises was indeed justified. A ‘syllogism’ of the
mood Barbara would, for instance, have the form N(S>AaB), N(S>BaC) N
(S>AaC). It is easy to show that this is a valid deductive schema using the
resources of modal propositional logic: From N(p>q), N(p>s), and N(q&s>r)
(Barbara), N(p>r) follows.²⁰ Since Φ-statements can now be conceived of as
necessity statements of a particular form, the difficulty with which we began
disappears. According to Aristotle’s theory of science there is, on the one hand,
scientific knowledge only of necessary states of affairs; on the other hand, there are
meant to be sciences whose initial premises are contingency premises. The
contradiction arises because Aristotle only considers statements of the form
AxB with one of the modal operators N, M, or Q as candidates for premises of
natural science. If AaB holds only with the qualification ‘by nature’, then, it seems,
the corresponding premises of natural sciences can only be understood as con-
tingency premises. The Φ-premises of natural sciences should, however, as closer
analysis shows, be represented not as contingent universal statements but rather as
necessity statements of the implicative form N(S>AaB). Aristotle would then not
have had to say that one could know that all sheep have four legs by nature, even
though this is a contingent fact; he could have said that all sheep normally have
four legs, and that one can know this because it is necessary. Whether a certain
sheep has four legs or whether all sheep have four legs remains, as Aristotle says,
contingent; but in the realm of contingent natural processes there can be science
because there is a sufficient number of normal cases. What is contingent in the
particular cases holds of normal cases necessarily.
The necessity that one refers to here—if one does want to follow Aristotle in
speaking about necessity—is evidently not a logical one. What Aristotle had in
mind when he spoke of necessity in nature must be inferred from his natural
scientific writings. If we leave aside the unproblematic case of unchanging astro-
nomical phenomena, there are two kinds of necessity to be distinguished. Both of
these, it seems to me, fall under the concept of necessity under normal circum-
stances as characterized above.
In his biological writings and in the Physics, Aristotle distinguishes between
‘hypothetical’ and ‘simple’ necessity.²¹ By ‘hypothetical’ necessity he understands

²⁰ If one omits the necessity operator, one can also justify syllogisms with one assertoric and one
Φ-premise, which Aristotle seems to defend in An.Pr. A 15. If, however, one represents Φ-premises as
statements of necessity, as seems to be Aristotle’s intention, then one requires a premise of necessity
rather than an assertoric premise—which corresponds, conversely, to Aristotle’s postulate that asser-
toric premises apply without temporal restriction.
²¹ Cf. Phys. B 9.199b34–35; PA I 1.642a34–36. The interpretation of this distinction is the subject of
much discussion; on this, see D.M. Balme, Aristotle’s De Partibus Animalium I and De Generatione
Animalium I, transl. with notes by D.M.B., Oxford 1972 ,76–84; U. Wolf, Möglichkeit und
    31

the necessity of certain means for a given end, in which case the means can be
characterized as necessary, not in themselves, but in view of this goal. For
example, in order for a house to come into being, there must necessarily be stones
and beams available. Likewise, it is hypothetically necessary that humans have
eyelids which protect their sensitive eyes from injury (PA II 13.657a25ff.). The
necessity arises from the fact that eyes fulfil a specific function, and not, for
example, that they consist of a certain material.
Aristotle speaks of ‘simple’ or ‘natural’ necessity, by contrast, in connection
with properties and types of behaviour that apply to certain materials like the
elements ‘according to their nature’, for example the downward movement of
stones (An.Po. B 11.95a2). This second type of necessity is subordinated to
hypothetical necessity in the case of the coming into being of organisms because
it can only be called upon to provide an explanation when a hypothetical necessity
already dictates that certain materials are present. Thus, one can say, for example,
that the front teeth of certain types of animals fall out owing to the constitution of
the jaw at certain times (GA V 8.789a8–b8); the fact that just this sort of material is
in this place must, however, be explained in turn on the basis of a hypothetical
necessity.
As this example shows, Aristotle uses the distinction between hypothetical and
‘simple’ or material necessity in order to prove the utility and indispensability of
teleological explanations in the natural sciences. This point will not interest us
further here. It seems to me that one can, however, show that both of the kinds of
natural necessity Aristotle describes presuppose the existence of normal condi-
tions. For hypothetical necessity this applies in the cases where something is
viewed not as the necessary condition for something coming into being or the
presence of a certain result, but rather as a necessary condition for something
functioning correctly. It is, for example, without a doubt unqualifiedly true that if
a person is to come into being or be present, flesh and bones must be present. One
can, however, not say without qualification that people necessarily have eyelids,
because as we all know there can—on account of developmental defects or
injury—be people without eyelids. The necessity is interrupted in these cases, as
Aristotle says; it seems, however, nevertheless reasonable to speak here of neces-
sity, since the gaps evidently arise from the interruption of a natural process that
can normally be taken for granted to an equal extent in all cases and that can be
explained as necessary for the complete achievement of a predetermined goal.

Notwendigkeit bei Aristoteles und heute, Munich 1979, 90-8; R. Sorabji, Necessity, Cause and Blame,
London 1980, ch. IX, 143–54; J. Cooper, Aristotle on Natural Teleology, in: M. Schofield/M.
C. Nussbaum (ed.), Language and Logos, Cambridge U.P. 1982,210 n. 8. My presentation relies on
the more recent study by Cooper, ‘Hypothetical Necessity’, in: A. Gotthelf (ed.), Aristotle on Nature and
Living Things, Mathesis Publications, Pittsburgh 1986, 151–167, in which he refines his earlier
argumentation.
32    

Sheep with three legs are indisputably still sheep; but they are missing something
which one may understandably view as necessary for healthy sheep.
Above all, the ‘simple’ or ‘natural’ necessity of the behaviour of elements seems
to be connected with normal conditions. Exceptions to naturally ‘necessary’
behaviour arise, for example, through the use of force: By their nature stones fall
to the ground and in this sense necessarily, but, with force, they may also be
moved upwards.²² Air necessarily expands when it warms, but one can prevent its
expansion with force if one encloses it in a sufficiently airtight vessel. In these
cases, too, it thus seems fitting to speak of necessity in light of normal conditions.
What Aristotle has to say about the necessity of natural processes thus seems to
me to confirm the proposed analysis of the type of statement in which something
is described as such and such ‘by nature’. Such a statement is a matter of necessity
with gaps; the gaps can, however, be represented by means of a conditional
statement without requiring us to treat universal statements about natural pro-
cesses as contingency statements. It thus becomes intelligible that natural pro-
cesses can rightly be described, on the one hand, as necessary, while on the other
hand, since they are liable to interruptions, they can be seen as contingent
processes in the individual case.
The section of the An.Pr. we began with concludes with the promise of a more
precise discussion that Aristotle presumably never wrote. This is one of many
places in the Analytics in which Aristotle, it seems, recognized that the framework
of the syllogistic was too narrow for the domain that it was apparently intended to
encompass—the realm of all formally valid arguments. A similarly unfulfilled
promise is to be found in chapter A 44, at the conclusion of his treatment of
syllogisms ‘from a hypothesis’, where Aristotle explicitly states that the arguments
he has just described cannot be put into syllogistic form (50b2–3). It is hardly
surprising that Aristotle was so impressed by his discovery of formal syllogistic—
truly a great discovery—that he expected more from it than it was capable of
delivering. But perhaps it speaks even more to the greatness of Aristotle that he
was not misled by syllogistic into giving up prior commitments for which he
thought he had good reasons. The thesis that there is scientific knowledge only of
necessary states of affairs is independent of Aristotle’s formal logic.²³ One should
thus—pace Barnes—also not count it among the formal shackles that the syllo-
gistic cast on a theory of science that was originally not geared towards any
particular formal logic (Barnes, op.cit. 1981, 58). The opposite in fact seems to
be the case: Aristotle did not renounce this thesis even though his modal syllogistic

²² ἀνάγκη—the same Greek word that is used for necessity. This is, however, a different sort of
necessity, as Aristotle notes (An.Po. B 11,94b37–95a2).
²³ For a defence of this thesis, see A. C. Lloyd, Necessity and Essence in the Posterior Analytics, in
E. Berti (ed.) Aristotle on Science: The ‘Posterior Analytics’. Padua 1981, 157–71 and Sorabji, op.cit.
ch. XIII, 209–22. I thank Jonathan Barnes, Wolfgang Carl, and Ulrich Nortmann for critical comments
on an earlier version of this chapter.
    33

seemed to force the conclusion that the premises of natural sciences must also
include contingency statements.
As I have attempted to show, the thesis that scientific statements are necessary
truths can in fact be rendered compatible with the thesis that there can be a science
of what occurs ‘for the most part and by nature’, but not always in the same way.
However, Aristotle’s formal theory seems to have led him to the conclusion that
the ultimate premises of some natural sciences can only be statements of
contingency.
We should not suppose that Aristotle never realized that his position was
contradictory. However, he evidently preferred to leave the problem open and
to think about it further rather than allowing his system to drive him to accept one
position or the other.
Taken together, Aristotle’s considerations regarding the necessity and contin-
gency of natural processes offer us a framework for explaining why there can be
science in the realm of the contingent: namely because the regular course of
certain processes makes it possible to distinguish normal cases from exceptions
or interruptions. Admittedly, though, what ought to count as normal in a given
domain is a difficult problem, and one that still occupies scientists today.
 3
Assertoric vs. Modal Syllogistic

This chapter argues that the inconsistencies in Aristotle’s modal syllogistic

are probably due to the fact that he did not realize that the switch from the
predication relations of the Topics to the quantitative term relations in
the Analytics would lead to difficulties for the interpretation of his premises.
The conversion rules regularly lead to what he calls (at An.Po. I 22.83a1–18)
predication per accidens (as in ‘the white is a log’) and rules out for scientific
proofs. Such cases do not affect the validity of assertoric syllogistic, but they
do lead to unacceptable results in the case of de re—modalities. Aristotle
notices such problems in a few places, but the solutions he offers hold only for
the particular cases.

Aristotle’s assertoric syllogistic was a spectacular success; by comparison, his

modal syllogistic was a failure. How could this happen?
Just in case someone would want to object to the second claim, let me briefly
remind you of the reason why I think it should be uncontroversial: Aristotle’s
modal syllogistic is incoherent. Beginning with Alexander of Aphrodisias,¹ who
was probably not the first to say this, we find commentators pointing out that if
Aristotle had consistently applied his own methods of proof and refutation, his
results would have been different—so different, indeed, that one may reasonably
doubt whether he would have preferred to accept them or whether he would
rather have changed his basic assumptions. There is, therefore, no hope in
finding a consistent formal model to represent Aristotle’s modal syllogistic,
and attempts to construct versions that capture as many as possible of the
moods he accepts seem pointless. The important question ought to be what
went wrong, and why.
It appears from the ancient commentaries, and indeed from Aristotle’s text
itself (A.Pr. I 15.34b5–18; I 17.37a9–31), that modal syllogistic ran into troubles
from the very beginning. The first critics were Aristotle’s pupils and colleagues

¹ In Aristotelis Analyticorum Priorum Librum I Commentarium, ed. M. Wallies, Berlin 1883,

144.32–145.4; 207.28–36; 213.11–27; 216.24–26. Among modern studies see, for example, J. Hintikka
‘An Aristotelian Dilemma’, Ajatus 22(1959), 87–92; J. van Rijen, ‘Aspects of Aristotle’s Logic of
Modalities’ (Synthese Historical Library 39), Dordrecht: Reidel 1989, 194–9.

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0003
  .   35

Theophrastus and Eudemus.² Theophrastus, it seems, was less interested in

understanding Aristotle’s reasons for accepting certain theses than in trying to
construct a system that would avoid some obvious difficulties. As a result, he
rejected Aristotle’s special conversion rules for possibility-premises, and intro-
duced a sweeping rule to determine the modality of conclusions: the conclusion
will never be stronger than the weakest premise.³ These changes produced a
simpler and superficially less paradoxical system. The first move substitutes the
simpler concept of possibility proper (equivalent to ‘not impossible’ or ‘not
necessarily not’) for Aristotle’s narrower concept of contingency (neither neces-
sary nor impossible). This may not have been a bad idea from a systematic
perspective but, of course, it does not help us in understanding Aristotle. The
notorious peiorem rule, on the other hand, though intuitively appealing and
certainly easy to apply, is really a non-logical rule, similar to the kind of thing
one can find, for example, in Philoponus⁴ (and many later authors): ‘a syllogism
must not have two negative or two particular premises’, and so on. Theophrastus
seems to have argued for his rule by appealing to the analogy with assertoric
syllogistic: if one ranks premises as stronger and weaker in the order a (universal
affirmative), e (universal negative), i (particular affirmative), o (particular nega-
tive), then in assertoric syllogisms the conclusion will ‘always follow the weaker
premise’ (Al.Aphr. in An.Pr. I 124,8–17). This happens to be a correct observation,
but it is not a rule of deduction. According to Bochenski (La logique . . . 101)
Theophrastus’s peiorem rule is not even correct for his own system.
Theophrastus’s alternative, then, does not offer a clue to our question.
The medieval logicians apparently used the de dicto/de re distinction to defend
Aristotle’s decision, in An.Pr. I, to accept first-figure moods with necessary major
and assertoric minor premise and necessary conclusion as valid while rejecting a
necessary conclusion in cases of an assertoric major premise combined with a
necessary minor.⁵ But they were primarily interested in constructing their own
modal logics, and not in offering diagnoses as to why Aristotle might have
proceeded the way he did. For sophisticated attempts to deal with the problems
of Aristotle’s modal syllogistic we have to look at the work of twentieth-century
scholars whose interest in the question was aroused by the pioneering studies of
Lukasiewicz. I will not try to discuss the verdict of Lukasiewicz himself, who
thought that Aristotle’s modal logic was a hopeless muddle. The first plausible
attempt to explain what Aristotle did came from A. Becker⁶ who worked with

² See Alexander of Aphrodisias, In Aristotelis, An.Pr. 1284.8–125.2; 173.33–174.3; 220.9–16, and

234.4–14.
³ For Theophrastus’ system, see I.M. Bochenski, La logique de Théophraste, Fribourg (Suisse) 1947, ch. V.
⁴ In Aristotelis Analytica Priora Commentaria, ed. M. Wallies, Berlin 1905, 70.1–71.17.
⁵ See W. Kneale, ‘Modality De Dicto and De Re’; in Logic, Methodology and Philosophy of Science,
ed. E. Nagel, P. Suppes, and A. Tarski, Stanford, CA 1962, 622–33.
⁶ Die Aristotelische Theorie der Möglichkeitsschlüsse, Berlin 1933.
36    

Heinrich Scholz, not incidentally a friend of Lukasiewicz. One of Becker’s most

important findings was that Aristotle’s rules for the term-conversion of necessary
premises are all right if the modal operator is assumed to range over the entire
proposition, whereas they break down if it is construed as part of the predicate. On
the other hand, Aristotle’s choice of valid and invalid moods in the first figure with
mixed necessary and assertoric premises can be justified if the modal operator is
construed as part of the predicate, but not if it ranges over the entire proposition
(Becker, op. cit. 41–3). In other words, and using the medieval labels, the conver-
sion rules require a de dicto reading, while the ‘mixed’ syllogisms of chapter I 9
must be construed de re. No wonder, then, that the resulting system is incoherent.
Becker’s diagnosis still seems to me to be very plausible, at least as a first step, but
it does not answer all the questions that one might legitimately ask. First of all,
the de dicto/de re distinction is not Aristotle’s, and it might be more desirable to find
some truly Aristotelian ambiguity at the root of the problem, if ambiguity there
is. Second, a diagnosis in terms of possible formal renderings of Aristotle’s
premises obviously will not tell us why Aristotle might have been inclined to
adopt one or the other interpretation of modal premises. Now, modal syllogistic
was presumably developed to serve as a deductive theory for the sciences as
described in the An.Po.⁷ Hence it seems plausible to think that Aristotle’s views
on science might offer some illumination about his understanding of modal
syllogistic. Furthermore, the Analytics is, of course, not the only treatise in which
Aristotle deals with necessity and possibility, so it should also be worthwhile to look
at other passages—for example, in the Metaphysics, the Physics, the De Caelo—to
see whether they throw some light on his theory of modal propositions. In this way
one might hope to discover what Aristotle intended to do—what kinds of argu-
ments he might have had in mind when talking about syllogisms with necessary or
contingent premises.

⁷ Several people—John Corcoran, Ulrich Nortmann, Robin Smith—have suggested that modal logic
should not be seen as the official logic for demonstrative science because the An.Po., as a matter of fact,
seems to use assertoric syllogistic. Scientific demonstrations must indeed have premises that state
necessary truths, but it does not follow that those premises must be explicitly modal, that is, contain
modal operators. This is an attractive proposal with regard to the An.Po., given the scarcity of
references in its text to modal syllogistic, and it would also seem to be corroborated by a passage in
Rhet. II 25,1402b13ff., where Aristotle clearly envisages premises that hold only ‘for the most part’, but
that are stated as universal propositions. However, it seems clear from the structure of An.Pr. that
modal syllogistic was an afterthought, and so we cannot be sure that it was already available when
Aristotle wrote the bulk of the An.Po. Aristotle’s remarks, at An.Pr. I 13,32b18–22 (on the possibility of
demonstration for things that are ‘by nature and for the most part’) and at An.Po. I 6,75a1–10 (on
necessary conclusions from necessary premises), seem to me to indicate, first, that it is not safe to
ascribe to him the distinction between non-modal propositions stating necessary or contingent facts
and explicitly modalized propositions; and second, that he probably intended his modal syllogistic to
cover scientific argument, even though he may have worked it out only after (most of) the An.Po. was
already written. Given that pure necessity syllogistic is supposed to work exactly like assertoric (An.Pr.
I 8), he would not have needed to revise much in his theory of scientific demonstration.
  .   37

As it happens, three studies of this sort were published in 1989–90, by van

Rijen, Patterson, and Nortmann.⁸ Two of these deal only with necessity; the third
(Nortmann) also discusses syllogisms with contingent premises. It is perhaps
reassuring to see that they seem to agree in several ways, though their conclusions
are different. All three acknowledge the fundamental incoherence of Aristotle’s
system as it stands. What they set out to do is to discover what kind of
system Aristotle should have offered, given what he says elsewhere, notably
in An.Po. I 1–4, about the premises of scientific demonstrations. On the basis of
An.Po. I 4–6, van Rijen proposes an extremely restrictive interpretation for
scientific premises, too complicated to set out here in detail. Roughly speaking,
the idea is that such premises will be true only if they fulfil the conditions laid out
in the An.Po. to the effect that the predicate must belong to the subject per se (καθ’
αὑτό). By these standards, a proposition like ‘all ravens are necessarily birds’ will
qualify, but ‘some black things are necessarily birds’ will not be acceptable—
ravens are birds qua ravens, not qua black. Given these restrictions, however,
van Rijen can claim to have shown that the ‘inference base’ of Aristotle’s system of
necessary and assertoric premises is sound (p. 210f.). That is to say, both the rules
of term-conversion and the moods accepted as perfect will come out as correct in
the same system. On the other hand, Aristotle himself does not observe the
strictures imposed in An.Po. when he offers terms for counterexamples, and so
his results do not cohere with his initial assumptions.
Patterson’s interpretation is less restrictive. He points out that there is good
Aristotelian evidence for accepting both types of necessity propositions illustrated
above. We might call the first sort (‘all ravens are necessarily birds’) strict necessity
propositions—they seem to hold both de re and de dicto—and the second pure de
re propositions.⁹ If some black things are necessarily birds, this is not because
being black has anything to do with being a bird, but because some things that can
be correctly described as black are such as to be essentially birds. Patterson’s ‘strict’
necessity is similar to the only sort of necessity admitted by van Rijen, and both
agree that term-conversion of necessary premises will work if this interpretation is

⁸ For van Rijen, see fn. 1 above. The others are R. Patterson, ‘The case of the two Barbaras’, Oxford
Studies in Ancient Philosophy 7, 1989, 1–40; U. Nortmann, ‘Über die Stärke der aristotelischen Modallogik’,
Erkenntnis 32, 1990, 61–82. Nortmann is publishing preliminary results of a more extended study.
⁹ Patterson claims that it is misleading to apply the de re/de dicto distinction to Aristotelian
propositions because Aristotle himself takes the modality to go with the copula—he says several
times that ‘necessarily belonging’ or ‘possibly belonging’ function in the same way as ‘belonging’ or
‘being’ in assertoric propositions. Apart from the fact that the very same passages were cited earlier by
the Kneales (W. Kneale, op. cit above, n. 5; M. Kneale, The Development of Logic, Oxford 1962, 83) as
conclusive evidence of a de dicto reading, the objection that a de re reading treats the modality as part of
the predicate term, which Aristotle manifestly does not do, seems to me to be a misunderstanding. The
predicate-expression in a proposition of the form ‘B is A’, on a modern understanding of translation
into symbolic notation, is ‘ . . . is A’, not just ‘A’, and so to say that the modal operator is treated as part
of the predicate is not to say that it is part of the predicate term. Hence, I take it that Patterson’s ‘weak’
necessity can be understood as a kind of de re necessity; and that the passages about the alleged
parallelism between the copula and modal ‘belonging’ are in themselves inconclusive.
38    

adopted. But Patterson thinks that the perfect moods of chapter I 9 can only be
accepted as valid if the weaker reading is presupposed. Though Patterson does
not proceed to offer a full formalized version of the desideratum he describes as
a ‘truly Aristotelian modal syllogistic’ (p. 21), it is clear that this would have to
be much more complicated than what we have in the An.Pr. Perhaps we would end
up with two different but internally consistent systems instead of one incoherent one.
For syllogisms with contingent premises, a similar approach can be offered, also
based on a consideration of Aristotle’s conception of scientific theories. Although
this does not agree with the official doctrine of the opening chapters of the An.Po.,
according to which demonstration must start from premises that are necessary,
Aristotle tells us in An.Pr. I 13 (32b18–20) that ‘knowledge and demonstrative
deduction’ are possible for things that are as they are only ‘by nature and for the
most part’. He presumably refers back to this view at An.Po. I 32.88b7, where he
says that the first principles (of a scientific theory) will be either necessary or
possible. Propositions that state what happens ‘by nature and for the most part’
are classified by Aristotle as contingent, and so it would seem that he expected his
modal syllogistic to provide a theory of deduction for sciences that start from
contingent premises. Given that the biological sciences would probably belong to
this group, Aristotle would surely have been interested in arguing that they could
contain valid demonstrations. As I have argued in an earlier paper,¹⁰ some of the
rather ingenious moves Aristotle makes in A.Pr. I 13 and I 15, ostensibly in order
to salvage the validity of moods that might otherwise seem hopelessly invalid, can
be taken to indicate that he was thinking of premises that describe cases in which
the predicate belongs to the subject ‘by nature’. The arguments he seems to have in
mind would appear to be valid—but their validity would not be adequately
displayed in a straightforward rendering of Aristotle’s syllogisms with contingent
premises, either de re or de dicto. It seems, rather, that their structure would be
much more complicated that anything Aristotle explicitly considers.
Nortmann has now tried to construct a formal model that captures Aristotle’s
intuitions. He argues, again partly on the basis of An.Po. I 4, but also of passages in
An.Pr. I 13 and I 15, that Aristotle’s remarks about the truth conditions of scientific
and assertoric premises suggest that an adequate rendering of both his necessity-
premises and his contingent premises requires more than one modal operator in each
proposition. Again we find a version of strict necessity similar to those proposed by
van Rijen and Patterson, but this time some of the strictures imposed upon inter-
pretation are made explicit in the formal rendering of the propositions themselves.
In mentioning some results of these recent attempts to develop a ‘truly
Aristotelian’ modal syllogistic, my main point was not, however, to show that
there might be some convergence of views about what such a logic would have to

¹⁰ ‘Notwendigkeit mit Lücken’, neue hefte für philosophie 24/25, 1985, 146–64. English translation
ch. 2 in this volume.
  .   39

look like. The point I am interested in is, rather, this: in showing what kind of a
system one should expect, given Aristotle’s own views in the An.Po. and elsewhere,
each of these scholars is forced to say that there is a gulf between Aristotle’s
ostensible requirements for modal propositions and many of the term examples
he uses. Nortmann puts this perhaps most neatly: he claims that while a proper
Aristotelian modal logic would have to be as strong as PL+S5, Aristotle’s examples
will work only if one presupposes a weaker logic, probably PL+S4.
I think it cannot be denied that there is a tension between Aristotle’s validity
claims and the term examples he offers, mostly to show invalidity. In fact, the
ancient commentators were already acutely aware of this tension. Both Alexander
of Aphrodisias and Philoponus spend a lot of time discussing Aristotle’s examples,
often trying to improve upon Aristotle by replacing them with what they consider
to be more adequate ones. One can illustrate the difficulties that arise by an
example that seems to be one of Aristotle’s favourites: white animals.
These creatures make their first appearance in modal syllogistic in chapter I
9,30b5–6, with the premise (implied by a list of terms) ‘animal necessarily belongs
to some white things’. Alexander explains this by reference to swans: they are
(necessarily, though this is not relevant here) white—let us forget about Australian
swans in this context, and assume that all swans are white—and they are neces-
sarily animals. In the next chapter, however, we find Aristotle saying (30b35) ‘for
it is possible for animal to belong to nothing white’. This serves to introduce the
assertoric premise ‘animal belongs to nothing white’. Now, according to
Aristotle’s own system this should be incompatible with the first proposition; so
how could both examples be correct? Yet as Alexander notes (p. 140,3), Aristotle
introduces the second proposition with a certain defiance (or ‘special pleading’, as
Alexander seems to say: μετὰ παραμυθίας), as though he were saying: it is, after all,
possible for animal to belong to nothing white. So he seems to have noticed the
difficulty, but insists that it does not matter. How did he think this was possible?
Later in the same chapter (I 10,31a14–17), Aristotle claims that the terms
animal, human, white can be used to construct a counterexample to the mood
Baroco NXN (2nd figure).¹¹ Alexander says that with these terms both premises
will turn out to be necessary: animal necessarily belongs to all humans, animal
necessarily does not belong to some white thing (namely snow, 144,18), and so he
introduces a different example. At first one might think that Alexander has made a
mistake: even if animal necessarily does not belong to snow, surely there would be
other white things to which it does not necessarily not belong? But a little
reflection shows that this can hardly be so, for how could anything be contingently
not an animal? This should mean that whatever the subject is, it might or might
not be an animal. In other words, the reference to snow is misleading, since the

¹¹ I use letters to indicate the modality of premises as follows: N for necessary, X for assertoric, Q for
contingent, and P for possible in the weaker sense of ‘not impossible’.
40    

point should be that if anything is not an animal, then that is necessarily so.¹² But
if Aristotle took his example seriously, he must have thought that the second
premise was not necessary, even though he would presumably have agreed that
snow is necessarily white and necessarily inanimate. And I think one can under-
stand why: in both cases we have considered, there would be no problem if the
modalities were construed de dicto. If it is not necessary that there be either swans
or snow, it might be the case (when no swans are around) that animal belongs to
nothing white; and if there were no snow, it might not be necessary for animal not
to belong to certain white things.
As one might expect, white animals have a role to play in examples for
contingent premises as well. They show up in a term-example in I 14,33b6–9. If
one fills in the premises, one gets

(1) Animal contingently belongs to some white things

(2) White contingently belongs to all humans

The problematic case here is (1): Aristotle is using a substance-term in the

predicate position of a contingent proposition. But surely, in an Aristotelian
ontology, nothing can be contingently an animal, or for that matter a dog or a
human being. Alexander, without stating his reason, replaces Aristotle’s terms
with ‘white, walking, swan’ (171.30), thereby avoiding the difficulty. But if term-
conversion of particular contingent premises is legitimate, as Aristotle says it is
(An.Pr. I 3), then (1) is just the equivalent of the innocent-sounding proposition
‘white contingently belongs to some animal’—and, by the way, (2) can be used to
infer ‘human contingently belongs to some white thing’. (In a later passage,
Alexander acknowledges this and reluctantly accepts a substance-term as predi-
cate; 215,8–10). Again, as before, there would be no problem on a de dicto reading:
it would seem to be neither impossible nor necessary for there to be some white
animals.
So far we have found two kinds of difficulty: (i) the propositions Aristotle uses
seem to form an inconsistent set. This would not matter in assertoric syllogistic, of
course (unless two incompatible premises were used in the same syllogism), but it
does seem to matter in modal syllogistic, if one makes the plausible assumption
that what holds of necessity must hold at all times or in all situations. For then any
proposition incompatible with a necessary proposition would have to be neces-
sarily false; (ii) some examples¹³ of contingent propositions involve substance-
terms in predicate position, and this should be ruled out on metaphysical grounds.

¹² Aristotle himself makes this point in connection with the term-conversion of contingent universal
premises, An.Pr. I 17.37a7–9: ‘It is not true to say that human possibly belongs to no white thing. For
there are many things to which it necessarily does not belong, and the necessary was not possible.’
¹³ For more examples of the same sort, see I 20,39b4–6.
  .   41

On the other hand, Aristotle’s examples do not look absurd or even implaus-
ible. If we switch for a moment to anachronistic terminology, we might say that in
a possible world where there are no swans or snow, the first two examples might
be accepted as correct (if construed de dicto); and there is nothing in Aristotle’s
metaphysics, as far as I can see, to rule out such a world. Although Aristotle did
believe that all existing species are eternal, it does not follow that he also thought
that the world necessarily contained precisely those species. Substance-terms will
inevitably show up as predicate terms in contingent premises as a result of term-
conversion, and so Aristotle must somehow have considered that to be acceptable.
Also, there is some evidence to suggest that Aristotle did pay attention to his
examples. I have already mentioned the special emphasis with which he seems to
introduce the problematic premise in I 10, and in another passage (I 11,31b8–10),
he notes that ‘all animals are good’ might not seem plausible, and proposes to
replace the term ‘good’ by ‘being awake’ or ‘sleeping’. In another passage
(I 15,35a2) he remarks ‘The terms should be better chosen’. Though this indicates
that the terms just used were not well chosen, it also shows that Aristotle did not
think this was unimportant. Hence one should not simply assume that Aristotle
was careless, and that his examples can safely be set aside. It seems, rather, that he
thought the difficulties that worried his commentators did not matter. Of course,
they do matter, and he should not have been so confident, for one can use his own
examples to construct counterexamples to the moods he accepts as valid. But we
have already noted that his system is in fact incoherent. For the moment I would
like to ask: why would he, at least initially, have thought that these examples were
all right?
To see why Aristotle might have assumed that funny examples would not make
a difference, I propose to take a look at the examples in assertoric syllogistic. This
was, after all, obviously the paradigm for modal syllogistic, both as regards basic
assumptions and as regards proof procedures. Suppose for a moment that we did
not know that assertoric syllogistic was a spectacular success, and were trying to
see how well it fits the kinds of arguments Aristotle probably had in mind. It is
clear that assertoric syllogistic was not specifically invented for scientific demon-
strations; it was supposed to cover dialectical and rhetorical arguments as well.
Therefore, it should at least accommodate the types of propositions grouped
according to the ‘predicables’ (definition, genus, proprium, accident) in the
Topics. Indeed, ‘mere belonging’ (ὑπάρχειν) is the most general expression for
any kind of predicative relation (Top. VII 155a11–16; 31–36),¹⁴ and so Aristotle
can claim with some justification that all types of dialectical theses (προβλήματα)
can be deduced by syllogisms in the first figure (I 4,26b31–33). But ‘belonging’, in
fact, covers even more than the predicables. In An.Po. I 22 (83a1–18) Aristotle tells

¹⁴ This presupposes the ‘inclusive’ interpretation of the predicables, for which see J. Brunschwig’s
introduction to: Aristote, Topiques, vol. I, Paris 1967, lxxvi–lxxxiii.
42    

us that a proposition like ‘the white is a log’ does not present a real case of
predication, properly speaking, although it may be true, because ‘white’ isn’t really
the subject of the predication. ‘A log is white’ is a proper case of predication—
‘simple predication’ or ‘predication in an unqualified sense’ (κατηγορεῖν ἁπλῶς),
as Aristotle says there; ‘the white is a log’ is at best a case of predication
per accidens. In scientific demonstrations, only simple predications will be
admissible.¹⁵ But, of course, given conversion, per accidens predication will
occur all the time in assertoric syllogistic, and Aristotle has no qualms about
that. ‘White’ occurs as a subject term in several examples (I 4,26b24–25;
I 6,29a9–10), and indeed the previous examples from modal syllogistic had the
same form. ‘Belonging’, then, covers not only proper predication but also predi-
cation per accidens. The metaphysical awkwardness clearly makes no difference to
the validity of deductions. But from this awkwardness follows another: namely
that Aristotle cannot offer a uniform interpretation for his syllogistic premises
such that ὑπάρχειν has the same meaning throughout. In ‘white belongs to some
animal’, white should be understood to be an attribute of animal, but in ‘animal
belongs to some white’, ‘animal’ cannot stand for an attribute of white. Also, as
Martha Kneale pointed out,¹⁶ there seems to be no way we can so interpret
Aristotle’s categorical premises that the terms keep the same reference in both
subject and predicate positions while ‘belonging’ has the same meaning through-
out, even apart from accidental predication. In fact, the predication relations
Aristotle discusses in the Topics are all asymmetrical, but the term relations
customarily represented by e and i in assertoric syllogistic are symmetrical. For
a modern reader, it is easy to interpret those syllogistic premises as referring to
classes, and hence treat e as exclusion, i as overlap, which are indeed symmetrical
relations. But in Aristotle’s ontology there seems to be no room for such classes as
those of sleepers, white things, walkers, and the like. What he explicitly discusses
are either genera, species, and individuals of various categories, or subjects (in the
metaphysical sense) and attributes. But the class of sleepers is neither an attribute
nor a genus nor a species.
What this means, I think, is that assertoric syllogistic is, as it were, metaphys-
ically neutral from the point of view of Aristotle’s ontology—and we can see that
Aristotle accepted this when he admitted predication per accidens into his deduc-
tions. The premises of assertoric syllogistic just do not have a uniform Aristotelian
semantics. Nevertheless, as I said before, assertoric syllogistic is entirely successful.
Given this situation, it seems natural to expect that a modalized version of these
premises will do the same for modal propositions as unqualified ‘belonging’ does
for assertoric ones—namely allow one to construct a modal logic that will be

¹⁵ Hence van Rijen (op. cit., 205–9) stipulates that per accidens predication must not occur even in
assertoric premises belonging to a scientific deduction.
¹⁶ Kneale, The Development of Logic, 63–7.
  .   43

metaphysically neutral, covering all the various sorts of necessity and possibility
without corresponding to any specific case. Though it may be true that Aristotle
needed to construct a modal syllogistic primarily for his theory of demonstrative
science, it does not follow that he intended it to apply only to scientific reasoning.
After all, necessity and possibility are not the exclusive domain of the scientist.
Possibility, at least, looms large in practical reasoning, and rhetorical arguments
will often have to use premises that hold only ‘for the most part’.¹⁷ It would
therefore be desirable to have a deductive theory that does not contain restrictions
that may be needed only in the sciences. We have already seen that assertoric
syllogistic accommodates predication per accidens. Modal syllogistic seems to do
the same—hence presumably the occurrence of substance-terms in the predicate
position of contingent premises. But while, in the assertoric case, conversion
produced metaphysically odd propositions that could still be recognized as true,
in modal syllogistic it seems to lead to propositions that cannot be true—except
perhaps on a de dicto reading.
Though the examples I have so far discussed were all cases of predication per
accidens, it seems clear that this was not the only difficulty. ‘White possibly
belongs to no animal’ should be just as objectionable as ‘animal belongs to nothing
white’, so long as one holds that some animals are necessarily white. Indeed, in An.
Pr. I 16 one can detect the inconsistent pair ‘white possibly belongs to no animal’
(36a29–30) and ‘white necessarily belongs to some human’ (36b12–18), this time
well hidden in lists of terms.
Now, in An.Pr. I 13 we find Aristotle using a different strategy to bring widely
divergent cases under the same rules (32b4–18): he explains that ‘is possible’ is
said in two ways. That is to say, contingent premises will apply both to what
happens ‘by nature and for the most part’ and what happens by chance and might
just as well be otherwise. Complementary conversion, however, Aristotle insists,
holds in both cases, though not for the same reasons. Furthermore, one can also
construct deductions for both sorts of cases—though one usually does not bother
to do this for things that happen by mere chance (32b21–22). It is a fair guess that
Aristotle did not bother to try this out, either. He evidently expects demonstra-
tions or deductions in both fields to take the form of syllogisms with contingent
premises and conclusions. If one chooses the right premises, say about things that
happen as they do ‘by nature’, the argument should amount to demonstration.
And so it might seem. For example, one might argue: ‘All dogs are (by nature)
runners; all runners have (by nature) sharp eyes; therefore, all dogs have (by
nature) sharp eyes’. This is a valid argument—but not, I think, because it has the
form of a syllogism in Barbara with contingent premises and conclusion. If this
were so, we should also have to accept the following: ‘All Athenians may happen

¹⁷ See Rhetoric I 2,1357a22–b5; II 25,1402b13–1403a5. Note that both passages refer explicitly to the
Analytics.
44    

to be white; all white things may happen to be swans; therefore, all Athenians may
happen to be swans’. This is a syllogism about ‘what might as well be such as
otherwise’. I would insist that it is a genuine counterexample to Barbara QQQ as
initially introduced by Aristotle. To say that this mood represents a valid form of
argument (provided that one sticks to the right interpretation of the premises) is
like saying that ‘p and q, therefore s’ represents a valid form of argument. Of
course it does, since any valid syllogism, for example, can be so represented. But
this is not what we usually mean when we say that a formula represents a valid
deduction—what we mean is that any argument that can be so formalized will be
valid.
Parodying Aristotle, we might offer the following argument to deal with the
case of our white animals: ‘It is possible for animal to belong to nothing white’ is
said in two ways—either because no white thing is necessarily an animal, or
because, though some white things are necessarily animals, there may be no
such animals around at a given time.’ Hence it is possible for animal to belong
to nothing white, although not, of course, while there are any swans. (For the last
sentence, compare An.Pr. I 10,30b37–38: ‘for it is possible for a man to become
white, although not so long as animal belongs to nothing white’.) But how does
one express, within the confines of Aristotle’s modal syllogistic, the proposition
that it is not necessary for there to be animals that are necessarily white, as distinct
from the proposition that no animals are necessarily white? On the one hand, one
can see why Aristotle would not have wished to rule out the offending proposition:
it can be true, though unfortunately not if taken in the de re sense that Aristotle
seems to presuppose for his contingent premises. On the other hand, since we are
given no clear distinction between what is necessary in one (the real) world and
what is necessary in all possible worlds, there is nothing in Aristotle’s official
theory to prevent one from constructing the following counterexample to the
mood Celarent QNQ (accepted as perfect at I 16,35b23–26): Animal possibly
belongs to nothing white; white necessarily belongs to all swans; therefore, animal
possibly belongs to no swan.¹⁸
Now, I think that it was under pressure from just such counterexamples as
I have cited that Aristotle introduced some additional requirements that seem to
point in the direction of scientific premises. In the last paragraph of An.Pr. I 13
(32b25–32), he suddenly tells us that contingent universal premises can be
understood in two ways: either so that the subject term refers to all that it actually
belongs to, or so that the subject term refers to all that it possibly belongs to. As
Becker pointed out (32–5), the second interpretation can be used to salvage the

¹⁸ This will not be a valid counterexample if one follows Nortmann, who considers Aristotle’s
acceptance of the paradoxical results of term-conversion for contingent premises as an indication that
he was prepared to countenance variants of the real world such that in some such variant ‘animal’ is not
a predicate that does or does not belong, if at all, with necessity. But I am inclined to think that this
definitely takes us too far away from Aristotle.
  .   45

allegedly perfect syllogisms with two contingent premises; but it will not work for
all of Aristotle’s contingent propositions. What it does is, in effect, to rule out the
kind of counterexample I have given for Barbara QQQ by postulating that the
second premise (‘All white things may happen to be swans’) cannot be true. Since
Aristotle probably added this requirement when he had already worked out a first
draft of modal syllogistic, he did not check to see whether the required interpret-
ation could be adopted in all cases, let alone whether his examples were in
conformity with it. A similar thing happens in the well-known passage I 15
(34b7–18), where Aristotle explicitly considers a counterexample to Barbara
XQP. There he stipulates that universal assertoric premises must be chosen
‘without qualification, and not as determined with respect to time’. In this way
he eliminates the following example: ‘Man belongs to everything in motion [at
some specific time]; it is possible for moving to belong to every horse; therefore, it
is possible for man to belong to every horse.’ But as before, Aristotle’s examples
both before and after this passage show that he did not envisage this requirement
at the beginning—he keeps using exactly the sort of ‘temporally qualified’ prem-
ises that he rules out in this passage. In both these cases the additional require-
ments are ingenious and not arbitrary—as Aristotle himself indicates in the
second passage, he seems to be adjusting his assumptions so that they fit certain
valid arguments that he presumably had in mind when setting out to develop his
modal syllogistic. But they cannot serve to show that Aristotle’s system as it stands
is valid (or has a valid inference base) for a certain restricted interpretation.
Rather, they point to a different system or systems that Aristotle did not
develop—either a modal logic specifically adapted to his theory of science, as
van Rijen postulates, or one that makes room for the apparently deviant proposi-
tions we have been considering, for example by introducing a distinction between
de re and de dicto modalities. Such a system might be, in a sense, a ‘truly
Aristotelian’ modal logic, but it is not to be identified with the extant
Aristotelian modal syllogistic. Aristotle’s modal syllogistic was a failure—not
because Aristotle did not pay attention to the requirements of his metaphysical
theory, but because what worked so well in assertoric syllogistic did not work in
modalibus. He failed to discover a general, metaphysically neutral deductive
theory that covered both the respectable cases and the funny ones, as assertoric
syllogistic works for both proper predication and predication per accidens.
We might see this failure as just punishment for Aristotle’s Fall, as Geach has
called the move away from the predicables in the Topics to the term-logic of the
Analytics.¹⁹ Or we might emphasize the positive with Martha Kneale, saying that
‘fortunately philosophical obscurity is not an insuperable bar to logical progress’.²⁰
Given the fact that, to this day, logicians have not come to agree on one generally

¹⁹ See P.T. Geach, ‘History of the Corruptions of Logic’, in his Logic Matters, Oxford, 1972, 47–51.
²⁰ Kneale, The Development of Logic, 66.
46    

accepted system of modal logic, the second option seems both more charitable and
more reasonable. Aristotle’s own system will not work as it stands, but it seems to
contain the beginnings of the most important systems that were developed after
him. Also, as I have tried to show, the difficulties that arise in his system were not
the result of mere blunders. On the contrary, they seem to have provoked further
investigations that led to important later developments, such as, indeed, the
medieval distinction between modalities de dicto and de re. Hence, at the very
least, Aristotle laid the foundations—which is, after all, what one would want to
say about his assertoric logic as well.²¹

²¹ I am grateful for criticisms and suggestions from participants at the conference, and especially for
the detailed written remarks of Ulrich Nortmann (who disagrees with most of what I say here).
 4
Perfection and Reduction in Aristotle’s
Prior Analytics

This chapter argues that perfecting a syllogistic mood is not the same as reducing
it to another mood. To perfect a mood is to make its validity evident, which may
or may not be done by reducing it to a perfect one; reduction works between
moods of all figures, perfect or not. The epistemic priority of the first-figure moods
led some later Peripatetics to claim that the imperfect moods owed their validity to
the perfect ones and would not be valid without them. Boethus of Sidon (first
century ) refuted them by arguing that all the valid moods are perfect, at least in
some cases appealing to the terminology of parts and wholes that Aristotle himself
uses at the beginning of chapter 4 of Prior Analytics I when introducing the moods
of the first figure.

Ever since Łukasiewicz reinstated Aristotle as the founder of formal logic, there
has been a wealth of studies of Aristotle’s syllogistic as a formal system. Against
Łukasiewicz’s claim that syllogistic is a system in which certain theses function as
axioms, others as theorems derived from these, it has been argued—convincingly,
to my mind—that it would be historically more accurate to represent syllogistic as
a system of natural deduction that starts from a set of primitive rules and
demonstrates the validity of inferences by showing how their conclusions can be
deduced from their premises.¹ However this particular dispute may be decided,
both parties start from the assumption that syllogistic is a system, and then argue
about what kind of a system it is. Now Aristotle, of course, did not start from the
notion of a deductive system of one or another sort. He describes what he does in
An.Pr.  1–2, 4–7 either as ‘perfecting’ syllogisms² or as ‘reducing’ one syllogism

¹ See K. Ebbinghaus, Ein formates Modell der Syllogistik des Aristoteles, Hypomnemata 9, Göttingen,
1964; T. Smiley, ‘What is a Syllogism?’, Journal of Philosophical Logic 2, 1973, 136–54; J. Corcoran,
‘Aristotle’s Natural Deduction System’, in Ancient Logic and its Modern Interpretations ed. Corcoran,
Dordrecht 1974, 85–131.
² I use ‘syllogism’ as a convenient transliteration of Aristotle’s συλλογισμός, with the understanding
that it does not mean a two-premise argument in one of the syllogistic moods but has the wider
Aristotelian meaning ‘valid deductive argument’. The currently favoured translation ‘deduction’,
though perhaps less misleading than ‘syllogism’, does not bring out the point that Aristotle takes
syllogisms to be arguments. I shall also sometimes use the non-Aristotelian term ‘mood’ where
Aristotle would have used συλλογισμός for a type of argument.

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0004
48    

to another. Since it has seemed perfectly clear, in terms of modern formal logic,
what Aristotle is doing in those chapters—namely, either deducing theorems from
axioms, or proving the validity of rules of inference by showing how a certain kind
of conclusion can be derived from a certain kind of premises using only the
primitive rules of the system—the terminology of perfecting and reducing has
not attracted too much attention. Most commentators seem to suppose that to
perfect a syllogism is to reduce it to a perfect syllogism, and that, in turn, consists
in showing that the mood so reduced is valid, given the perfect mood, which is
already accepted as valid. Hence it is customary to describe the proofs in A 4–6
as ‘reductions’ of the second- and third-figure moods to the perfect moods of the
first figure.
It can hardly be disputed that what Aristotle does in these chapters is to
prove the validity of second- and third-figure moods. But it seems to me that
an investigation of the relations between perfection and reduction might help
us to see more clearly how Aristotle arrived at the system that marks, after
all, one of the most important differences between the Prior Analytics and the
Topics. My claims will be (1) that perfecting a syllogism is not the same as
reducing it to a perfect syllogism, in one or the other sense of ‘perfect’; and (2)
that perfection and reduction together are what makes syllogistic a deductive
system, as opposed to a mere collection of inference rules. Finally, there will be
some speculation about why Aristotle’s system was largely ignored by the later
tradition.³
The verb ἀνάγειν (to reduce) appears for the first time in the Prior Analytics in
chapter A 7,29b1. Aristotle has just stated that ‘all imperfect syllogisms are
perfected through the first figure’ (29a30–31). In the next paragraph, he continues:
‘But one can also reduce all syllogisms to the universal syllogisms in the first figure.
For those in the second figure it is clear that they are perfected through these . . . ’.
Obviously, Aristotle implies that if a syllogism S₁ can be perfected through another
syllogism S₂, then it can be reduced to S₂. Now, since all the moods in A 6 were, in
fact, perfected through the first figure, one might think that ‘reducing’ and
‘perfecting’ are simply synonyms. Accordingly, commentators since Alexander
of Aphrodisias (113,5–9; among recent authors, Patzig⁴) have assumed that if the
two particular moods of the first figure (Darii and Ferio) can be reduced to the
universal ones (Barbara and Celarent), the universal syllogisms must somehow be
more perfect, or perfect in a more stringent sense, than the particular ones. But the
following sentence shows that Aristotle did not think so: he declares that Darii and
Ferio are ‘perfect in themselves’ (literally, ‘perfected through themselves’), though
they can also be proved indirectly through the second figure. That is to say, they

³ For the later ancient tradition I am relying on the excellent monograph by Tae-Soo Lee, Die griechische
Tradition der aristotelischen Syllogistik in der Spätantike, Hypomnemata 79, Göttingen 1984.
⁴ Patzig, Die aristotelische Syllogistik, 3rd ed., Göttingen 1969, 77.
      ’    49

can be reduced to second-figure moods that are themselves reducible to Barbara

or Celarent. So, a perfect mood can be reduced to an imperfect one, according to
this passage, and that should be enough to show that not every case of reduction
has to be a case of perfection.⁵ But one might still hold that perfection is a special
case of reduction, namely reduction to a perfect syllogism. I do not think this is
correct either, but in order to show that, I will have to ask first what exactly
Aristotle meant by ‘perfecting a syllogism’.
The verbs τελειοῦν and ἐπιτελεῖν (to perfect) suggest that perfecting should be a
way of transforming an imperfect syllogism into a perfect one. This has been
understood in two ways, depending upon one’s interpretation of the adjective
τέλειος (perfect). A perfect syllogism is one in which ‘nothing besides the assump-
tions is needed to make the necessity obvious’ (A 1.24b23–24). The difference
between perfect and imperfect syllogisms lies in the fact that the validity of the
perfect syllogisms is evident, which is not so for imperfect syllogisms. Aristotle
declares the four moods of the first figure to be perfect (A 4), and proceeds in the
following chapters to prove the validity of the remaining moods by means of
deductions that include one of the perfect moods as a syllogistic step. If perfecting
consists in turning an imperfect syllogism into a perfect one, then Aristotle
apparently thought that the syllogism to be proved is transformed into the perfect
syllogism that occurs in the proof. Patzig⁶ claims that this is indeed what Aristotle
thought. When he comes to a syllogistic step involving a first-figure mood, he
often says things like ‘the first figure has come about’ (γεγένηται). He also speaks
several times of imperfect syllogisms as ‘possible syllogisms’, by which, since all
syllogisms must be valid arguments, he can hardly mean anything but ‘possibly
perfect syllogisms’. Patzig notes that this way of speaking is misleading, since an
argument in Darapti, say, does not become an argument in Darii just because its
conclusion can be deduced from its premises by a deduction that contains a
syllogism in Darii. This would certainly be a justified complaint, but I think that
Aristotle’s language does not imply this line of thought, and that it is, in fact, not
plausible to attribute this view to him. When Aristotle says ‘the first figure has
come about’, he need be taken to mean no more than that a premise-pair of the
first figure has been reached in the deduction; and the phrase ‘possible syllogism’
can just as well be understood as ‘syllogism that can be perfected’ (δυνατὸς
ἐπιτελεῖσθαι), which takes us back to the question of what Aristotle meant by

⁵ At A 23,40b17, in a reference back to A7, Aristotle says that (all) syllogisms in the second and third
figure ‘are perfected through the universal syllogisms in the first figure and reduced to these’,
apparently obliterating the distinction between perfecting and reducing. Alexander of Aphrodisias
(in An.Pr. 255,1–17) notices the difficulty and suggests two solutions: either Aristotle just means to say
that all second- and third-figure syllogisms are reduced to these (in which case Aristotle would be
saying twice the same), or he is using καθόλου instead of ἁπλῶς, to say that second- and third-figure
syllogisms are ‘generally’ perfected through first-figure ones. I am more inclined to think that
Aristotle’s formulation is negligent—the statement at the end of A 23,41b3–5 is accurate.
⁶ Patzig, Die aristotelische Syllogistik, 140–3.
50    

‘perfecting’. What seems to me to be decisive, however, is the fact that it is wildly

implausible to think of an imperfect syllogism as being transformed into the
relevant perfect one in the case of indirect proofs. For while one can understand
how Aristotle might have seen an argument in Cesare as merely another version of
an argument in Celarent, this is certainly not so in a case where the syllogism that
occurs in the deduction has as its conclusion the contradictory of one of the
original premises. The proof for Baroco, for example, looks like this: Assume that
M belongs to every N but does not belong to some X; then it will follow that N does
not belong to some X. For if N belongs to every X and M belongs to every N, it
follows (by Barbara, in the first figure) that M belongs to every X. But we assumed
that it did not belong to some X (cf. A 5.27a36–b1). The first-figure syllogism has the
form MaN, NaX / MaX (Barbara), and I do not see how this could, by any stretch of
the imagination, be seen as a different version of an argument in the form MaN,
MoX / NoX. If ‘perfecting’ means ‘reducing to a perfect syllogism’, then it should
mean no more than ‘prove to be valid by means of a perfect syllogism’.
If we want to keep the supposition that perfecting must be a way of transform-
ing an imperfect syllogism into a perfect one, the second interpretation of τέλειος
looks more promising. John Corcoran⁷ has argued that the distinction between a
perfect and an imperfect syllogism is that between a full deduction and a valid
inference that needs filling out. An inference may be valid but not obviously so,
and one can show that it is valid by deducing the conclusion from the premises
using only primitive rules of deduction. Following this interpretation, Robin
Smith, in his recent new translation of the An.Pr.,⁸ has decided to render τέλειος
as ‘complete’ rather than ‘perfect’. This is a plausible suggestion, because what
Aristotle does in chapters A 5–6 is just that—showing how the conclusion of an
‘incomplete’ or ‘imperfect’ syllogism can be deduced from its premises. Corcoran
concludes that the long deduction will then count as a perfect syllogism. But this
no longer agrees with Aristotle’s own explanation of the term τέλειος. He says that
he calls a syllogism τέλειος if nothing besides the premises assumed is needed to
make the necessity obvious (24b23–24; cf. also A 4.26b30; A 5.27a16–18;
A 6.29a15–16; A 16.36a6–7). In general, he seems to think of a syllogism as
consisting only of its premises and conclusion. True, at A 25.42a32 he says that
every proof and every syllogism ‘will have two premises and no more—unless
something is added, as we said in the beginning, for the purpose of perfecting
(completing) the syllogisms’. Yet no amount of filling in will bring it about that the
necessity of the conclusion is evident from the initial premises alone. Once the
deductive steps have been filled in, it will be obvious that the conclusion follows
from the premises, but not before; and this seems to be precisely the reason why
Aristotle calls syllogisms that need filling in ‘imperfect’. Hence, I would argue that

⁷ Corcoran, ‘Aristotle’s Natural Deduction System’, 91–4.

⁸ Aristotle, Prior Analytics. Hackett, Indianapolis 1989.
      ’    51

the remark in A 25 is meant simply to forestall the objection that the deductions in
A 5–6 have more than two premises—they do not, Aristotle wants to say, because
the additional steps in the deduction do not add any new assumptions.
Furthermore, before we decide to translate τέλειος as ‘complete’, we should
consider that every syllogism must be complete in the sense that ‘no further
term is needed from the outside to bring about the necessity’ (24b21–22).
Confusion between the two kinds of completeness apparently led some ancient
commentators to suppose that second- and third-figure syllogisms were not even
valid by themselves and needed some help to make the conclusion follow
(Ammonius 31, 28–33). It might be better to indicate the second kind of com-
pleteness by the word ‘perfect’, since it is a feature that need not belong to every
valid deductive argument.
I think we should give up the supposition that perfecting consists in turning an
imperfect syllogism into a perfect one. It is a procedure that serves to make it
evident that a certain conclusion follows from certain premises. What is perfected
is, strictly speaking, not the syllogism, but the obviousness of the necessity. And
Aristotle actually comes close to saying that in A 5, 27a17, when he explains why
second-figure syllogisms are not perfect: ‘for the necessity is perfected not just
from the initial premises but also from others.’ To make it evident that the
conclusion follows, one will presumably use only deductive steps that are them-
selves evidently valid—among them the perfect syllogisms. The proofs in A 5–6
each contain one of the perfect syllogisms, and that must be the reason why
imperfect syllogisms are said to be ‘perfected through’ and ‘reduced to’ perfect
syllogisms. We seem, then, to get the same result as before: to perfect a syllogism is
to reduce it to a perfect one.
But now one should note that in chapter A 6 Aristotle points out several times
that certain moods (Darapti, Datisi, Bocardo) can also be proved by the procedure
he calls ἔκθεσις. The ecthetic proof for Darapti is set out in 28a24–26: ‘if both
P and R belong to every S, then if one takes one of the S, say N, both P and R will
belong to it, so that P belongs to some R.’ If the letter N is here construed as
standing for a general term, the proof coincides with the demonstrandum. Hence
in this case, at least, one should take N to stand for the name of an arbitrarily
chosen individual. But then the proof contains no syllogism (in the narrow sense)
at all, which shows that perfecting need not take the form of proof through a
perfect syllogism.⁹ The proofs for Datisi and Bocardo can be construed as

⁹ As Ian Mueller points out to me, one might object that it is not clear whether Aristotle considers all
proofs of validity in A 6 as cases of perfecting. Ecthesis is mentioned only as an alternative method of
proof, alongside reductio ad impossibile, and in A 7.29a30–31 Aristotle simply asserts that ‘all imperfect
syllogisms are perfected through the first figure’, either directly or indirectly. But in his summaries at
the end of chapters A 5 and A 6 (28a4–9; 29a14–18) he seems to refer back to the preceding proofs quite
generally as cases of perfecting. The reason why ecthesis is not mentioned in A 7 is presumably that it
does not lead to first-figure syllogisms and hence is not a method of reduction to the first figure.
52    

syllogistic without circularity, but if so, they probably do not contain one of the
perfect syllogisms. In fact, in the one passage where Aristotle describes ecthetic
proofs as syllogistic (A 8.30a9–14), he explicitly says that the syllogisms will be in
the same figure as the syllogism to be proved, that is, second and third, respect-
ively. So perfection is, in principle, independent of reduction to perfect syllogisms.
It consists in proving the validity of a syllogistic mood by making it evident that
the conclusion follows from the premises, by a deduction that may or may not
contain one of the perfect syllogisms. I have already noted that reduction is also
independent of perfection, since Aristotle acknowledges that some of the perfect
moods can be reduced to imperfect ones. What, then does he mean by ‘reduction’?
Here I would like to begin by looking at A 45, the other chapter in the An.Pr. in
which Aristotle discusses the reduction of syllogistic moods to one another. A 45
comes at the end of a long section (A 32–45) in which Aristotle explains how
ordinary-language arguments can be put into syllogistic form. A 32 states the
topic: ‘how we are going to reduce (ἀνάγειν) syllogisms to the aforementioned
figures.’ In the next sentence, Aristotle picks this up by saying ‘if we can resolve
(ἀναλύειν) syllogisms which have been produced into the aforementioned figures’
(46b40–47a4). ἀνάγειν and ἀναλύειν are used interchangeably throughout this part
of book A. As a special case of this sort of reduction or analysis, A 45 treats the
question of how an argument that has already been put into syllogistic form can be
‘reduced to’ a syllogism in another figure—that is, reformulated as a syllogism in a
mood from another figure. Aristotle considers only pairs of syllogisms that have
the same form of conclusion. He notes at the beginning (50b8–9) that this kind of
reduction is possible only for some, not for all, moods. For example, Barbara is not
mentioned at all, because it is the only mood with a universal affirmative conclu-
sion. Aristotle goes through the figures and shows how, by conversion of one or
both premises, first-figure syllogisms can be transformed into second-figure ones,
second-figure ones into first-figure ones, third-figure ones into first-figure ones
and vice versa, and so on.¹⁰ The only operation considered here is conversion—
understandably, since the indirect proofs in A 5–6 did not lead to other syllogisms
with the same conclusion as the one to be proved.
Though it is true, of course, that a syllogism that can be reduced to another in
this way can also be proved to be valid by the same method if the mood to which it
is reduced is already accepted as valid, this does not seem to be the point in which

¹⁰ In those cases where Aristotle converts a universal affirmative premise to a particular one, such as
Darapti to Darii, Felapton to Ferio, the premises of the second syllogism are weaker than those of the
first, and one might object that this can no longer count as a reformulation of the original argument,
while it remains the case that the first syllogism can be reduced to the second in the sense of ‘proyed’.
However, it is still true that, given the premises of Darapti or Felapton, one can always use the same
terms to produce a syllogism in Darii or Ferio. The fact that Aristotle limits himself to syllogisms with
the same conclusion, and also his use of ἀναλύειν, seem to me to support the interpretation that he is
thinking of reformulating a given argument, or using the information given in its premises to produce
another argument for the same conclusion, rather than proving validity.
      ’    53

Aristotle is interested here. The words ‘perfect’ and ‘perfecting’ do not occur at all,
since, after all, this sort of reduction works in both directions between moods of all
three figures. Rather than read this chapter as trying out different axiomatizations
for syllogistic (as suggested, for example, by Patzig¹¹), I would suggest that we
should compare the reductions performed here with a kind of ἀναγωγή that
occurs elsewhere in Aristotle’s treatises, and which consists in trying to classify
a larger number of cases under a few general headings. So, for example, Aristotle
repeatedly claims that all pairs of opposites can be reduced to ‘the one and the
many’ as genera or principles (Metaph. Γ4,1055b26–29; cf. Γ 2,1054b27–1055a2).
The mysterious reduction of opposites is always assumed as given, but there is an
extended example of this kind of reduction in GC 2.2,329b15ff., where Aristotle
shows how all the differentiae of tangible bodies can be ‘reduced’ to the four
primitive ones, hot and cold, wet and dry, by being explained as special forms of
them. An example that is closer in subject matter to An.Pr. A 45 occurs at
Sophistici Elenchi 6,168a17–20. Aristotle has just listed and classified various
types of fallacies. Then he remarks ‘One must either classify (διαιϱετέον) apparent
syllogisms and refutations in this way, or else reduce (ἀνακτέον) them all to
ignoratio elenchi, taking that as the starting point (ἀϱχή). For all the ways (of
fallacious reasoning) we mentioned can be resolved (ἀναλῦσαι) into the definition
of refutation’. That is to say, every type of fallacy can be described as a case of
ignoratio elenchi because it will violate one of the conditions in the definition of a
valid refutation. Here, as in An.Pr. A 45 ἀνάγειν and ἀναλύειν are used as
synonyms.¹² This suggests that the kind of reduction Aristotle had in mind in
A 45 consisted in seeing whether syllogisms from one figure could be subsumed
under the heading of a mood from another figure as being different versions of it.
If this is what he was doing, then what he shows in that chapter is, among other
things, that reduction in the sense of analysis does not offer a way of ordering the
syllogistic moods or figures in such a way that the moods of any two figures can be
subsumed under those of one other figure. There is also no way in which one of
the figures can be said to be prior to the others on the ground that their moods can
be reduced to its moods, but not the other way around.¹³

¹¹ Patzig, Die aristotelische Syllogistik, 55.

¹² For ἀναλύειν alone in connection with classification under genera, see Metaph. Δ 28.1024b9–16:
ἕτεϱα δὲ τῷ γένει λέγεται ὧν τε ἕτεϱον ὑποκείμενον καὶ μὴ ἀναλύεται θάτεϱον εἰς θάτεϱον μηδ’ ἄμϕω εἰς
ταὐτόν . . .
¹³ Robin Smith (‘Some Studies of Logical Transformations in the Prior Analytics’, History and
Philosophy of Logic 2, 1981, 1–9) has suggested that A 45 is one of several chapters in the An.Pr. that
contain attempts by Aristotle to find some general principles for syllogistic before he discovered the
method of perfecting. In his commentary (1989), he says that the topic of A 45 might be stated as ‘Is
indirect proof necessary?’. I do not think this is an accurate statement of the topic, since reformulation
is not the same as proving validity. But it might be that Aristotle had at first hoped to unify the system
of syllogisms by grouping all valid moods under some most general ones, preferably from the same
figure.
54    

By contrast, the method of reduction discussed in A 7 serves to do precisely

that. If reduction is construed, not as a way of transforming a syllogism in one
figure into a syllogism in another figure, but as proving its validity by a deduction
that uses a syllogism from another figure, then all the other moods can be reduced
to those of the first figure—or indeed, as Aristotle shows in A7, to the two
universal moods of the first figure alone. Aristotle evidently knew that this second
form of reduction is not the same as the method he considers in A 45, for at the
end of that chapter he says: ‘It is now clear that the same syllogisms cannot be
resolved in these figures [sc. second and third] that could also not be resolved into
the first; and that when we reduce the syllogisms to the first figure, these are the
only ones that are brought to a conclusion by reductio ad impossible.’ But the
exercises in transformation of A 45 would certainly also have shown him that
reduction in the sense of proving validity through another syllogism does not
suffice to single out one particular figure or set of moods as the one to which all the
others should be reduced. Once indirect deductions are admitted, it is just as easy
to reduce all first- and second-figure moods to those of the third figure as it is to
reduce second- and third-figure moods to those of the first. This is, after all, why
Aristotle was said to be trying out different axiomatizations for syllogistic in A 45.
Reduction to another syllogism will show that one mood is valid if the other is, but
since it works in both directions between each pair of figures, it will not establish
any order of priority among the figures.
This is the point at which the notion of perfection comes, as it were, to the
rescue. If one set of syllogisms can be singled out as being perfect, that is, evidently
valid and not in need of proof, then this offers a non-arbitrary way of determining
which moods should count as basic, so that the others should be reduced to them
and not vice versa. By combining reduction and perfection, then, Aristotle man-
aged to unify syllogistic in such a way that it can be called a deductive system.
Now, since the first-figure moods are chosen as primitive on epistemological
grounds, it is tempting to see Aristotle as devising axioms for his deductive theory.
The validity of the perfect syllogisms is supposed to be immediately evident, just
as, it seems, the principles of a demonstrative science are supposed to be grasped
without proof. Furthermore, the word ἀνάγειν itself suggests that what something
is reduced to should be a kind of principle (ἀϱχή).¹⁴ Yet, as far as I know, Aristotle
himself never describes the perfect moods as ἀϱχαί.¹⁵ It should also be kept in
mind that a modern formal model of his syllogistic would treat at least one of the
conversion rules as primitive alongside the perfect moods. Aristotle never

¹⁴ For the association of ἀνάγειν with άϱχή, cf. the passage quoted above from the Sophistic Elenchi,
and also, for example, Physics B 3,194b22–23; Rhet. A 4,1359a38; EN 3.5,113b19–20; Metaph.
E 3.1027b14–15.
¹⁵ When he speaks of ‘syllogistic principles’ (συλλογιστικαὶ ἀϱχαὶ) at Metaph. Γ 3.1005b7, he is
clearly thinking of the principles of non-contradiction and excluded middle presupposed by all
deductive arguments.
      ’    55

considered this, I think, because he did not set out with the idea of devising a
deductive system, either axiomatic or natural deduction, in which certain theses or
rules were to function as primitive. He seems to have been interested, rather, in
establishing an order of priority among the moods, and so A 7 talks only
about reducing all other moods to those of the first figure. What Aristotle created
was, in fact, a deductive system, but not because he set out to create one from the
very start.
There is also a deeper reason why, even on the unlikely assumption that he
considered the first-figure moods as propositions, Aristotle should quite properly
have hesitated to call the perfect syllogisms principles of the others: namely that
they do not stand to the derived moods in the relation in which the principles of
an Aristotelian demonstrative science must stand to the theorems derived from
them. In spite of some attempts by Aristotle to fit it into his theory, the relation of
logical consequence or implication, which exists between the premises and con-
clusion of a valid argument, but also between certain syllogistic moods considered
as theses, is not one of the four explanatory relations set out in the doctrine of the
Four Causes. Although Aristotle wishes to say that the premises are somehow the
cause of the conclusion, he wavers about how the relation fits into his scheme. At
Physics B 3,195a18–19 (= Metaph. Δ 2,1013b20), he claims that the premises are
the material cause of the conclusion; but in Physics B 9,200a15–30, he suggests,
rather, that the conclusion is related to the premises as matter to form, or means to
end. The second suggestion is actually the more plausible of the two, since it is
based on the consideration that the end implies the means as the premises imply
the conclusion, but not vice versa. But it would obviously not be plausible to say
that the premises consist of the conclusion, or that the conclusion is there for the
sake of the premises. As a matter of fact, the first premises of an Aristotelian
science must be ‘causes of the conclusion’ (An.Po. A 2,71b22) in two ways: they
must imply the conclusion, of course, but they must also state the reason why what
is asserted in the conclusion is the case (cf. Alex.Aphr. in An.Pr. 21,12–18). But one
cannot claim that Barbara and Celarent, say, explain why Baroco and Camestres
are valid. Arguments in all four moods are valid because their conclusions follow
from their premises or, as one might also say, they satisfy the definition of
‘syllogism’. But this explains very little, and, of course, does not show that the
validity of any of the moods can be deduced from that definition. So syllogistic is
indeed a deductive system, but it is not an Aristotelian science.¹⁶

¹⁶ None the less, there is some evidence that Aristotle might have thought of the Analytics as offering
a science of arguments. In Rhet. A 4.1359b10, for example, he refers to ἀναλυτικὴ ἐπιστήμη. More
important than this phrase, which might after all be non-terminological, is Robin Smith’s observation
(‘The Mathematical Origins of Aristotle s Syllogistic’, Archive for History of Exact Sciences 19, 1978,
201–9) that the language in which Aristotle expounds syllogistic in the systematic chapters of the An.
Pr. is strikingly similar to the language in which theses are formulated and proved in Euclid’s Elements.
Since mathematics is Aristotle’s paradigm of a demonstrative science, however inadequate syllogistic
56    

Aristotle himself never explains why he finds it important to show that all
syllogisms can be reduced to the perfect moods of the first figure. But later
Peripatetics raised the question, and this led to the dispute about perfect syllo-
gisms that I mentioned before. This debate must go back at least to the time of
Boethus of Sidon in the first century .¹⁷ Ammonius (31,13–25) reports that
Boethus ‘rightly thought, and indeed proved, that all syllogisms in the second and
third figure are perfect (τέλειοι)’ and that he was followed by Porphyry,
Iamblichus, Maximus, Syrianus, Proclus, Hermias the father of Ammonius—
and, of course, Ammonius himself. Now, a few pages before (14,28–29)
Ammonius has explained the word τέλειος as ‘self-sufficient, not in need of
anything in order to be (a syllogism)’; hence one might think that Boethus was
merely defending the view that second- and third-figure syllogisms are valid in
themselves.¹⁸ No wonder that Ammonius, who claims at first that Boethus was
contradicting Aristotle, ends up by pointing out that Boethus’s view can in a way
be derived from Aristotle’s own words (33,18–21). But in fact the story seems to
have been more complicated.
Some of the background to Ammonius’s remarks can be found in Themistius’s
treatise ‘In reply to Maximus about the reduction of the second and third figures
to the first’.¹⁹ In this treatise, Themistius argues against Maximus’s claim that ‘the
categorical syllogisms in the second and third figure are perfect in themselves and
do not need either demonstration or reduction to the first figure’. Themistius
defends the supposedly Aristotelian view that the first figure is naturally prior to
the others (168,30–36) because it is their ‘cause et génératrice’ (? αἰτία καί γένεσις,
167,2). The second and third figures, he says, are ‘generated’ by the first by
conversion, that is, changing the order of terms in the first and second premise,

may be to represent its deductive structure—this similarity is at least an indication that Aristotle tried
to set out the theory of syllogisms in the form of a demonstrative science. The use of the word ἀπόδειξις
to refer to the proofs of validity in chapters A5 and A6 points in the same direction. But what
corresponds to Euclid’s exposition of geometry is not the system of syllogisms considered in A 7, but
rather the metatheory of deductive argument, whose theorems are not themselves syllogisms but
statements about syllogisms such as the one with which Aristotle introduces the valid moods of the
first figure in A 4,25b32–35. This metatheory would seem to include the (non-syllogistic) proofs given
for the conversion rules in A 2, and also the definitions of a- and e-premises in A 1.24b28–30, to which
Aristotle appeals for the claim that the first-figure moods are evidently valid, that is, perfect. As Ian
Mueller points out to me, these definitions could indeed be said to explain why Barbara and Celarent
are valid—which is presumably the reason why the ‘dictum de omni et nullo’ was later considered as
the first principle of syllogistic. However, the underlying logic of ‘analytics’ as a metascience would have
to be considerably stronger than ‘syllogistic’. (I take this point from Ian Mueller’s unpublished paper
‘Interpreting Syllogistic’.
¹⁷ For the (uncertain) date of Boethus, see P. Moraux, Der Aristotelismus bei den Griechen, Berlin,
1973, 143.
¹⁸ So Moraux, 167.
¹⁹ Published in a French translation from an Arabic source by A. Badawi, La Transmission de la
philosophie grecque au monde arabe, Paris 1968, 166–80. The text seems confused in several places, and
I cannot decide where this is due to Themistius, to the Arabic translation, to the French translation, or
to a combination of these. I think, however, that the passages I am using here are clear enough.
      ’    57

respectively (170,27–171,26);²⁰ and the valid moods of those figures derive their
validity from those of the first (167,40–168,5; 169,13; 176,9–10; 179,22).
Obviously, Themistius—unlike Aristotle—did take the first-figure moods to be
‘principles’ of the others. Maximus had argued for the view of Boethus and other
Peripatetics (171.3–5), mainly, it seems, on the basis of the two theses (1) that
first-figure moods can be reduced to second- and third-figure ones just as well as
the other way around (166.28–32; 169.16–17), and (2) that the second- and third-
figure moods can be shown to be valid without recourse to the first figure
(170,4–5; 176,11–12; 177,2–3). For (1) he probably appealed to An.Pr. A 45—
Themistius grants the point and hence does not discuss it. In support of (2),
Maximus, and before him Boethus, apparently produced various methods of
proving the validity of second- and third-figure moods without reduction to the
first figure. One of these, not surprisingly, was ecthesis (177, 4).²¹ The second,
however, is very interesting: it consists in showing that certain moods can be seen
to be valid by appeal to the sort of consideration that Aristotle himself uses in An.
Pr. A4 to show that the first-figure moods are perfect. Themistius repeatedly
ascribes to Maximus the principle ‘what is totally separate is also separate in all
its parts’ (172,25–26; 173,34–35; 174,17–18). The proof for Festino (second
figure), based on this principle, is set out as follows: ‘une partie de C est contenue
en A, tout A est séparé de B, ce qui en lui est partie de C est done séparé de B’
(173,21–22). The reasoning seems to be: if A is totally separate from B (second
premise), then the part of it that is a part of C (first premise) must also be separate
from B. Similar proofs were apparently offered for Cesare, Camestres, and Baroco,
but the text is more confused in those passages. In the case of Darapti (third
figure), the reasoning seems to be: if the middle term is a part of both extremes
(e.g. B and C are both said of all A), then the major will be said of a part of the
minor (B will be said of a part of C, namely A); cf. 174,35–39. It appears that
Boethus and his followers were elaborating upon the terminology of wholes and
parts that Aristotle uses only occasionally,²² but after all quite prominently. For
example, this is how Aristotle first introduces the perfect syllogisms in Barbara
and Celarent, A.Pr. A 4.25b32–35: ‘When three terms are related in such a way

²⁰ This is not nonsense, as Lee thinks (Die griechische Tradition, 123), but trivial: ‘conversion’ in the
sense in which it is used here is not a logical operation. It should be noted that Themistius uses
‘reduction’ only in the sense in which it is synonymous with ‘analysis’. So he says (173,41–44): ‘Aristote,
quoiqu’il ne lui soit pas arrivé de réduire le quatrième mode (de la deuxième figure) à la première figure,
de toute façon il le démontre au moyen de la première figure, car la réduction à l’absurde se fait dans
cette figure.’ He seems to think that the imperfect syllogisms become valid only after being transformed
into first-figure ones; cf. 175,18–29.
²¹ The French translation has ‘hypothèse’ here, but the Arabic text must have had a word translating
ἔκθεσις, as is clear from an earlier passage, 175.34, where Themistius quoted An.Pr. A 6,28b14–15—not
A 7.29b7–8, as the translator claims. There is also a method of reductio ad absurdum, ascribed to
Boethus (177,7ff.), that I will not try to discuss.
²² For Boethus, see 178.11–12, where Themistius claims that he explained the expression ‘dans tout’
(ἐν ὅλῳ) in a different way from Aristotle’s.
58    

that the last is in the middle as a whole and the middle is or is not in the first as a
whole, there will necessarily be a perfect syllogism of the extremes.’ In A1, ‘being
(or not being) in something as a whole’ appears twice—among the expressions to
be explained (24a13), and at the end of the chapter, where Aristotle declares that
for one thing to be in another as in a whole is the same as for the second thing to
be said of every of the first (24b27–29). There follows Aristotle’s explanation of the
phrase ‘to be predicated of every’, and in the rest of the treatise the expression ‘in
something as a whole’ occurs hardly at all, except for chapter A 4. Aristotle
evidently came to prefer the locutions ‘to be predicated of every/no’ and ‘to belong
to every/no’. However, the standard expressions for particular and universal
premises, ἐν μέϱει and καθόλου, also show that wholes and parts must have played
some role in the prehistory of syllogistic. The sentence just quoted from An.Pr. A 4
invites the idea that syllogistic premises should be interpreted in terms of wholes
and parts. Aristotle himself suggests that the perfect syllogisms can be seen to be
valid by reflecting upon the expressions ‘said of every/no’ (A 4,25b9–40,
26a27–28; cf. also A 14,32b40–33a5, 33a24–25). Similarly, one might suggest
that the validity of syllogisms can be seen by reflecting upon the meaning of the
phrases ‘in something as in a whole’ and ‘separated from something as a whole (a
phrase not used by Aristotle himself, who has the rather misleading ‘not to be in
something as a whole’). Since this is exactly the kind of procedure Aristotle
employs to argue for perfection, one could then also claim that syllogisms
recognized as valid in this way should count as perfect. This second sort of
argument, then, shows why Boethus would indeed have disagreed with
Aristotle’s claim that only the first-figure moods are perfect.
Thesis (2), I think is correct—the validity of second- and third-figure moods
can be established independently of reduction to the first figure, and Boethus was
right in claiming that all valid moods are perfect, or at least that there are more
perfect moods than the four that Aristotle had singled out.²³ Aristotle’s reasons for
declaring the four first-figure moods to be perfect may have been stronger than
one realizes at first sight,²⁴ but he certainly did not state them very clearly. One
might suspect, indeed, that he found it rather convenient to have all perfect
syllogisms assembled in one and the same figure and did not bother to look
further. For a modern reader, Boethus’s and Maximus’s principles about wholes
and parts are reminiscent of Euler diagrams,²⁵ which can be used precisely to
make the validity of certain inferences evident. It is no accident, for example, that

²³ Themistius suggests in one passage (172,8–10) that Boethus did not claim perfection for all the
moods of the second and third figures. Given that Boethus used several methods of proof, including
ecthesis and reductio ad absurdum, I am inclined to think that this is correct. To show that the validity
of all syllogistic moods can be established without reduction to the first figure is not the same as
showing that they are all perfect, though Boethus’s opponents may have thought so, and Maximus
apparently did make the stronger claim.
²⁴ See Patzig, Die aristotelische Syllogistik, ch. 3.
²⁵ See Lee, Die griechische Tradition, 129, n. 20.
      ’    59

Themistius complains that on Maximus’s principles the moods Celarent, Cesare,

and Camestres would seem to coincide (171,38–172.8): the corresponding Euler
diagrams would be the same. If such methods count as establishing validity,
reduction to the first figure will not be needed for the purpose. Now, Euler
diagrams might not be considered as proofs,²⁶ precisely because they are not
deductions. Yet they could surely be used to demonstrate the validity of syllogisms
much in the way in which truth-tables are used in present-day propositional logic.
To show that most or all syllogistic moods can be seen to be valid by Euler
diagrams, or by reflecting upon the meaning of the logical constants used in
syllogistic premises, should be sufficient to establish that they are not dependent
on first-figure syllogisms for their validity. In this sense, they do not need
demonstration.
Now, if more than the four moods chosen by Aristotle count as perfect, then
one has to admit that the choice of one set of syllogisms rather than another as
primitive is, to some extent, arbitrary. But this does not mean that there is no
point in developing a deductive system to exhibit the logical—as opposed to
epistemological—relations between syllogistic moods. By contrast, the result of
proving validity piecemeal, as it were, using different methods for different cases,
will not be a deductive system. Themistius is right to complain that Maximus’s
and Boethus’s procedure lacks generality (173,35–38; 177,5–7).
Ammonius’s and Themistius’s reports are not full enough to allow us to decide
whether Boethus’s main aim was just to refute the misunderstandings of those
who claimed that the first-figure moods were the ‘cause and origin’ of all others, or
whether he set out to criticize Aristotle. If it was the second, Boethus may have
failed to see the point that reduction to the first figure is a method not only of
proving validity, but also of unifying syllogistic in a way that his methods would
not achieve.
However that may be, it seems obvious that the emperor Julian was right to
decide the dispute between Maximus and Themistius in favour of Maximus
(Ammonius 31,18–22). As we saw, all the major Neoplatonists agreed with him.
Unfortunately, the arguments of Boethus and his followers seem to have led to the
view that reduction to the first figure is unimportant, since it is not indispensable
as a method of proving validity. Perhaps this was the main reason why Aristotle’s
own deductive system fell into neglect and was eventually replaced by the notori-
ous rules of the syllogism.²⁷

²⁶ It should be noted, however, that Aristotle himself occasionally refers to the appeal to the meaning
of ‘predicated of every’ as a proof (ἀπόδειξις, A 14.33a27 and A 9.30b2, where the words ‘the proof is
the same’ presumably mean that a syllogism is perfect).
²⁷ I am grateful to the participants in the Berlin conference held in honour of Günther Patzig on
25 July 1991 for objections and advice. Special thanks are due to Ian Mueller for his detailed comments
on the chapter and for letting me read his unpublished article ‘Interpreting Syllogistic’.
 5
Aristotle and the Uses of Logic

This chapter argues that although Aristotle was clearly interested in logical theory,
he did not consider his syllogistic as a science in its own right. He saw it as part of
a general theory of argument, optimistically expecting it to cover the entire field of
valid deductive argument. Such a theory would provide not only scientists, but also
speakers in political debates or in court as well as philosophers, with methods of
constructing proofs, refuting arguments, and discovering fallacies. This is
illustrated by passages from the Prior Analytics beyond the exposition of the
formal system of syllogistic.

Aristotle, as we all know, invented formal logic. Over the last fifty years or so,
scholars have learned to recognize that what he presented in the first few chapters
of the Prior Analytics (An.Pr.) is the real thing—a system of formal logic, whether
or not the inspiration for the discovery of the syllogism had anything to do with
Platonic division. We no longer hear about the magical force of the middle term or
the alleged demonstrative power of first figure syllogisms as opposed to, say, the
superficial subtleties of Stoic logic. Although Aristotle’s syllogistic covers only a
small part of the field of modern mathematical logic, what he offered contained all
the elements of a formal deductive system. He introduces the system of syllogistic
moods by defining its technical terms, stating and justifying the primitive rules,
and then providing formally correct proofs of the derivative rules. In other words,
he developed a complete system of natural deduction, limited indeed by the
assumption that all propositions must be simple subject–predicate sentences,
but otherwise flawless.¹ Aristotle was rightly proud of his discovery, and so it is
not surprising that he greatly overestimated its power, apparently thinking that
syllogistic would be sufficient to represent the deductive structure of demonstra-
tive sciences, including geometry or mathematics in general. It is only natural that
the modern logicians who vindicated Aristotle’s position in the history of their
subject should have gone on to find Aristotle also pursuing other interests of his
modern successors, such as proof theory and other parts of logical theory. For

¹ This summarizes the conclusion of J. Corcoran, ‘Aristotle’s Natural Deduction System’, in

J. Corcoran (ed.), Ancient Logic and Its Modern Interpretations, Dordrecht 1974, 122–3.

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0005
       61

example, Jonathan Lear in his Aristotle and Logical Theory² sets out to show that
Aristotle was ‘not only the father of logic, but also the (grand)father of metalogic’
(preface p. 1), and an article by Robin Smith describes ‘Aristotle as Proof
Theorist’.³ I do not wish to deny that these subjects can be found in the
Analytics, but it seems to me that some of his more recent commentators have
been inclined to ascribe to Aristotle an interest in the formal theory of deduction
independent of the wider purposes of the Analytics, thus painting him as more of a
logician in the modern sense than he probably was.
Just as there was a dispute in late antiquity over whether logic was a part or
merely an instrument of philosophy,⁴ so there is some uneasiness today about
whether logic properly belongs to philosophy or not. Logic has developed into a
formidable technical discipline that may perhaps better be seen as a branch of
mathematics—certainly one needs to be a trained mathematician to do any
significant work in the field. But most philosophy departments still require their
students to take at least a course in elementary logic. Even if the connection
between informal argument and formal logic has become harder to see, it is still
the case that we need a clear understanding of such notions as consistency,
inference, and validity; and though philosophical arguments usually are not
deductive proofs, they had better not be logical fallacies. Furthermore, formaliza-
tion can often help us to clarify distinctions that are hard to detect in ordinary
language, even if we do not believe that formalized languages are in some way
better or more scientific than natural ones. So, a part of logic at least continues to
be an instrument of philosophy—or rather of any discipline that makes use of
reasoning and argument. Aristotle was interested both in logic as a theory and in
its more humdrum uses in philosophical or indeed everyday argument, and more
than half of the text of the An.Pr. is concerned with the uses of logic in argument
rather than with either the exposition of a formal system or what we would call
logical theory. This is what one should expect, since Aristotle invented formal
logic for the purposes of his general theory of argument, not just as a formal
theory of deductive proof or an ‘underlying logic’⁵ for demonstrative science. In
order to show how the perspective of a general theory of argument differs from
that of logical theory, I will argue that although syllogistic can be shown to be
complete in the modern logician’s sense, it was not considered by its author to
be complete in the sense relevant to his project. A deduction system is complete in

² J. Lear, Aristotle and Logical Theory, Cambridge 1980, 1.

³ R. Smith, ‘Aristotle as Proof Theorist’, Philosophia Naturalis 21,1984, 590–7.
⁴ See, for example, Alexander of Aphrodisias, In Aristotelis Analyticorum Priorum Librum
I Commentarium, ed. M. Wallies (CAG 2.1, Berlin 1883), 1,1–6,12, and the discussion of the debate
by T.-S. Lee, ‘Die griechische Tradition der aristotelischen Syllogistik in der Spätantike’, Hypomnemata
79, 1984, 44–54.
⁵ So Corcoran, loc. cit. n. 1, 98.
62    

the modern sense if it allows one to deduce all (and only) the valid formulae. What
Aristotle had in mind when he set out to show that ‘every deductive argument
(συλλογισμός) . . . is in one of the (syllogistic) figures’ (A 23,40b20–22) was the
claim that every valid deductive argument can be formulated as one or more
syllogisms in the narrow sense. This, as Aristotle recognized, is not the case
(A 44,50b2–3). However, I will also argue that he thought syllogistic captured at
least a necessary component of every valid deductive argument, and perhaps that
it was indeed sufficient as an account of the logical form of scientific demonstra-
tion. Finally, I will illustrate the role of formal syllogistic in the theory of argument
by a few examples from the second half of An.Pr. A and from book B.

The system of syllogisms is introduced in An.Pr. A 4 (25b28–31; cf. A 1,24a22–

b15) as part of a general theory of valid deductive argument, more general than
the theory of scientific proof, which is set out in the Posterior Analytics (An.Po.).
Formal logic is only a part of such a general theory, and that is why, after
presenting the system of syllogisms as an account of the valid forms of deductive
argument, Aristotle goes on to discuss heuristics—how to find premises for theses
that one wants to prove—and what he calls analysis, that is, how to discover the
syllogistic structure of premises and terms in arguments formulated in ordinary
language (cp. A 32,46b38–47a2). The title of the whole work, Ἀναλυτικά, refers to
this part of the treatise, not the formal theory set out in its opening chapters. Much
of book B can be seen as concerned with argumentative strategies and the
detection of fallacies, and its last five chapters contain the sketch of a theory of
rhetorical argument. So, although the introductory lines of the An.Pr. describe the
treatise as a preliminary to the treatment of demonstrative science, the general
theory developed here should serve the practitioner of dialectic and rhetoric as
well as the scientist or philosopher.
In the context of such a general theory, the system developed in A 1–22 is only a
beginning, not the complete account of valid deductive argument that Aristotle in
his more optimistic moments claims to have found, and that too many of his later
followers have taken it to be. We do not have to think that Aristotle would have
agreed with the arguments Alexander of Aphrodisias presents for not pursuing the
subject of valid inference for its own sake.⁶ But it may be that Aristotle developed
his formal theory only to the extent that he thought he needed it for his account of
demonstration, leaving many loose ends that seem to have been picked up by his
students Theophrastus and Eudemus. That may be the reason why, for example,

⁶ Alexander argues that the study of inferences that cannot be used as arguments is useless,
2,33–4,29.
       63

he did work out a system of modal syllogistic, but did not try to offer a formal
account of the second type of valid deductive argument he recognizes in chapter
A23—what he calls ‘argument (syllogism) based on a hypothesis’—although he
saw that it was not covered by the system of the syllogistic figures.
The point that Aristotle himself considered his formal logic as incomplete may
have been obscured by the fact that he actually purports to offer a completeness
proof, and that the first half of this proof can be read as a completeness proof in
the modern technical sense.⁷ In order to show in what sense formal syllogistic was
not complete, then, I have to draw attention to some differences between the
system of syllogisms as described and developed by Aristotle in An.Pr. A 1–7 and
its best modern models.
I take it to be generally agreed by now that formal syllogistic is best represented
as a system of natural deduction rather than an axiomatized theory. Aristotle
explicitly introduces all the elements of such a system, and several scholars have
shown that within its limits it is complete—that is, allows one to deduce all and
only the valid formulae.⁸ Nonetheless, I think it is misleading to assume that
Aristotle regarded formal syllogistic as a system of deductive inference. Although
it would be correct to say that he uses such a system, it is not what he is talking
about when he sets out to prove completeness, or when he shows, in A 7, that all
syllogistic moods can be reduced to the two universal moods of the first figure,
rather than the four he had introduced as perfect in A 4. This is so because
Aristotle is focusing on the notion of valid argument, not logical consequence or
valid inference in general. His definition of ‘syllogism’ (A 1,24b18–20) stipulates
that a syllogism must have at least two premises, and that the conclusion must not
only follow, but also be different from the premises. This shows, I think, that he
thought of syllogisms as arguments or argument-forms, not just valid deductions
or inferences. The notion of a valid deductive argument may be hard to define
precisely, but it is clear that it is narrower than the notion of a valid deduction. For
example, any proposition can be validly deduced from itself, but no one would
wish to present an inference of p from p as an argument. When Aristotle says, as
he does several times (e.g. A 15,34a17–18; A 23,40b35; B 2,53b18–20), that
nothing follows of necessity from a single proposition, he should not be taken to
deny that any proposition implies itself, or to maintain that nothing can be
inferred from a single premise. After all, in A 2 he actually proves the validity of
four types of single-premise inference—the conversion rules he is going to use in
his proofs in chapters A 5 and A 6. But he apparently thought that such inferences
could not be used as arguments because their conclusions would not state

⁷ See T. Smiley, ‘Aristotle’s Completeness Proof ’, Ancient Philosophy 14, 1994, 25–38.
⁸ See, for example, J. Corcoran, ‘Completeness of an Ancient Logic’, Journal of Symbolic Logic 37,
1972, 696–702; P. Thom, The Syllogism, Munich 1981, 181–92; R. Smith, ‘Completeness of an Ecthetic
Syllogistic’, Notre Dame Journal of Formal Logic 24, 1983, 224–32.
64    

anything different from their premises (cp. An.Po. A 3,73a7–11), and hence they
would be question-begging.⁹ The syllogistic moods Barbara and Celarent have a
different status, and when it comes to showing that the two particular moods of
the first figure can be reduced to them, Aristotle does not mention any of the other
rules of his system, but speaks as though all syllogistic moods could be derived
from Barbara and Celarent alone. So he is not trying to establish the minimal
number of primitive rules of inference for the system he uses, but the number of
primitive forms of argument presupposed by it. What he thought about the
deduction rules—conversion, reductio, ecthesis—we do not know, since he
never discusses their status. Again, the first half of the completeness proof in
A 23 takes the form of showing that any direct deductive argument with two or
more syllogistic premises can be transformed into a chain of two‑premise syllo-
gisms which will, therefore, consist entirely of arguments in one of the syllogistic
figures. This can be read as a completeness proof in the modern sense, but it is
actually only one half of Aristotle’s own purported proof. The second half sets out
to show that all arguments ‘based on a hypothesis’, and most importantly all
reductio-arguments, will also fall into the figures; and this part is a failure.¹⁰
Aristotle first offers an analysis of hypothetical arguments according to which
such arguments consist of a syllogistic deduction plus one more step, which he
describes as ‘based on a hypothesis’. This is the step that distinguishes hypothetical
arguments from normal direct syllogisms. In the reductio‑arguments discussed
almost exclusively in A 23, the step would appear to be the move from ‘p implies a
falsehood’ to ‘p is false’, and hence to the assertion of not-p. In the direct
hypothetical arguments analysed in A 44, the step seems to be, in modern
parlance, an application of the rule of modus ponens. In any case, the hypothetical
arguments are clearly not complete without the last step that leads to the conclu-
sion, and so it is astonishing to see Aristotle assert, at the end of A 23, that these
arguments too will be in one of the figures. The most he can say at this point—and
he has not proved that either—is that such arguments must contain a syllogism in
one of the figures. And at the end of A 44 he does acknowledge that this type of
argument cannot be analysed as a syllogism in the narrow sense.
With regard to the modern formal models of syllogistic, this should show that
although the natural deduction system Aristotle uses can indeed be shown to be
complete, Aristotle’s syllogistic as a formal theory of valid deductive argument is
not complete, and Aristotle did not think it was. The divergence between Aristotle
and his modern commentators arises, I think, because they treat reductio as a
deduction‑rule, the validity of which need not be proved within the system.

⁹ For this point, see M. Frede, ‘Stoic vs. Aristotelian Syllogistic’ [1974] in his Essays in Ancient
Philosophy, Hackett,Minneapolis 1987, esp. 114–15.
¹⁰ I have argued in detail for the account of hypothetical argument I presuppose here in ‘Aristoteles
über Syllogismen “aufgrund einer Hypothese” ‘, Hermes 107, 1979, 33–50; chapter 1 in this volume.
       65

Aristotle so uses it himself in chapters A 2, A 5 and A 6, but in A 23 he considers

reductio as a distinct type of argument, and he is right in pointing out that it is not
covered by the syllogistic moods. This is not a minor omission in Aristotle’s
projected theory. For it appears from A 44 that the class of hypothetical arguments
is supposed to contain most, if not all, of the argument-forms treated in the Topics,
and Aristotle emphasizes that these need to be investigated and clearly distin-
guished (50a39–40). But it seems that he left this task to his students.
Now, the explicit recognition that hypothetical argument is not syllogistic
comes much later than the completeness proof in A 23, and so one might be
tempted to fall back on a developmental explanation, saying that Aristotle believed
what he claims in A 23 and came to realize only later that hypothetical argument
will not fit into the figures. But speculation about chronology should obviously be
a last resort, and it would seem particularly implausible in this case. The Analytics
are among the works that Aristotle himself cites by their traditional title, and
though they are obviously not a polished treatise and do seem to contain sections
(such as modal syllogistic) that were inserted into already existing drafts, there is
good reason to think that it was Aristotle himself who put together the materials
collected in the four books. Moreover, the argument of A 23 actually shows that
hypothetical arguments are not strictly speaking in the figures. Hence it seems
more plausible to suppose that Aristotle was content with the weaker claim that
every proof or every valid deductive argument must at least contain a syllogistic
deduction. If this were true, then anyone who tried to produce a formally valid
argument would need to rely on the syllogistic moods, even if the argument
consisted of a hypothetical step in addition to the syllogism.
But why did Aristotle think that hypothetical arguments had to contain a
syllogism in the narrow sense? This assumption appears unargued in both chap-
ters A 23 and A 44, and that in spite of the fact that the example Aristotle gives in
A 44 might actually be taken to show just the opposite. I think that the reason why
Aristotle still thought he did not have to argue for his claim has to do with his
conception of syllogistic necessity: he thought that the necessity that forces one to
accept the conclusion of a hypothetical argument depends upon the necessity of
the preceding syllogism.
Aristotle did not have the notion of a propositionally complex sentence or
proposition. He thought that one should in principle be able to reformulate any
given valid argument as a series of categorical propositions in which ‘one thing is
said of one thing’. Simple sentences are not contrasted with complex ones formed
by logical connectives, such as conjunctions, conditionals etc., but with sentences
that express more than one categorical predication. For Aristotle, what we would
call a conjunction appears to be a set of two propositions, and similarly for the
other complex sentence-forms. An Aristotelian syllogism in one of the figures
consists of at least two categorical premises and a categorical conclusion which is
said to result of necessity because what is asserted in the premises is (supposed to
66    

be) the case (τῷ ταῦτα εἶναι, 24b20). But the premises assert that some one thing
holds or does not hold of all or some of some other thing, and so the conclusion
will follow necessarily only if the term-relations asserted to hold in the premises
necessitate the term-relation asserted in the conclusion. In a hypothetical argu-
ment, on the other hand, it will indeed be necessary to accept the conclusion, but
not because one has asserted some premises such that the term-relations asserted
necessitate the term-relation in the conclusion. The conclusion is accepted
because its contradictory has been shown to be false or because some other
proposition has been proved and one has agreed beforehand to accept a proof
for that proposition in lieu of a proof for the conclusion. Aristotle contrasts the
two sorts of necessity in A 44 by saying that ‘it is necessary to agree [sc. to the
conclusion]; not on the basis of (ἐκ) a syllogism, however, but on the basis of a
hypothesis’ (50a25–26).¹¹ An anachronistic way of describing the difference
would be to say that in the strictly syllogistic case, the inference is justified by
the term-relations asserted in the premises, while in the case of hypothetical
argument it is justified only by the (assumed) relations between the truth-values
of the propositions involved. Aristotle did not realize that truth-value relations
could in fact be asserted in complex propositions such as conditionals, and so he
thought that the conclusion of a hypothetical argument, though validly estab-
lished, had not been reached in the same way as the conclusion of a proper
syllogism.
This does not yet show, of course, why a hypothetical argument would have to
contain a syllogism, but one can see why Aristotle would have thought so from
looking at his description of the two types of hypothetical argument he recognizes.
One type is described as follows: One begins with an agreement to the effect that a
proof for a proposition q will be accepted in lieu of a proof for another proposition
p, the original demonstrandum. One then produces a proof for q, and because of
the initial agreement, the interlocutor will have to agree that p has been proved as
well. An example of this type of argument—no doubt the paradigm case that
accounts for the label ‘hypothetical’—would be Socrates’ argument ‘from a
hypothesis’ in the Meno. Socrates and Meno agree that if they can show that
virtue is knowledge, they will have shown that virtue can be taught. Socrates then
sets out to find a proof for the thesis that virtue is knowledge, and it is clear that if
he had found such a proof, he and Meno would have been satisfied that virtue is
indeed teachable. Aristotle calls the proposition for which a proof is to be offered
the ‘substitute’ (μεταλαμβανόμενον)—the thesis that virtue is knowledge is substi-
tuted, as it were, for the original demonstrandum—and he claims that it must be
proved by a direct syllogism. Now, unfortunately his example in A 44 does not

¹¹ The same point is made in A 23 ,41a24–25: ‘For all those who come to a conclusion through an
impossibility deduce (συλλογίζονται) the falsehood, but prove (δεικνύουσι) the original thing from an
assumption’ (Robin Smith’s translation); cp. also 41a32–34.
       67

bear him out, for he clearly envisages an indirect proof for the substituted thesis, as
the ancient commentators already point out (cp. Al.Aphr. 387,1–11). But this will
not invalidate his general claim that every valid deductive argument must contain
a syllogism if it can be shown that every reductio-argument contains a syllogism;
so we must look at the other type of hypothetical argument.
A reductio is an argument in which one derives a falsehood or absurdity from
the contradictory of the demonstrandum, and then concludes that the demon-
strandum must be true because its contradictory has been shown to be false. In
order to derive the absurdity, one cannot use another indirect argument, since
presumably nothing absurd can be derived from the thesis to be proved. Given
that all premises are assumed to be categorical, it would seem that the
absurdity can only be deduced by a categorical syllogism, as Aristotle maintains.
For since Aristotle does not recognize the possibility of a direct modus ponens—
argument, he cannot consider the option of deriving either the so-called substitute
proposition or the falsehood in a reductio from a conditional together with the
proposition that forms the antecedent. The necessity involved in hypothetical
argument thus seems to be parasitic upon syllogistic necessity in the narrow sense,
because the final step can only be reached if a proof has been given, either for the
falsity of the contradictory or for the substitute thesis, and that proof cannot itself
be hypothetical, on pains of a regress.
So much for Aristotle’s claim that every valid deductive argument must contain
at least one syllogism. It would assure him that he had established the indispens-
able core of a general theory of argument, and although he could not in fact
prove the ambitious thesis of A 23, the first half of his argument there—the
completeness-proof for direct syllogistic deductions—retained its importance as
a proof that the core theory was indeed complete.
Furthermore, he may have thought that this part would provide all that was
needed for scientific demonstration, even if it did not adequately cover the wider
field of dialectical argument, or indeed scientific reasoning in general, because
demonstration in the strict sense would not have to rely upon hypothetical
argument. Though both forms of hypothetical argument recognized by Aristotle
may have been taken over by the philosophers from Greek mathematics,¹²
Aristotle may have thought that they belonged to a heuristic or preliminary
stage of scientific investigations. Plato’s Socrates (Meno 86E–87A) claims to
have learned the hypothetical method of reasoning from the mathematicians,
although the geometrical example he uses to explain the method is notoriously
much harder to understand than the dialectical procedure it allegedly serves to
illustrate. It seems clear, however, that the example does not illustrate a method of
proof, but a heuristic strategy. The Meno’s argument about virtue and knowledge

¹² I am indebted to Henry Mendell for reminding me of this point, and driving it home with
examples.
68    

appears, together with another geometrical example (not Plato’s), in An.Pr. B 25,
the chapter dealing with ‘leading away’ (ἀπαγωγή). This procedure clearly serves a
heuristic purpose, namely to ‘lead away’ from a given problem that appears for the
moment intractable to another one that can be more easily solved. Aristotle
describes the Meno example as follows: ‘ . . . let A be teachable, B stand for science,
and C justice. That science is teachable, then, is obvious, but it is unclear whether
virtue is a science. If, therefore, BC is equally convincing as AC, or more so, it is a
leading away . . . ’ (69a24–27, tr. R. Smith). It seems that this will bring us ‘closer
to scientific understanding’ (69a23–24) because it shows us that if we can find a
proof for the minor premise (BaC), then we will have a (syllogistic) proof for
the desired conclusion, ‘virtue is teachable’ (AaC). It is interesting to note that
the argument considered here is not ‘hypothetical’ in the sense of A 23 or A 44; the
additional assumption is simply the minor premise in a regular syllogism. What
was earlier described as a hypothesis or agreement—‘If (we can prove that) virtue
is knowledge, then (we have shown that) virtue is teachable’—turns out to link the
minor premise and conclusion of a proper syllogism. The hypothesis, then, was by
no means arbitrarily chosen, and the example invites the consideration that in the
direct case, at least, one might try to eliminate the need for an agreement by
finding an additional premise that will turn the hypothesis into a syllogism. Thus
direct hypothetical argument might be thought to figure in scientific reasoning
only as a preliminary to the discovery of syllogistic proof.
That still leaves indirect argument, and Aristotle obviously knew that this was
in fact quite common in Greek mathematics—his own favourite example of a
reductio-argument is the proof for the incommensurability of the diagonal with
the side of a square. Nevertheless, he maintains in several places (A 29,45a23–28;
B 1,63b12–16) that any indirect argument can be replaced by a direct one, so that
reductio could in principle be dispensed with. The argument for this claim, as far
as I can see, comes from his treatment of syllogism ‘through an impossibility’ (διὰ
τοῦ άδυνάτου συλλογισμός) in B 11–14.¹³ There he describes indirect argument as
an argumentative strategy, to be employed if an opponent is not willing to accept a
premise one needs to derive the conclusion one is aiming at. An indirect argu-
ment, Aristotle explains in B 14 (62b29–31), is one in which one assumes
what one wants to refute and leads on from there to an agreed falsehood. So if
the opponent will not agree to p, you ask him to accept its contradictory—he
can hardly refuse to do that if he has rejected the original proposition
(cp. B 11,62a15–17)—and then try to derive an obvious falsehood from this
assumption together with some other, already accepted premise. This will serve
to show that the contradictory of the premise you tried to introduce must be false,
and hence allow you to use the desired premise after all. In the simple cases

¹³ In An.Po. A 26,87a1–17 he uses the distinction between ‘deictic’ and ‘per impossibile’—syllogisms
discussed in these chapters to characterize the difference between direct and indirect proofs in general.
       69

Aristotle considers in B 11–14, it is also true that the desired proposition could
have been derived from the contradictory of the ‘obvious falsehood’ together with
the other accepted premise, and so Aristotle concludes that the same terms can
(always) be used to construct either a direct or an indirect argument (B
14,62b38–40; 63b12–13). When one tries to discover a scientific proof, as opposed
to refuting an opponent in a dialectical debate, one will presumably not have to
resort to such stratagems, and hence indirect proof will be dispensable. Since the
difference between direct and indirect argument thus seems to be merely strategic,
and the same terms will be used in both cases, Aristotle thinks that the heuristic
devices used for constructing direct syllogistic deductions will be sufficient to
cover indirect argument as well. The hitch is, of course, that not every indirect
argument will be as simple as the ones Aristotle analyses in B 11–14. The examples
he has in mind there probably do not come from mathematics, but from the
indirect proofs he himself had offered in A 5 and A 6, where the ‘impossibility’
arises in each case because the contradictory of one of the assumed premises has
been derived from the ‘hypothesis’ (i.e. the contradictory of the conclusion that
follows from the premises). It is probably significant that Aristotle does not seem
to distinguish between deriving an ‘obvious falsehood’, that is, a patently absurd
but not logically false proposition, and deriving a contradiction. Only belatedly, in
B 15, does he acknowledge that one might need more than one syllogism to derive
a contradiction (64b17–27)—namely in those cases where one has not assumed
the contradictory of the derived proposition in advance. But it is not clear that this
would have led him to revise his view about the dispensability of indirect proof. He
knew that such proofs were common in mathematics, and hence he found it
important to show that his theory could deal with them, but he probably did not
try to derive, for example, the incommensurability theorem from the premise ‘no
number is both odd and even’ together with other mathematical truths.
Aristotle may also have thought that in a truly scientific system of demonstra-
tions, indirect proofs should be replaced by direct ones. He thought that direct
arguments were preferable to indirect ones for epistemological reasons: as he
points out in An.Po. A 26 (87a20–30), when one derives a falsehood in an indirect
argument, the premises of the syllogistic deduction one uses will not state the facts
on which one’s eventual conclusion is based, and hence will not be ‘prior to the
conclusion’ in the way required for demonstration.¹⁴ Only when the argument has
been transformed into a direct syllogism which uses the contradictory of the
‘obvious falsehood’ as a premise will the correct order of epistemological priority
be restored. I suspect that what Aristotle disliked about indirect proofs was that
they do not do what an Aristotelian demonstration is supposed to do, namely
explain why the conclusion must be true. The incommensurability-proof does

¹⁴ For my interpretation of this difficult and corrupt passage, see Striker, Zeitschrift für philoso-
phische Forschung 31, 1977, 317–18.
70    

show that the diagonal in a square cannot be commensurable with the side, but
one can hardly say that it explains or states a reason why the diagonal must be
incommensurable, least of all by providing an appropriate middle term. So
perhaps Aristotle hoped that, in principle at least, demonstration could proceed
without indirect proof.¹⁵
However that may be, even if formal syllogistic did not provide the compre-
hensive theory Aristotle had promised at the beginning of A 4, the system of the
figures would be the foundation of all valid deductive arguments, whether scien-
tific or not, and its applications in heuristics and analysis would be as important
for dialectic or rhetoric as for science or philosophy. In the rest of the An.Pr., after
A 26 and before he turns to his official treatment of demonstration, Aristotle is
concerned to show how formal syllogistic should be used to construct or to
criticize arguments, and his main focus appears to be on dialectic. I will now
briefly illustrate these uses of logic by a few examples.

II

In An.Pr. A 27–31, Aristotle describes an ingenious method for finding premises

for any given categorical conclusion. Once one has the complete list of valid
syllogistic moods provided in A 4–6, one will no longer have to collect and
memorize long lists of propositions about various fields, as Aristotle had recom-
mended in the Topics. Instead, one can construct lists of terms that either follow,
or are followed by, a given term. The thesis under discussion—the πρόβλημα, as
Aristotle calls it here (A 28, 44a37), using the terminology of the Topics—will have
to contain two terms; hence by looking for terms that occur in the lists for both of
the requisite terms, one will be able to find an appropriate middle term to
construct one’s syllogism. Alexander of Aphrodisias (301,12ff.) illustrates the
procedure by a delightful example: lists of terms one can use to argue that pleasure
is or is not a good, as the occasion demands. Aristotle’s system is more efficient
and economical than Platonic division, which is therefore demoted to ‘a small
part’ of heuristics in A 31 (46a31); and Aristotle insists that his method is the only
one guaranteed to locate all the premises that can be used to construct valid

¹⁵ As Robert Bolton points out to me, at Top. 8.2.157b34–158a2, Aristotle seems to say the opposite,
namely that indirect proofs are just as good as direct ones in a scientific context, but that they should be
avoided in a dialectical debate. His reason is that the opponent might refuse to acknowledge the
‘impossibility’ unless the falsehood is overwhelmingly evident, and hence also refuse to accept the
desired conclusion. This indicates, I think, that the passage was probably written before Aristotle had
developed the formal analysis of reductio-arguments that he offers in An.Pr. A 23; for even a debater
could hardly deny the ‘impossibility’ of an explicit contradiction. On the other hand, Aristotle was well
aware that indirect proofs were common in science (that is, in mathematics)—his own preference for
direct demonstrations is based on epistemological considerations, not on reasons of formal validity.
       71

syllogisms (A 29,45b36–46a2). At the same time, it may also show, depending on

the exhaustiveness of the lists, where a syllogism cannot be found.
Another advantage, not emphasized by Aristotle himself, is that it is arguably
more general in its scope¹⁶ and certainly much easier to use than the system of the
predicables that Aristotle had used to organize his materials in the Topics. It is
hard to imagine an orator holding forth about questions of genus, species, and
definition, while it is clear that she would be interested in the quantity or quality of
her premises—whether some predicate is or is not true of some or all members of
a given class. Syllogistic shows that the formal validity of an argument can be
established or checked regardless of the specific type of predication that occurs in
its premises and conclusion.¹⁷ Hence special conditions regarding the different
types of predication can be treated separately from a general account of deductive
validity. This certainly simplifies the task of the orator, who can now just study the
system of the syllogistic moods (she might, indeed, memorize them with the help
of mnemonic verses, as the medievals did) instead of delving into the arcane
metaphysical questions that fill so many pages of the Topics. It is not surprising
that Aristotle seems to replace references to dialectic as a general theory of
syllogisms by references to the Analytics in the chapters that deal with argument
in the Rhetoric:¹⁸ Syllogistic provides all that the orator needs in order to construct
valid arguments; she does not have to study up on the predicables. This does not
mean that the orator will have to learn about scientific proof instead of dialectical
argument, of course. The syllogistic figures will have to be part of the equipment of
the dialectician as well; and given that the Analytics do not offer a detailed account
of hypothetical argument, the Topics may still be needed as a handbook for the
orator. What is missing from the Analytics is an account of the so-called ‘general’

¹⁶ More general: In the Topics (1.4,101b13–25) Aristotle maintains that all subject–predicate
propositions will contain one of the four types of predication he has distinguished. This claim is
actually the basis of a kind of completeness‑proof in Top. 1.8,103b2–19. However, given term‑conver-
sion, syllogistic will also cover non‑standard cases of predication such as predication per accidens (e.g.
‘The white is a log’, An.Po. A 22,83a5, a14–17), and An.Pr. A 36 may indicate that Aristotle thought it
could deal with propositions that might not fit into the schema of the predicables.
¹⁷ Contrast Top. 1.6,102b27–103a5, where Aristotle acknowledges that there might be considerable
overlap in the treatment of arguments according to the predicables, but insists that one should not try
to find one general method to apply to all forms of proposition. Such a thing would not only be hard to
find, but altogether imprecise (ἀσαφής) and of little use for the present enterprise (102b36–38).
¹⁸ In Rhet. A 2,1356a25–27, Aristotle says that rhetoric is an offshoot of dialectic and political
science, because the orator must be able to construct arguments and must also have some knowledge of
character and the passions. The dialectician has been mentioned just before (1356a22; cp. also
1355b16–17) as the person who knows how to produce deductive arguments (συλλογίσαθαι
δυνάμενος). In A 4,1359b8–12, explicitly referring back to his earlier remark, Aristotle describes
rhetoric as consisting of ‘the science of analytics’ (ἀναλυτικὴ ἐπιστήμη) and ethics (ἣ περὶ τὰ ἤθη).
He continues: ‘and it resembles partly dialectic, partly sophistical argument.’ This passage shows,
I think, that knowledge of the (Prior) Analytics is now supposed to be common to dialectic and to
rhetoric. Both the dialectician and the orator must know the syllogistic moods in order to construct and
criticize arguments. Rhetoric still resembles dialectic, and indeed sophistry, rather than demonstrative
science, in that its premises are ἔνδοξα, plausible or generally accepted rather than necessarily true. But
general expertise about syllogisms now belongs to ‘the science of analytics’, not dialectic.
72    

or subject-neutral topoi. In the framework of the classification of argument-forms

Aristotle uses in An.Pr. A 23, these should probably be described as general types
of ‘hypothesis’ that one might expect one’s interlocutor or one’s audience to agree
to.¹⁹ (An example might be: ‘what holds of one of two similar things, will hold of
the other also’.) The sketch of a formal treatment of arguments concerning things
preferable to their opposites in An.Pr. B 22 may show that Aristotle at some point
planned to work out a formal theory of such arguments, but he evidently did not
get around to this project. However, it invites the suggestion that a revised
handbook for dialecticians might have paid less attention to the predicables
than the Topics does. The Topics, as we have it, is largely a handbook for
Academic dialecticians; a manual based on the syllogistic moods, by contrast,
could equally well have been used by people who had no special interest in genus,
definition, or proprium. Hence the method of heuristics introduced in An.Pr.
A 27–31 is at least as important for dialectic and rhetoric as for the theory of
scientific proof. In fact, although Aristotle does briefly discuss the use of his
method in the context of science in A 30, it seems rather obvious to me that
lists of terms are more likely to be used by a dialectician than by a scientist.
A science, as Aristotle says (46a22–28), will have to begin by assembling, as
completely as possible, the observed facts about its subject-matter. These will
have to form most of the premises and conclusions of its demonstrations, and at
this stage it may be helpful to look for common terms in order to find out what can
or cannot be derived from what. But as the An.Po. shows, this will hardly be
enough to establish, for example, the correctness of a proposed definition, or the
claim that a thing’s being F is to be explained by its being G rather than the other
way around. Alexander was right, I think, to illustrate the use of heuristics by an
example from the dialectician’s repertoire.
The next section of An.Pr. A 32–46, deals with ‘analysis’—that is, the question
of how to reduce given arguments to syllogistic form. Aristotle declares at the
beginning (47a5–8) that his previous claims about the universal scope of
the syllogistic figures will be confirmed if it can be shown that it is indeed possible
so to analyse any given argument. This is no doubt correct, but unfortunately
the following chapters do not in fact confirm Aristotle’s optimism. Since it is
already clear how one has to deal with explicitly categorical propositions, this
section concentrates on difficult or recalcitrant cases—examples of obviously valid

¹⁹ I realize that this suggestion needs more argument than I can give it here, and it may be based on
the commentators more than on Aristotle’s explicit remarks. The argument would be roughly this:
Rhet. A 2 (1358a10–17) decribes the ‘common topoi’ as being subject-neutral, as opposed to the so-
called εἴδη, which seem to be plausible premises concerning the fields that rhetoric is supposed to deal
with, for example, virtue, good and evil, pleasure etc. The examples given for common topoi tend to
coincide with the ones that Alexander of Aphrodisias cites for hypothetical arguments (e.g. 265,30ff.) It
might be that the common topoi are schemata from which one could derive appropriate ‘hypotheses’,
for example (assuming that A and B are similar) ‘If A is F, then B is F also’. See Striker, op.cit. fn.10,
chapter 1 in this volume.
       73

arguments that are not syllogistic in form, or fallacies that appear to be formally
valid syllogisms but are not really so because, as Aristotle wants to maintain, their
premises have not been correctly analysed. Aristotle never shows any serious
doubt about the adequacy of the syllogistic moods to capture the form of all
valid arguments, but he often emphasizes the difficulty of uncovering the proper
structure, and the need for more detailed investigations.
From a modern point of view, several of these chapters could almost be read as
collections of counterexamples to demonstrate the inadequacy of syllogistic. But
Aristotle evidently did not see things this way. Chapter A 36 may show most
clearly why he was so confident: There he declares that the technical term
ὑπάρχειν (belonging) covers much more than straightforward cases of predication.
It appears that his method of analysis allows him to treat just about any sort of
proposition as a categorical premise.²⁰ In other words, he thinks that he knows
what the ‘logical form’ of an argument has to be, and does not consider formal-
ization, as modern logicians might, as a way of exploring or discovering different
logical structures. The shortcomings of syllogistic as a way of representing com-
plicated arguments are too well known to need spelling out in detail. With respect
to the possible usefulness of syllogistic, however, it seems more important to
emphasize that, in spite of the difficulties so candidly discussed by Aristotle
himself, analysis or reduction to syllogistic form can be very helpful in cases
where the form of the premises presents no problem. This comes out most clearly,
perhaps, in the discussion of rhetorical arguments from ‘signs’ (B 27) and the
Rhetoric’s brief treatment of λύσις—dissecting an opponent’s arguments (Rhet.
B 25)—which refers back to the Analytics. Aristotle shows that all but one of the
customary types of sign-argument are formally invalid, so that an orator who has
studied the Analytics will know exactly how to show that the argument, even if it
has some merit, will not be sufficient to establish its alleged conclusion. Also, given
some acquaintance with modal premises, one may be able to show that an
apparently cogent objection is not in fact conclusive. For example, if an orator
has used what appears to be a universal premise, say ‘Those who walk around late
in the night are thieves’, and the objection is raised that the generalization does not
hold without exception, one can retort that what one was actually claiming was
only that all those who walk around late at night are likely to be thieves—a
proposition much harder to disprove, since one might have to show that the
alleged suspects are in fact more likely not to be thieves than to be thieves
(cp. Rhet. B 25,1402b25–1403a2. Aristotle does not give an example in this
passage). Despite its limitations, then, syllogistic can be quite successfully used
at times to identify types of invalid argument, and in this respect it can serve an

²⁰ Henry Mendell has recently shown in detail how Aristotle might have thought this worked for
mathematical proofs. See his ‘Making Sense of Aristotelian Demonstration’, Oxford Studies in Ancient
Philosophy 1993.
74    

important function that modern logic also fulfils—namely to spot faulty inferences
or gaps in arguments. I should admit, however, that Aristotle’s attempts to trace
fallacies to faulty ways of rendering the supposedly categorical structure of premises
do not look like an improvement over the treatment of fallacy in the Sophistici
Elenchi.
Now, the use of syllogistic as a diagnostic instrument to discover weaknesses or
faults in arguments seems, as in the case of heuristics, more important for those
who engage in debate or in courtroom pleading than for those who try to
construct scientific proofs. Presumably a serious scientific researcher will carefully
try to avoid fallacious inferences, and therefore use only forms of argument that
she knows to be valid. Given Aristotle’s preconceived notion that proofs just have
to be syllogistic in form, it is not likely that the chapters on analysis were meant to
persuade the reader that mathematical proofs, for example, could be presented in
syllogistic form. Aristotle’s examples tend to be of the sort known from the Topics,
and what is being confirmed, to the extent that anything can be said to be
confirmed here, would seem to be the claim that the syllogistic figures are
adequate to deal with all sorts of arguments beyond the field of scientific
demonstration.
The sections on heuristics and analysis in An.Pr. A complete a project to which
Aristotle briefly adverts in the opening paragraph of A 32 (46b38–47a2).²¹ The
three parts he mentions there are probably the system of the syllogistic moods,
heuristics, and analysis. The contents of book B are more difficult to classify, partly
no doubt because the book contains some notes or sketches that do not cohere
very well with their context. For example, the book begins with a chapter that
seems to offer some corollaries to formal syllogistic and in effect bring in the
syllogistic moods that were later classified as belonging to the fourth figure
(Aristotle notoriously has only three). At the beginning of B 23 (68b9–13)
Aristotle makes a remark that indicates that he has so far been dealing with
dialectical and demonstrative syllogisms and is going to extend his investigation
by applying the doctrine of the syllogistic moods to rhetorical argument as well;
but he does not indicate precisely what was the point of the preceding chapters.
Still, it seems fairly clear that these chapters belong to the general theory of
argument as opposed to the special theory of demonstration. They either try to
establish metatheorems regarding deductive argument in general, or deal with
questions of argumentative strategy. I have already mentioned the unfortunate
theorem to the effect that an indirect argument can always be replaced by a direct
one. But there are better examples as well. One theorem of which Aristotle was

²¹ ‘It is evident from the things which have been said, then, [1] what all demonstrations come from,
and how [ἐκ τίνων . . . καὶ πῶς] and [2] what things one should look to in the case of each problem. But
after these things, we must explain [3] how we can lead syllogisms back into the figures stated
previously [πῶς ἀνάξωμεν τοὺς συλλογισμουσ εἰς τὰ εἰρημένα σχήματα], for this part of our inquiry
still remains’ (tr. R. Smith, with ‘syllogism’ instead of ‘deduction’).
       75

understandably proud is proved in B 2–4, namely that a true conclusion can be

validly derived from false premises.²² One may find it pedantic and tedious that
Aristotle goes through all the syllogistic moods to show how this is possible in
each case, with one or both premises false, when two examples would have
sufficed to establish the general claim. But this should not divert our attention
from the fact that Aristotle’s proof is a considerable achievement. It is one thing
merely to assert that the truth of the conclusion tells us nothing about the truth of
the premises, another to offer a rigorous proof to show that this is so. Aristotle also
shows that if a false conclusion is derived from apparently true premises, then
either one or more of the premises is false, or the conclusion does not follow.
These are points that may appear trivial to readers trained in modern logic, but
they still need to be made in any introductory class on reasoning and argument,
and Aristotle was the first not just to state, but to prove them. The chapters on the
so-called conversion of syllogisms (B 8–10) may look like a debater’s exercise
(cp. Top. 8.14,163a32–36): showing how a given set of propositions and their
negations can be combined in different ways to deduce different conclusions. But
once again it is important to note that formal syllogistic enables Aristotle to prove
his claims, not just to tell his students that they should try out various combin-
ations to see if anything follows. Finally, as Myles Burnyeat has pointed out,²³ the
last section (on rhetoric) shows how syllogistic enabled Aristotle to distinguish
clearly between two sorts of things that would ordinarily have been called proofs,
by Aristotle as well as by his contemporaries, namely valid deductive arguments
and formally inconclusive but relevant evidence used to support or corroborate a
given thesis. Although Aristotle did not go on to develop a theory of evidence, he
obviously recognized that not every thesis can or should be justified in the same
way, and even that formally inconclusive evidence may be just as convincing as an
elaborate deductive proof (cp. Rhet. A 2,1356b22–24).
Despite its obvious limitations as a general theory of deduction, formal syllo-
gistic allowed Aristotle to establish a number of fundamental points in the theory
of argument. The crucial step, of course, was Aristotle’s discovery that the validity
of an argument depends on the form of its premises and conclusion, not their
content or the peculiarities of the subject matter under discussion. For a general
theory of argument, the distinction between formal validity and truth may well be

²² He refers to this theorem three times in other works (Top. 8.11,162a8–11; An.Po. A 12,78a6–13;
EE 1.6,1217a10–17), which is relatively often for a single thesis, and the EE passage emphasizes its
importance: Aristotle explains at length that it makes sense to consider the reasons given separately
from the conclusion derived, because what is supposed to have been shown may often be true, but not
for the reasons given in the argument.
²³ ‘The origins of non-deductive argument’, in J. Barnes et al. (eds.), Science and Speculation,
Cambridge U.P. 1982, 193–238. I suspend judgement on his more recent claim that rhetorical
syllogisms are based on probable rather than necessary consequence (see Burnyeat, ‘Enthymeme:
Aristotle on the Logic of Persuasion’, in D.J. Furley and A. Nehamas, Aristotle’s Rhetoric, Princeton,
NJ 1994, 3–55).
76    

more important than the notion of a deductive system or an axiomatized theory.

I hope to have shown that Aristotle’s formal logic need not be seen exclusively as
an attempt either to provide the underlying logic for an axiomatized theory or to
develop a system of deductive inference. It is at least as much a contribution to a
theory of valid argument in general, and in this role it was correctly described by
the ancient commentators as a tool of philosophy.²⁴

²⁴ Thanks are due to the students in my spring 1993 seminar for discussing the questions raised in
this chapter with me, and in particular to Mitzi Lee for her criticisms of the first draft; to Henry Mendell
for his detailed and helpful comments at the Princeton Classical Philosophy Colloquium in December
1993; to James Allen for his friendly comments at the Methodos conference; and to the participants on
both occasions for further questions and clarifications.
 6
Aristotle’s Three Theories of Argument

This chapter presents a brief outline of the development of Aristotle’s theories of

argument from dialectic in the Topics through the Rhetoric to formal logic in the
Analytics.

At a conference celebrating Aristotle’s birth 2400 years ago, it is no doubt

appropriate to dedicate a session to Aristotle as the inventor of formal logic.
Aristotle’s syllogistic was one of his most important and influential
achievements—so successful, indeed, that it, supplemented by some elements
of Stoic propositional logic, was seen as the whole of logic until at least the end of
the eighteenth century.
After the nineteenth century, syllogistic seemed to be eclipsed by the new
discipline of mathematical logic, but then the modern logicians rediscovered
Aristotle as a genuine predecessor. Naturally, most of their interest was focused
on the work that contains Aristotle’s exposition of syllogistic in the Prior
Analytics.
But since we are here to celebrate its author, it may be worthwhile to take a look
at what one might call the pre-history of formal syllogistic in Aristotle’s works. For
the invention of formal logic did not come to Aristotle as a sudden inspiration out
of the blue. Just like Archimedes’ famous ‘heureka!’, it was preceded by years of
study, and in this particular case we can to some extent follow its chronology. The
Analytics are the third in a series of treatises on argument that Aristotle wrote,
probably still as a member of Plato’s Academy, for the disciplines of dialectic,
rhetoric, and finally scientific demonstration. All three treatises use virtually the
same definition of ‘syllogism’ (valid deductive argument), and their chronological
order is apparent from the fact that the general theory of the syllogism, presented
in the first seven chapters of the Prior Analytics, plays no role in either the Topics
or the Rhetoric (apart from a few passages that are probably later insertions),
while in the Analytics Aristotle explicitly announces that his general theory covers
all three fields. The Rhetoric refers for its theory of argument back to the Topics,
so it is clear that the Topics came first, then the Rhetoric, and only after that

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0006
78    

the Analytics.¹ I think that a look at this sequence of works will also show some of
the differences between Aristotle and his modern successors.

The Topics: Dialectic

The first treatise opens with a grand announcement:

The aim of this study is to find a method with which we shall be able to construct
deductive arguments (συλλογίζεσθαι) from acceptable premises concerning any
problem that is proposed, and when submitting to argument ourselves, not to
contradict ourselves. (Top. I 1,100a18–21)

This is followed by the definition of syllogism:

A syllogism is an argument in which, certain things being posited, something

other than what was laid down results of necessity because of what was laid
down. (I 1,100a25–26)

But the modern reader soon discovers that the scope of the theory is much more
limited than one might expect. It is a manual for a debating exercise that deals
with a very specific kind of thesis. The theses and the premises used to confirm or
refute them must be plausible or reputable (ἔνδοξα), that is, such that they are
either believed by most people or by the experts. Aristotle assumes that all
propositions used in a deduction will be simple subject–predicate statements, in
which the predicate will signify either an accident (or attribute, συμβεβηκός), or
the genus (γένος), or a unique property (ἴδιον), or the definition (ὅρος) of the

¹ This chronology of Aristotle’s logical works was first presented by Friedrich Solmsen in his book
Die Entwicklung der aristotelischen Logik und Rhetorik (Berlin 1929); for a recent discussion, see
J. Allen, ‘Aristotle on Disciplines of Argument: Rhetorik, Dialectic, Analytics’, Rhetorica 25, 2007,
87–108.
My treatment of these works here is evidently sketchy and simplified; for more detailed introductions
to each, see, for example, J. Brunschwig’s introduction to the first volume of his Budé edition (Aristotle,
Topiques I–IV, Paris 1967), C. Rapp’s introduction to his translation and commentary on the Rhetoric
(Aristoteles: Rhetorik, Erster Halbband, Berlin 2002), and R. Smith’s introduction to his translation with
notes of the Prior Analytics (Aristotle, Prior Analytics, Indianapolis 1989) or, for readers who are
familiar with formal logic, P. Thom’s The Syllogism, Munich 1981. For some of the passages from the
Analytics mentioned in this chapter, see also the commentary to my translation of Prior Analytics book
I (Aristotle: Prior Analytics, Book I, translated with introduction and commentary by G. Striker, Oxford
2009).
I have used the following standard texts of Aristotle’s works: J. Brunschwig, Aristote, Topiques, Budé,
Paris, vol. 1, 1967, vol. 2 , 2007; R. Kassel, Aristotelis ars rhetorica, Berlin 1976; W.D. Ross, Aristotle’s
Prior and Posterior Analytics, Oxford 1949. Translations from Topics book I are by R. Smith (Aristotle,
Topics I, VIII and Selections, Oxford 1997), with modifications; from the Rhetoric by J.H. Freese in the
Loeb series (Aristotle, The ‘Art’ of Rhetoric, Loeb vol. 193, 1926 and later), also with modifications.
Translations from the Analytics are my own.
  ’      79

subject. These four types of predicates were called ‘predicables’ in the later
tradition.
The debate would take place between two persons: the answerer or defender of
a thesis, and the questioner. At the beginning, the answerer would choose a thesis
from a pair of contradictories—for example, ‘Is knowledge the genus of virtue, or
not?’. The thesis the answerer would have to defend would be either ‘virtue is
knowledge’ or ‘virtue is not knowledge’. The questioner would then propose
further propositions that the answerer could either accept or reject, answering
‘yes’ or ‘no’, unless he could object that the proposition was ambiguous or in fact
consisted of more than one statement. The aim of the questioner was to get the
answerer to accept a proposition that contradicted or was incompatible with the
initial thesis.
As the name ‘dialectic’ (derived from the verb διαλέγεσθαι, to have a conver-
sation) indicates, this was a systematized version of the kind of question-and-
answer exchange that one finds in Plato’s dialogues, including rules both for
Socrates’ famous method of examining and refuting another person’s beliefs and
for his ways of trying to find definitions. The predicables, as Aristotle himself
points out (Top. I 6,102b27–103a5), are all involved in the search for definitions,
which are supposed to be formed by genus and differentia.
The so-called topoi (τόποι) or ‘places’, listed in books II to VII of the Topics
according to the predicables, are considerations or rules that a questioner may use
to find the premises he needs to reach the desired conclusion, or the answerer to
avoid accepting propositions that might lead to the contradictory of his thesis.
For example, if the subject A of the answerer’s thesis is similar to some other
thing B, the questioner might suggest that what holds of one of two similar things
will hold of the other one also. Then he asks the answerer to agree that if B is F,
A will be F too. Next, he proceeds to show that B is F, and thereby also proves that
A is F (cf. Top. I 18,102b12–19). The rule here would be ‘what holds of one of two
similar things will hold of the other one too’, from which the debater derives what
Aristotle calls the ‘hypothesis’ that ‘if B is F, A will be F too’. Since such
‘hypotheses’ are usually stated in the form of a conditional, it is tempting to
think that the arguments based on the topoi would be in the forms of either modus
ponens (‘if p, then q; p; therefore, q’) or modus tollens (‘if p then q; not q; therefore,
not p’). This is what J. Brunschwig confidently asserts in the introduction to the
first volume of his Budé edition of the Topics (Paris 1967, xli–xlii), and it is indeed
correct from the point of view of modern propositional logic, but it is not likely to
be what Aristotle had in mind. Propositional logic had yet to be invented, and
Aristotle did not recognize complex sentences like conditionals, conjunctions, or
alternations as single propositions that could function as premises in a deduction.
He explicitly stated that both premises and conclusions in dialectical debates
would be simple predicative propositions. So he described the ‘hypotheses’ stated
in the form of if . . . then—sentences as a kind of agreement between the partners
80    

in a debate, to the effect that a proof for (what we would call) the antecedent would
be accepted as a proof of the consequent. The agreement itself is not a part of the
proof, but rather a kind of licence to infer the consequent once the antecedent has
been proved.²
A good example of this kind of ‘argument based on a hypothesis’ can—not
coincidentally—be found in Plato’s Meno (87B–96C): Socrates proposes an
attempt to answer Meno’s question ‘Can virtue be taught?’ with the help of a
hypothesis—if virtue is knowledge, then it is teachable. Meno agrees to the
hypothesis, and Socrates then offers an argument to show that virtue is know-
ledge, but when Meno is ready to accept the conclusion that virtue can indeed be
taught, Socrates turns around and proceeds to argue at great length that virtue
cannot be taught. The result is, as so often in Plato’s dialogues, an impasse
(ἀπορία). I take it that this type of argument was the model for Aristotle’s
collection of topoi. The argument in the Meno could be summarized as consisting
of a modus ponens followed by a modus tollens, but this would leave out the main
part of what Socrates has to say. Similarly, a dialectical argument ‘based on a
hypothesis’ derived from a topos would mainly consist in establishing a propos-
ition from which the desired conclusion could eventually be inferred. Aristotle
never discusses the form of those ‘syllogisms based on a hypothesis’ in the Topics,
perhaps for the good reason that, as the example of the Meno shows, they would
not have to be uniform at all.
It remains an open question exactly how one can decide whether a deductive
argument is valid. In his book on fallacies, usually cited separately as Sophistici
Elenchi but clearly meant to be the last book of the Topics, Aristotle explains that
deductive arguments will be only apparent, not real syllogisms if they are invalid
(ἀσυλλόγιστοι), that is, if their conclusion does not follow of necessity from what
was laid down, but he does not explain any further (168a21–23). In An.Pr.
I 32,47a22–35, where he sets out to show that syllogisms not formulated in the
canonical form of the figures can also be reduced to syllogisms in the three
recognized figures, he offers a few examples of arguments that necessitate a
conclusion, but are not properly formulated, and also remarks that ‘the necessary
extends beyond the syllogism’ (here, of course, understood in the narrow sense).
In chapter 44 (50a16–b4) he turns to the ‘syllogisms (sic) based on a hypothesis’
and ends up recognizing that those arguments cannot be put into the form of his
new official syllogisms in one of the figures, although he admits that it is necessary
to accept their conclusion—‘though not on the basis of a syllogism, but on the
basis of a hypothesis’ (50a24–26). There he also states, as he had already done in
the chapters on heuristics (A 29,45b19–20), that the arguments from a hypothesis

² For a detailed account of this interpretation, see my article—G. Striker, ‘Aristoteles über
Syllogismen “aufgrund einer Hypothese” ’, Hermes 107, 33–50. English translation chapter 1 of this
volume.
  ’      81

ought to be examined and clearly marked (50b2–3). It seems, then that Aristotle
eventually realized that his high hopes for his general theory were exaggerated. It
is a pity that he seems never to have come back to the subject. So much for the
Topics.

The Rhetoric: Rhetorical Argument

Next in the chronology is Aristotle’s Rhetoric. Since it contains much more than a
treatment of arguments, and since its last book deals with style and delivery of
speeches, in the traditional order of Aristotle’s works it is listed alongside the
Poetics and not included in the so-called Organon with the other works considered
to belong to logic. However, in the first book Aristotle insists that argument is
actually the most important means of persuasion, and his account of rhetorical
argument there offers an illuminating contrast to the Topics and also contains
some new elements.
The Rhetoric opens with the statement:

Rhetoric is a counterpart of dialectic; for both have to do with matters that are in
a manner within the cognizance of all men and not confined to any special
science. (Rhet. I 1,1354a1)

Rhetoric and dialectic correspond to one another like the first and last part of a
Greek choral ode; they are similar though not identical. Another similarity
consists in the fact that ‘Rhetoric and dialectic alone of all the arts prove opposites;
for both are equally concerned with them’ (I 1,1354b31). But the two disciplines
have different aims: the Topics is only concerned with refuting or defending given
theses, regardless of context, while rhetoric, of course, aims at persuading an
audience. So Aristotle defines rhetoric as ‘the faculty (δύναμις) of considering
what may be persuasive in reference to any subject whatever’ (I 2,1355b). Aristotle
describes rhetoric in this book as ‘an offshoot of dialectic and the study of
character’ (i.e. ethics) (I 2,1356a). Dialectic, as it were, provides the part that
consists in constructing or refuting arguments; ethics provides the knowledge of
human character that is needed to adapt the speech to its audience. But though
Aristotle refers to the Topics (and sometimes the Analytics, probably in passages
that were inserted later than the first version of the Rhetoric) for the technicalities
of argument, he also adapts this part to its new purpose. One important difference
between the subject matter of rhetoric and dialectic as described in the Topics lies
in the fact that rhetoric usually deals with individuals—persons, actions, policies—
while the theses of the Topics are usually general statements. Also, the orator in a
political assembly will often be advising about future actions to be taken, where
the result is not entirely predictable, or discuss the actions of a defendant in a trial
82    

that may or may not show her guilt. Aristotle keeps the fundamental distinction
between deduction and induction:

Just as dialectic possesses two modes of argument, induction and the syllogism,
real or apparent, the same is the case in rhetoric; for the example (παράδειγμα) is
induction, and the enthymeme (ἐνθύμημα) a syllogism, and the apparent enthy-
meme an apparent syllogism. Accordingly I call an enthymeme a rhetorical
syllogism, and an example rhetorical induction. (I 2,13556a35–b7)

But then he offers a slightly modified definition of the syllogism:

When, certain things being the case, something different results alongside of
them from their being true, either universally or in most cases, such an argument
in dialectic is called a syllogism, in rhetoric an enthymeme. (I 2,1356b16–19)³

Since many of the arguments an orator uses will have premises that state likeli-
hoods rather than known facts, their conclusions will also be likely, that is, hold
only for the most part (cf. I 2,1357a29–b20). Furthermore, while in dialectic all
premises should be made explicit, this is not necessary in rhetoric and might
indeed sometimes appear pedantic. An ‘enthymeme’ (literally, something to keep
in mind), though it can be recognized as a deductive argument, may therefore
sometimes be stated in an abbreviated form, omitting a premise that the audience
will see as obvious. The rhetorical counterpart of an inductive argument is called
an example (παράδειγμα), because it may consists only in mentioning one or two
examples of particular cases similar to the one under discussion to suggest that
what holds in those cases might also hold in this one, as in Aristotle’s example: ‘to
prove that Dionysius is aiming at tyranny because he asks for a bodyguard, one
might say that Pisistratus before him and Theagenes of Megara did the same, and
when they obtained what they asked for made themselves tyrants’ (1357b31–35).
Again, this is no doubt due to the fact that orators are speaking about individuals,
while inductive arguments in the Topics were usually based on a number of kinds.
Perhaps the most significant addition to the orator’s repertory of argument
types are what Aristotle calls ‘signs’ (σημεῖα)—observable facts that may indicate

³ Aristotle omits the phrase ‘of necessity’ (ἐξ ἀνάγκης) in this definition. M.F. Burnyeat has argued in
an influential paper (‘Enthymeme: Aristotle on the Logic of Persuasion’, in Aristotle’s Rhetoric:
Philosophical Essays, ed. D.J. Furley and A. Nehamas, Princeton 1994), that this means that Aristotle
here recognized that there could be a weaker link than logical necessity in a deduction: the premises
make the conclusion probable, but it does not strictly follow. Given that Aristotle regarded the necessity
of the consequence as the main characteristic of syllogisms in the Topics, I doubt that he would even
have called enthymemes rhetorical syllogisms. Besides, the subsequent statement that ‘what
happens for the most part or what is merely possible can only be deduced from similar propositions’
(1357a 27–30) seems to me to confirm the interpretation that Aristotle does not mean to say that the
conclusion need not logically follow, but only that it may be weaker than a straightforward assertion.
  ’      83

some unobserved cause. Signs will occur in the form of inferences, such as ‘This
woman has milk in her breasts; therefore, she must have given birth’ which can
easily be understood as deductive syllogisms (adding the unstated premise that all
women who have milk in their breasts have given birth). Aristotle notes that only
some such signs provide conclusive evidence:

Among signs, some are related as the particular to the universal; for instance, if
one were to say that all wise men are just, because Socrates was both wise and
just. Now this is a sign, but even if the particular statement is true, the argument
can be refuted, because it cannot be reduced to syllogistic form. But if one were to
say that it is a sign that a man is ill, because he has a fever, or that a woman has
had a child because she has milk, this is a necessary sign. This alone among signs
is a tekmērion; for only in this case, if the fact is true, is the argument irrefutable.
(1357b10–20)⁴

But he does not want to exclude the use of inconclusive signs—evidence, even if
not conclusive, can obviously play an important role in court cases. What is
important to note here, then, is that Aristotle is including evidence as a factor in
argument that need not be deductively valid.
What is entirely absent from the Rhetoric, however, is the theory of the
predicables—understandably enough, since one would hardly expect a public
orator to talk about genus and species or definitions. Instead, we find in the
Rhetoric chapters about the most important subjects with which orators will
have to deal in the three kinds of speeches that Aristotle distinguishes: speeches
in political deliberation, accusation or defence in court, and speeches of praise (or
blame) such as, for instance, Pericles’ famous speech about the young Athenians
who had died in battle. In those chapters, Aristotle assembles plausible or widely
accepted views about what is good or bad, expedient or harmful for deliberative
rhetoric, what is just or unjust for forensic rhetoric, and what counts as noble or
shameful for speeches of praise or blame, along with a few general argument-
schemata like those in the Topics, which, however, he finds less important for
rhetoric (cf. Rhet. I 2,1358a10–28). The collections of plausible views are not only
presented as means for finding premises of a particular form, since those beliefs
can obviously be used in various contexts, not only in argument—for example, in
descriptions of the expected results of particular policies, or the character of
persons. So those chapters provide the orator with material for his speeches,
and, in this sense, they have the same function as the lists of topoi in the Topics.

⁴ This is one of the passages in which Aristotle is probably relying on his chapter on sign inferences
in the Analytics (An.Pr. II, ch. 27), to which he refers for a more detailed treatment. This explains the
emphasis on the validity or invalidity of those syllogisms.
84    

But Aristotle must evidently have realized that the predicables did not belong to a
theory of argument in general, as he seems to suggest at the beginning of the
Topics, but only to the very special kind of dialectic that was presumably practised
in the Academy.
This distinction between the theory of argument and the subject matter to be
discussed is, I think, an important step in the direction of the truly formal theory
of valid deductive argument we eventually find in the Prior Analytics.

Prior Analytics: Scientific Demonstration

In the Analytics, Aristotle turns to the demonstrative (or apodeictic) syllogism.

But this time he explicitly separates the general treatment of deductive argument
from the specific subject of his inquiry:

Syllogism must be discussed before demonstration, because syllogism is more

universal than demonstration; for a demonstration is indeed a kind of syllogism,
but not every syllogism is a demonstration. (An.Pr. I 4,25b28–31)

Syllogism in general is treated in the Prior Analytics, scientific demonstration and

scientific knowledge in the Posterior Analytics.
Aristotle distinguishes, as he had already done in the Topics, different kinds of
syllogism by the status of their premises—dialectical and rhetorical premises must
be plausible, but need not be true, while demonstrative premises have to be true,
either as principles of a science or correctly derived from those principles. But the
form of the premises is the same: ‘a syllogistic premise in general will be an
affirmation or a denial of something about something’ (cf. An.Pr. I 1,24a28–b3).
However, the fundamental innovation in the Analytics is a new analysis of the
simple predicative proposition. Aristotle introduces his new terminology on the
first page of the Prior Analytics. Instead of four different predication relations all
represented by the copula ‘is’ (or ‘are’), he now introduces four quantitative
relations that appear explicitly in the proposition:

‘A premise, then, is a sentence that affirms or denies something of something,

and this is either universal or particular or indeterminate. By ‘universal’ I mean
belonging to all or to none of something; by ‘particular’, belonging to some, or not
to some, or not to all; by ‘indeterminate’, belonging without universality or
particularity, as in ‘of contraries there is a single science’ or ‘pleasure is not a
good’. (I 1.24a16–22)

(The premises here classified as ‘indeterminate’ are typical examples of dialectical

premises from the Topics.)
  ’      85

The subject and predicate of each sentence are now equally described as ‘terms’
(ὅροι—a word that Aristotle had used in the sense of ‘definition’ in the Topics): ‘I
call a “term” (ὅρος) that into which a premise is resolved, that is, what is predicated
and what it is predicated of, with the addition of “to be” or “not to be” ’. The
definition of syllogism remains the same as in the Topics, apart from a stylistic
variation in its last clause: ‘And a syllogism is an argument in which, certain things
being posited, something other than what was laid down results of necessity
because these things are so’ (1,24b16–20).
Whether it was the realization that the four predicables belong to the subject
matter of (Academic) dialectic rather than to the theory of deductive argument in
general, or whether Aristotle saw that he needed an account of syllogistic validity
for a theory that was to cover demonstrative deductions as well as dialectical ones,
we do not know. But his new analysis—which I think gave the work its title—
allowed him to introduce the rules of term-conversion which he could use to
prove the validity of arguments in the form of the now canonical ‘syllogism’, that
is, arguments with two categorical premises that have one of their terms in
common. He states the rules of conversion in the second chapter:

. . . it is necessary for the universal privative premise of belonging to convert with

respect to its terms. So, for instance, if no pleasure is a good, then neither will any
good be a pleasure. And the positive premise converts necessarily, though not
universally, but to the particular; for instance, if every pleasure is a good, it is
necessary that some good is also a pleasure. Of the particular premises the
affirmative necessarily converts to the particular, for if some pleasure is a good,
then some good will also be a pleasure; but for the privative premise this is not
necessary. For it is not the case that if man does not belong to some animal, then
animal also does not belong to some man. (An.Pr. I 2,25a5–14)⁵

Term-conversion would not work for propositions that are taken to express one of
the predicable relations, for example, if B is an attribute of A, A cannot be an
attribute of B. If one disregards the specific predication-relation, as in the quan-
tified propositions Aristotle adopts in the Analytics, conversion leads to sentences
that do not fit the Topics pattern; so, for instance, the proposition ‘all swans are
white’ converts to ‘some white (things) are swans’, where the predicate signifies
the subject—in the metaphysical sense—in which the attribute whiteness
resides. In the Posterior Analytics (A 22,83a1–23) Aristotle calls this ‘accidental
predication’—and rules it out for propositions that occur in demonstrations.
In the first seven chapters of the Prior Analytics, Aristotle presented what is
now recognized as the first instance of a logical calculus; a system very limited in

⁵ Conversion is evidently Aristotle’s favourite method for the proofs of validity in the Prior
Analytics; he resorts to indirect proof or ecthesis only where conversion alone will not do.
86    

scope, but flawless from the point of view of modern formal logic. Together with a
few rules of the propositional logic invented by the Stoics in the century after
Aristotle, Aristotle’s syllogistic was seen for centuries as all of logic in general.
It is tempting to see the system of syllogistic as a shining example of a science as
Aristotle himself describes it in the Posterior Analytics—an axiomatized system of
deductions in which every theorem is derived from a few axioms. This is indeed
how Jan Lucasiewicz presented it in his path-breaking book Aristotle’s Syllogistic
from the Standpoint of Modern Formal Logic (2nd edn Oxford 1957), which
revived the interest of logicians and philosophers in Aristotle’s logic. But there
are good reasons for thinking that this is not the way Aristotle himself saw it. Soon
after the publication of Lucasiewicz’s book, several logicians⁶ pointed out that a
better formal model of Aristotle’s syllogistic would be a natural deduction system,
which is based on primitive rules rather than axioms. Apart from the fact that he
would have had to recognize propositional deductions as valid in order to treat
some of the syllogistic moods as axioms, Aristotle seems to have continued to
regard syllogistic, just like dialectic and rhetoric, as a method or skill (τέχνη) that
can be used by scientists, lawyers, orators—indeed in any discipline that has to do
with argument and proof—but that does not have a specific subject matter of its
own. He presents his final and most rigorous version as a kind of introductory
study to the theory of demonstration and scientific knowledge in the Posterior
Analytics, and though he was obviously also interested in the system of syllogisms
itself, proving, for instance, in An.Pr. I 7 that all valid moods in the syllogistic
figures could be derived with the help of the two universal ones in the first figure
alone, or that it is possible to derive a true conclusion from false premises (An.Pr.
II 2–4), his interests even in the later chapters of the Prior Analytics were mostly
epistemological. One indication of this is that he thought that any syllogism would
have to have at least two premises—not because nothing follows from a single
proposition, although he misleadingly says so in An.Pr. I 23,40b35–36, but
probably because he saw the single-premise inferences represented, for example,
by the rules of conversion as not leading to a conclusion ‘different from what was
laid down’ as stipulated in the definition of syllogism, and hence not useable as
arguments. The chapters on heuristics in the Prior Analytics (An.Pr. I 27–29) are
much shorter than in either the Topics or the Rhetoric, and intended for aspiring
scientists who can only use true premises. Instead of patterns for finding possible
premises, Aristotle recommends that for each of the two terms contained in a
proposition one should consult lists of terms that either ‘follow’ or ‘are followed
by’ the respective term, that is, terms that are truly and universally predicated of
the term, or of which the term is truly predicated. He expects that the relevant

⁶ See T. Smiley, ‘What is a Syllogism?’, Journal of Philosophical Logic 2, 1973, 136–54; and for a
formal model, J. Corcoran, ‘Aristotle’s Natural Deduction System’, in J. Corcoran (ed.), Ancient Logic
and its Modern Interpretations, Dordrecht 1974, 85–131.
  ’      87

terms for each science will be found by experience and observation (An.Pr.
I 30,46a17–30). Obviously, unlike modern logicians, Aristotle did not see his
syllogistic as one among several ways of formalizing ordinary language arguments.
All three of Aristotle’s treatises on argument are intended for specific discip-
lines; they all contain sections on finding appropriate premises, and also chapters
on fallacies—subjects that one would not necessarily expect to find in a modern
logic book. On the other hand, a modern logic book would presumably begin with
a complete list of axioms or rules of deduction. But, of course, Aristotle was not
writing a logic textbook—he was quite literally introducing formal logic for the
first time. And although he eventually realized that the formal syllogistic of his
Analytics was not as all-encompassing as he had hoped, it may well be that he did
not return to the problem of the recalcitrant syllogisms ‘based on a hypothesis’
because he thought that such arguments would not occur in a scientific
demonstration.
So the ancient commentators were probably right in describing Aristotle’s
logical works as an Organon—a tool or instrument for philosophers as well as
other scientists. And in this role, elementary logic has remained a part of the
curriculum of every philosopher. Aristotle was indeed the founder of formal logic,
but the reinvention of logic as a science in its own right and a branch of
mathematics came only many centuries later.
 7
The ‘Analysis’ of Aristotle’s Analytics

‘Analytica’ is one of the few titles in Aristotle’s works that the author himself used.
It is not clear what kind of analysis Aristotle had in mind, but it is usually thought
that the title refers to the section at the end of Prior Analytics I (chapters 32–46)
where he discusses ways of putting ordinary language arguments into the form of
syllogistic moods. However, that section consists mainly in the discussion of
recalcitrant examples and is not very successful. It may be more plausible to think
that the title refers to the analysis of propositions into terms and their quantitative
relations that replaced the classification of propositions according to the
‘predicables’ in the Topics.

Aristotle was the founder of formal logic. But his own name for the new discipline
of constructing deductive arguments on any given subject, of which he was quite
proud, was not ‘logic’—this label was already in use in the Academy for one of the
three parts of philosophy (the others were ethics and physics; see Topics
A 14,105b20–25), and that included not only everything to do with language
and reasoning, but also epistemology. Aristotle’s theory of argument developed in
stages, two of which we can still recognize in the extant works. First came the
Topics, a manual of dialectic, the debating technique of defending or refuting
given theses, which probably included the treatise on fallacies; then the four books
entitled ‘Analytics’, associated with scientific demonstration. In the later ancient
tradition, the four books of the Analytics were divided into two parts of two books
each, referred to as the Prior Analytics and the Posterior Analytics, respectively.
The first half of the work contains Aristotle’s most celebrated achievement: the
system of figures and moods now usually referred to as syllogistic; the second half
contains his theory of scientific proof (demonstration) and scientific knowledge.
Now, the title ‘Syllogistic’, that is, ‘On deductive argument’, would have fit both
the Analytics and the Topics. They both use the same definition of syllogism (valid
deductive argument), but the Topics does not yet have the schematic representa-
tions of propositions and the proofs of validity that make the Prior Analytics easily
recognizable as a treatise of formal logic. So Aristotle himself distinguished
between the two treatises by using the title ‘Topics’ for the manual of dialectic,
and ‘Analytics’ for the treatise on scientific demonstration. Why did he choose the

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0007
  ‘  ’   ’   89

label ‘Analytics’ for the second work, and what exactly did he mean by analysis in
this context?¹
Analysis of arguments is first explicitly mentioned in chapter A 32 of the Prior
Analytics as the third part of a project that begins in chapter 4, where Aristotle
explains why he has to deal with syllogism in general before going on to scientific
demonstration: ‘a demonstration is indeed a kind of syllogism, but not every
syllogism is a demonstration’ (25b26–31). At the beginning of chapter 32,
Aristotle notes that two parts of the investigation have been completed; it remains
now to show how all syllogisms can be reduced to the figures:

For if we have considered how syllogisms come about, have the ability to find
them, and then also to analyse existing syllogisms according to the aforemen-
tioned figures, our initial project should have come to an end.
At the same time it will turn out that what we have said before is confirmed
through what we are about to say now, and it will become even more evident that
things are as we say. For all that is true must agree with itself in every way.
(An.Pr. A 32.47a1–9)

Aristotle’s tone is confident and almost triumphant. Rhetorical flourishes like this
are decidedly rare in the Analytics. However, the following chapters hardly justify
Aristotle’s optimism.

¹ An interpretation according to which Aristotle took over the technical term ‘analysis’ from the
Greek geometers, and therefore meant a procedure for finding premises for a given conclusion, was
presented by Patrick H. Byrne (Analysis and Science in Aristotle, New York 1997). But finding the
premises for given conclusions is precisely the procedure discussed in the Topics, whereas the Prior
Analytics begin with what Philoponus aptly describes as synthesis—the composition of valid deductive
arguments. In the Prior Analytics the emphasis is on validity, not heuristics. In two recent articles,
M. Crubellier (‘The Programme of the Aristotelian Analytics’, in Dialogues, Logics and Other Strange
Things, London 2008, 121–47) and M. Wesely (ΑΝΑΛΥΣΙΣ ΠΕΡΙ ΤΑ ΣΧΗΜΑΤΑ, Peitho vol. 3,
2012, 83–113) have addressed the question what Aristotle means by ‘analysis’, but both seem to me to
go in the wrong direction. Crubellier, who bases his interpretation on a passage in Met. Δ 2
(1013b20–21 = Phys. II 3,195a16–19. For this passage see now: M. Malink, ‘Aristotle on Principles as
Elements’, Oxford Studies in Ancient Philosophy LIII, 2017, 163–213) in which Aristotle describes the
premises as the material cause of the conclusion, suggests that analysis consists in finding premises for
an intended conclusion by dividing the conclusion ‘into two distinct propositions by choosing a
convenient middle term . . . freely introduced by the analyst’ (p. 126). But the claim that the premises
are the material of the conclusion is puzzling and highly problematic, since the premises are clearly not
a constituent part contained in the conclusion, and so could hardly be discovered by ‘analysing’ the
conclusion. Apart from this, I have not been able to find a passage in which Aristotle uses the word
‘analysis’ or its corresponding verb in this sense.
Wesely considers only the ‘analysis into the figures’ in the Prior Analytics. He takes this to be
analogous to geometrical analysis, and claims that the diagrams to which Aristotle occasionally refers
were used to illustrate the validity of syllogistic moods in the same way as the construction diagrams of
the Greek geometers illustrated their proofs. But this is a mistake: the diagrams show only the figures,
not the moods, and since there are more inconclusive moods in each figure than conclusive ones, they
show nothing about validity. Fortunately for the modern reader, the proofs of validity are found in the
deductions of chapters 4 to 6 of the An.Pr. and are independent of the diagrams.
90    

Analysis, as the commentator Alexander of Aphrodisias tells us, consists in

resolving a composite whole into its constituent parts or elements. This can be
done in different ways for different purposes—from ‘analysing’ a composite
organic body into simple bodies or elements, to analysing a sentence into words,
then words into syllables and syllables into letters. On the epistemological side, as,
for example, in geometry, analysis is the converse of synthesis, which is the way
from the principles to what comes from them, while analysis is the way back from
the end to the principles (see Al. Aphr. In Aristotelis Analyticorum Priorum
commentarium ed. M. Wallies, CAG II, Berlin 1883, 7.11–22).
Analysing existing arguments into the syllogistic figures, according to Aristotle,
consists in finding the parts that determine to which of the figures an argument
belongs, and to see whether the alleged syllogism is valid or not. Aristotle does not
have a special word for moods as opposed to figures, but by ‘analysis into the
figures’ he clearly also means determining the mood of an argument and its
validity or otherwise. Throughout this and the following chapters, Aristotle uses
the verbs ‘to analyse’ and ‘to reduce’ (ἀνάγειν) interchangeably, but one should
keep in mind that ‘reduce’ does not here have the sense of proving validity as in
the systematic chapters 5 to 7, but rather subsuming or classifying as belonging to
a figure or mood.²
Analysis in this sense is, of course, perfectly easy so long as the argument
consists only of simple categorical propositions, and Aristotle’s instructions for
this case are brief and to the point:

. . . first find the two premises, then divide them into their terms, and take as
middle term the one that is said in both premises; for that the middle term occurs
in both premises is necessary in all the figures. (47a37–40)

The position of the middle term shows to which figure the argument belongs, and
once one has the premises, one can see which one is universal, which particular,
which affirmative, and which negative. This is presumably why Aristotle does not
list the quantifiers or the copula (if there is one) as another part to be identified in
the analysis.
A special and particularly simple case of analysis is treated in chapter A 45:
analysing a mood from one of the figures into a mood from another figure.
Aristotle goes through all the valid moods except Barbara and shows how a
mood can be transformed into another with the same conclusion by converting
one or two propositions. Validity is not at issue here, of course, but the procedure
suggests what one would have to do in the case of an argument that is not

² For the different senses of ‘reduce’, see my paper ‘Perfection and Reduction in Aristotle’s Prior
Analytics’, in M. Frede and G. Striker (eds.), Rationality in Greek Thought, Oxford 1996, 203–20, ch. 4
in this volume.
  ‘  ’   ’   91

formulated in the canonical way: one would have to transform premises and
conclusion into equivalent categorical propositions. And this, as it turns out,
can be quite difficult. Indeed, Aristotle’s first exposition of the steps needed for
analysis turns into a longish digression about the difficulties: sometimes a premise
may not have been stated; sometimes one might think that an argument must be a
syllogism because something follows of necessity from some assumptions, but the
argument is not properly formulated, as in this example: ‘if being a man, a thing is
necessarily an animal, and if an animal, a substance, then being a man, it is
necessarily a substance’ (47a28–30).
In general, says Aristotle, ‘necessity extends beyond the syllogism; for while
every syllogism is necessary, not everything necessary is a syllogism. Thus, if
something results from certain assumptions, one should not try to reduce it
[scil. to the figures] right away . . . ’. This statement is sometimes cited as the
admission that syllogistic does not cover all valid deductive arguments, though
in this context it may mean no more than that one sometimes encounters what
appears to be a valid argument, but finds it hard to discern the corresponding
syllogistic mood.
The following chapters from A 33 to A 46 then deal with some of these
difficulties. Aristotle’s examples are of the kind one finds in the Topics: most of
them have no quantifiers, and some contain singular terms. To a modern reader,
the recalcitrant cases that Aristotle considers look like a collection of counter-
examples to show that the proposed method of analysis will not work, and one
suspects that some of the examples go back to Aristotle’s colleagues in the
Academy.
It would take too long to go through all these chapters (some of them are merely
brief remarks); so I will concentrate on chapters 36 and 34 for the analysis of
premises, and chapter 44 for the analysis of arguments.
One of the terminological characteristics of the Analytics consists in the
prevalent use of the Greek verb ὑπάρχειν³—usually translated as ‘to belong’—in
the representation of syllogistic premises instead of the more natural forms of ‘to
be’ as copula. This may seem to be just a matter of linguistic convenience (no
switch from singular to plural forms of ‘to be’), but some of the difficulties
Aristotle discusses also stem from the difference between ‘is’ and ‘belongs to’. So
in chapter A 34, Aristotle is faced with an apparent counterexample to one of his
valid moods:

. . . for example, if A is health, what is designated by B, illness, and what is

designated by C, man. One may truly say that A cannot belong to any B (for
health belongs to no illness), and that B belongs to every C (for every man is

³ ὑπάρχειν also occurs in the Topics, though less frequently—for example, where Aristotle wishes to
distinguish between different modes of predication.
92    

susceptible of illness). Now it would seem to follow that health cannot belong to
any man. The reason is that the terms have not been well set out in their
expression, for if one substitutes the terms corresponding to the respective states,
there will be no syllogism (that is, if one puts healthy instead of health, and instead
of illness, being ill). For it is not true to say that being healthy cannot belong to
what is ill; but if this is not assumed, no syllogism comes about, except perhaps
for a possibility. But this is not impossible, for it is possible that health should
belong to no man. (48a1–15)

Here Aristotle in effect proposes a reinterpretation of the first premise, presum-

ably based on the fact that the proposition ‘illness belongs to every man’ is
equivalent to ‘every man is ill’, not to ‘every man is illness’. The details of the
example are obscure, but it incidentally also shows that ‘to belong’ does not always
stand for the copula ‘is’.⁴
In chapter A 36, Aristotle actually suggests that this may be an advantage. At
the beginning of the chapter, he declares:

That the first belongs to the middle, and this to the extreme, should not be taken
to mean that the terms will always be predicated of one another, or the first of the
middle in the same way as the middle of the extreme; and similarly for not
belonging. Rather, one must think that ‘to belong’ signifies in as many ways as ‘to
be’ is said, or as ‘it is true to say’ the same thing. (48a40–b4)

This does not yet show that the use of ὑπάρχειν extends beyond cases of predica-
tion, for the ways in which ‘to be’ is said are presumably the categories (as
Aristotle himself says in chapter A 37, which reads like a preliminary note for
A 36), and those are, as the Greek word indicates, kinds of predication; and what
can truly be said of something should also be predicable of the same thing. But the
example Aristotle then offers shows what he has in mind:

For example, of contraries there is a single science. Let A be ‘there being a single
science’; let the things contrary to one another be designated by B. Now A belongs
to B, not in the sense that contraries are there being a single science of them, but
that it is true to say of them that there is a single science of them. (48b4–9)

⁴ It is not so clear that the first premise is equivalent to ‘nothing that is ill can be healthy’—it
probably means that health cannot be illness. The example is likely to come from a Platonist who
remembered that contraries may belong to the same thing, but cannot possibly belong to each other
(see Plato, Phaedo 102B–103A). Aristotle notes at 48a20–22 that ‘this does not agree with what we said
before since <we said that> when it was possible for several things to belong to the same subject, they
could also belong to each other’—that is, the mood Darapti QQQ would be invalid.
  ‘  ’   ’   93

So far so good, one might say—one could still reconstruct this as a predicative
proposition with a complex predicate, for example, ‘contraries are the object of a
single science’. But although Aristotle had just pointed out that terms need not
always be single words (A 35), he seems to be thinking of the sentence ‘a single
science belongs to contraries’, which could (in Greek as well as in English) be used
to express the more complicated proposition above without linguistic impropri-
ety, and so in the subsequent examples of complete arguments he treats only the
nouns as terms. Now, the verb ὑπάρχειν can, of course, often be used in lieu of the
copula, but in ordinary Greek it also functions, like the English ‘have’, as a stand-
in for various relations, from having a house to having a lawyer to having
the measles. This is why it can also replace the existential use of ‘to be’ with the
genitive, as in the phrase ‘there is an A of B’; but, of course, this does not turn the
relevant sentence into a categorical proposition. Nonetheless, in the following
examples of arguments, Aristotle proceeds as though it did. Here is his first
example:

It happens sometimes that the first is said of the middle, but the middle is not
said of the third. For example, if wisdom is a science, and of the good wisdom is
(scil. the science), the conclusion is that of the good there is a science. Now the
good is not a science, but wisdom is a science. (48b10–14)

There are no quantifiers, but I think it is clear that the premises (which contain a
singular term) should be treated as universal affirmatives. Using only the nouns as
terms, and ‘belongs to’ instead of ‘is’, we would get the following version of the
argument: ‘Science belongs to wisdom, wisdom belongs to the good, therefore, (a)
science belongs to the good.’ Here only the first premise represents a proper
predication, and for all one can see, ‘belongs to’ in the second premise and
conclusion stands for whatever relation holds between a science and its object—
such as ‘studies’, perhaps. But the term-relations of syllogistic were defined as
predications, and so this argument, in spite of appearances, cannot count as a case
of Barbara. What is true of the good is that it is studied by or an object of wisdom,
but if one uses ‘object of wisdom’ as a predicate term in the second premise, then
the argument no longer has the form of a proper syllogism because it lacks a
middle term. Needless to say, Aristotle’s argument as first stated is valid, but his
syllogistic does not have a rule that justifies the inference from ‘x is an object of
wisdom’ and ‘wisdom is a science’ to ‘x is an object of science’.
The other examples in this chapter all illustrate the existential use of ‘is’, in
affirmative as well as in negative propositions, for example, ‘there is no sign of a
sign’. They could in fact be transformed into valid syllogisms by using predicate
terms such as ‘has a sign’, but Aristotle never does this. He seems to think he has
shown what he said—namely that syllogistic premises may have a form that is not
straightforwardly predicative. But if he were right, one would have to accept the
94    

following argument as valid: ‘Of every master there is a slave; of every slave there is
a master; therefore: of every slave there is a slave’ (borrowing an example from
Paul Thom⁵). At the end of the chapter, Aristotle generalizes, suggesting that there
might be other cases—presumably of ‘is’ with the dative or some other inflected
form of a noun—that could be treated in the same way: “For this we say generally
about all cases, that the terms must always be set out in the nominative, such as
‘man’ or ‘good’ or ‘contraries’, not ‘of man or ‘of the good’ or ‘of contraries’, but
the premises must be taken in accordance with the inflections of each.” Paul Thom
charitably interprets this as Aristotle’s ‘brilliant intuition’ that such propositions
are only acceptable as premises if the terms can be put back in the nominative—as
would be the case if one used phrases for predicate-terms—but I think this might
be taking charity too far.
Chapter 36 is perhaps the most intriguing. Most of the other chapters that deal
with premises take off from some error that might lead one to think that a
paradoxical conclusion could be derived (33, 34), or that an argument could not
be analysed (38). Aristotle offers some interesting interpretations of premise
forms, for instance ones that contain his notorious qualifier ‘insofar as’ (often
transliterated from the Latin as ‘qua’) as in ‘the good insofar as it is good’, or ‘being
qua being’ (chapter 38), and a distinction between the forms ‘A belongs to
everything to which B belongs’ and ‘A belongs to everything to all of which
B belongs’ as possible interpretations of ‘A belongs to every B’ in chapter 41, as
well as the notorious distinction between ‘is not A’ and ‘is not-A’ in chapter 46.
These chapters, then, do not look like a systematic attempt to classify different
types of proposition or to show how they should be transformed into categorical
premises; they look more like a discussion of specific difficulties that might have
come up in conversations between Aristotle and his colleagues or students. We
hear from Alexander of Aphrodisias (379,9–11), for example, that Theophrastus
adopted the interpretation ‘A belongs to everything to all of which B belongs’ for
propositions of the form ‘of what B is said, A is said’—that is, probably for universal
affirmative propositions. Aristotle still seems to have no doubt that all premises
could in principle be stated as categorical propositions; and, of course, there is no
mention of propositionally composite forms like disjunctions or conditionals, since
Aristotle did not recognize these as distinct forms of proposition.
However, he quietly admits that his grand generalization does not hold up once
he comes to the analysis of an argument-type rather than a kind of premise,
namely what he calls ‘syllogisms from a hypothesis’ in chapter 44. In modern
terminology, these would be described as arguments in modus ponens or modus
tollens on the one hand, indirect argument or reductio ad impossibile on the other.

⁵ The example comes from P. Thom, The Syllogism, Munich 1981, 78; for the medieval treatment of
arguments with such premises, see also P. Thom, ‘Termini Obliqui and the Logic of Relations’, Archiv f.
Geschichte der Philosophie 59, 1977, 143–55.
  ‘  ’   ’   95

The ancient commentators, all equally familiar with Stoic logic as with Aristotelian
syllogistic, describe the first without hesitation as arguments with one ‘hypothet-
ical’ premise—that is, a conditional—and one categorical premise. But Aristotle did
not know propositional logic.
Here is his account:

Furthermore, one should not try to reduce the syllogisms from a hypothesis, since
they cannot be reduced from what has been laid down. For they have not been
proved by a syllogism, but are all accepted on the basis of an agreement. For
example, if one had assumed the hypothesis that if there is no single power of
contraries, there is also not a single knowledge, and then one went on to argue
that not every power is a power of contraries—not, for example, of the healthy
and the unhealthy, for then the same thing would be healthy and unhealthy at the
same time. Now that there is not a single power for all contraries has been
proved, but that there is not a single knowledge has not been shown. And yet it is
necessary to agree to this—though not on the basis of a syllogism, but on the
basis of a hypothesis. This part, then, cannot be reduced, but the argument that
there is not a single power can be, for this was perhaps a syllogism after all, while
that was a hypothesis. (50a16–28)

Aristotle had already offered the same account for arguments by reductio ad
impossibile in chapter A 23: they are a combination of a hypothesis or agreement
and a syllogistic deduction.⁶ The hypothesis or agreement—presumably between
the partners in a dialectical debate—is to the effect that if a certain proposition has
been proved, another proposition must be accepted. In chapter 23, Aristotle had
also said that the first proposition is substituted (μεταλαμβανόμενον) for the
demonstrandum—that is to say, it is agreed that a proof for this proposition will
count as establishing the second. Aristotle clearly saw that what he calls the
hypothesis is a part of the argument, and that it is, therefore, (logically) necessary
to accept the conclusion. He points out in chapter 44 that in cases of reductio, no
explicit agreement is needed, ‘because the falsehood [sc. of the syllogistically
derived proposition] is evident’. But he does not go on to investigate this non-
syllogistic type of argument. The conclusion of chapter 44 is one of his unfulfilled
promises:

Many other arguments are brought to conclusion from a hypothesis, and these
should be examined and clearly marked. Now what the differences are between
these and in how many ways a syllogism from a hypothesis comes about, we will
say later. For the moment so much should be evident, namely that syllogisms of

⁶ For a detailed defence of this interpretation, see my ‘Aristoteles über Syllogismen “aufgrund einer
Hypothese” ’; Hermes 107, 1979, 33–50. English translation in chapter1 of this volume.
96    

this kind cannot be analyzed into the figures. And we have said what is the reason
for this. (50a39–b4)

Now, arguments in modus ponens or modus tollens are common in the Topics—so
much so that one sometimes wonders why Aristotle did not discover propos-
itional logic as well as categorical syllogistic. So it is probably no accident that
the chapter on arguments from a hypothesis appears in a section that seems to be
concerned with showing that the arguments discussed in the Topics could also
be dealt with by the new theory of the figures. Chapter 44 shows that this is not the
case—and the triumphant tone of the introduction to chapter 32 is gone.
This section, and also the first book of the An.Pr., seems to end in chapter
A 45 with the brief remark ‘How one must reduce syllogisms, then, and that the
figures can be analysed into one another, is evident from what has been said’
(51b3–5). Chapter A 46 was perhaps added at the end because it deals with the
correct interpretation—or analysis—of certain propositions. In the summary at
the beginning of book B, there is not even a mention of analysis. Throughout
book B, Aristotle considers only arguments with categorical premises.
This brings me back to my initial question: why did Aristotle choose the title
‘Analytics’ for his work? How plausible is it to think that the book got its name
from this third part of the investigation?
Philoponus raises the question in the introduction to his commentary
(5,15–6,1): ‘We must find the reason for the title—why it is “Analytics” and not
“On Syllogism”? . . . If a syllogism is a composition (σύνθεσις), that is, an assem-
bling of several sentences, not an analysis, it is worth asking why Aristotle should
have entitled his treatise on syllogisms Analytics, for it should rather have been
entitled “Synthetics”.’ Philoponus’ first answer is weak—‘because analysis is more
important and more perfect than synthesis, since by analysis every syllogism is
found to have its proper figure’, but the second answer sounds more plausible: ‘He
who knows the analysis also knows the synthesis; for what he has analysed [i.e.
resolved into parts] it is not difficult to put together. But the reverse does not hold.
For even a layman knows how to complete the composition of nouns and verbs
into a sentence and to say, for instance, “Socrates is taking a walk” (περιπατεῖ), yet
he does not know, nor can he say, which word is the noun and which is the verb.’
In any case, I think he is bringing out the point of analysis to the figures: people
may be able to produce valid arguments, but that does not mean that they can
explain why those arguments are valid, or not, as the case may be. But once you
have assigned an argument to a syllogistic mood, you can clearly decide about its
validity or otherwise.
Analysis ‘to the figures’ as described in chapter A 32 presupposes the system of
syllogistic and could, therefore, only come as the final and perhaps crowning part
of Aristotle’s project. I have so far concentrated on chapters 32 to 45 because it is
only there that Aristotle explicitly uses ‘analysis’ as a technical term, but this
  ‘  ’   ’   97

should not mislead us into thinking that analysis consists only in the transform-
ation of premises into categorical form. After all, analysis applies to entire
arguments, and though the case of syllogisms from a hypothesis shows that the
strong claim that all valid deductive arguments can be assigned to the syllogistic
moods does not hold up, this should not blind us to the fact that analysis works
very well indeed for arguments with categorical premises.
There are three passages in the Prior Analytics where Aristotle claims to show
that all syllogisms whatsoever are in the figures: chapters A 23, A 32, and B 23; and
all three illustrate the use of analysis. The first passage actually comes before the
section on analysis: in chapter A 23, Aristotle presented a completeness-proof for
direct deductions with categorical premises by showing that all deductions with
more than two categorical premises could be transformed into a series of two-
premise syllogisms that would then be in the figures.⁷ Here, as in chapter 44, he
recognizes that this will not work for syllogisms from a hypothesis, but he
considers only reductio as an argument from a hypothesis, and at the end includes
it in his general claim by maintaining that it must contain at least one proper
syllogism. In this chapter, the completeness-claim is the main point, but in the
later chapters of both books A and B the focus is more often on assessing validity.
Transforming existing premises into the right categorical form is a preparatory
step that may be needed before one can decide whether an argument is valid.
Chapters A 32 to A 46 mainly illustrate the difficulties of this procedure, and
perhaps Aristotle thought he might get back to those problems. At the end of book
B, however, comes another short section of five chapters in which Aristotle
announces that ‘It should now be explained that not only the dialectical and
demonstrative syllogisms come about through the aforementioned figures, but
also the rhetorical ones, and quite generally any means of persuasion in any kind
of discipline’ (B 23.68b9–13). He then goes through several types of argument that
come up in his Rhetoric—induction, paradigm, apagogic argument, ‘objection’
(that is, proof of the contradictory of an opponent’s assumption), and ‘sign’ (an
argument in which a certain fact is taken to indicate another) and explains them
all in terms of categorical syllogisms. The words ‘analysis’ and ‘analyse’ are absent,
but clearly what Aristotle does here should count as analysis as described in book
A. There is no further discussion of premises that do not have the proper form.
But chapter B 27 (70a3–38) illustrates another form that analysis may take: there
Aristotle reconstructs what he calls signs—that is, arguments that are usually
stated simply as inferences from a single premise such as ‘she has milk in her
breasts, so she must be pregnant’—as complete syllogisms to bring out their tacit

⁷ For a formal version of the proof sketched by Aristotle, see T. Smiley, ‘Aristotle’s Completeness
Proof ’, Ancient Philosophy XIV, 1994, 25–38. At A 42, 50a8 Aristotle explicitly mentions the reduction
of longer arguments that contain more than one two-premise syllogism to chains of basic syllogisms as
a case of analysis.
98    

presuppositions. He shows that only inferences that can be construed as first-

figure syllogisms may count as proofs of their conclusion, others—such as ‘wise
men are good men, for Pittakos is wise’—may contain some truth, but are not
valid as inferences. The tacit premises (‘all women who have milk in their breasts
are pregnant’ in the first example; ‘Pittakos is a good man’ in the second) are
indeed obviously assumed by those who use the sign arguments, and Aristotle’s
analysis shows why the first one is incontrovertible if the premises are true, while
the second can be refuted even if the conclusion is true—from the fact that
Pittakos is both a good man and a wise man, it does not follow that all wise
men are good.⁸ That chapter has been deservedly praised for its distinction
between deductive proof and possibly relevant but inconclusive evidence.⁹
References to the Prior Analytics in other works also show that judging the validity
of arguments is the function of analysis—so, for example, at Posterior Analytics
(91b13) Aristotle refers to ‘the analysis concerned with the figures’¹⁰ for the
statement that Platonic division cannot prove the truth of a definition, and at
Rhet. (II 25) he appeals to the Analytics for the inconclusiveness of certain sign-
inferences.
Analysis as a method of assessing the formal validity of arguments, both in
constructing one’s own and in arguing against others, no doubt constitutes what
Aristotle calls—just once, in the Rhetoric (1359b10)—‘analytical science’, and
regards as an indispensable part of properly technical rhetoric. In order to decide
about an argument’s validity, one must bring it into canonical form. Sometimes
this may require reformulating premises, sometimes supplying a premise that has
not been stated, and sometimes it means transforming a long deduction with
many categorical premises into a series of basic two-premise syllogisms. As a
method of judging the validity of arguments, formal logic has remained on the
curriculum of all students of philosophy, not just mathematicians. And this
function of logic also explains, I think, why Aristotle chose the title ‘Analytics’
for his work on the general theory of valid deductive argument.
This accounts for the title of what we now call the Prior Analytics, but what
about the second half of the work, now called the Posterior Analytics? Here
epistemology comes very much to the fore, since Aristotle now deals with dem-
onstration, that is, scientific proof, and scientific knowledge, both in the sense of
scientific theory and in the sense of an intellectual capacity. For the inclusion of
this part, it is not enough to appeal to chapter A 4 of the An.Pr., where Aristotle

⁸ This passage should be compared with the similar discussion of sign inferences in Rhet.
I 2.1357b1–1358a2.
⁹ See M. Burnyeat, ‘Enthymeme: Aristotle on the Rationality of Rhetoric’, in A.O. Rorty (ed.), Essays
on Aristotle’s Rhetoric, Oakland, CA 1996, 88–115. On Aristotle’s sign-inferences in general, see J. Allen,
Inference from Signs, Oxford 2001, study I, 13–86.
¹⁰ The passage he refers to is chapter A 31 of the An.Pr., but since the chapter divisions are not
Aristotle’s own, this may well be a reference to the first book of the Analytics as a whole.
  ‘  ’   ’   99

states that syllogism in general must be studied before scientific demonstration

because every demonstration will be a syllogism, but not every syllogism is a
demonstration. After all, he could have called the first part Analytics, the second
Apodeictic, as some of the Greek commentators do,¹¹ and there are plausible
reasons to think that the two halves of our Analytics originally formed two
separate treatises.¹² Yet there can be no doubt that it was Aristotle himself who
decided to put them together under the title Analytics, since references to this
work are as often to the Posterior as to the Prior Analytics.
There is no ‘analysis into the figures’ comparable to the last chapters of An.Pr.
book I in the Posterior Analytics. Aristotle assumes throughout that all proposi-
tions of a scientific theory would be categorical in form, and that all demonstra-
tions, regardless of their length, would consist in a sequence of syllogistic steps.
Now, ‘analysis into the figures’ could only be treated in the Prior Analytics after
the systematic chapters A 4–6, where Aristotle establishes the list of valid moods
in all three figures. Once an argument has been put into canonical form, its
validity or invalidity can be easily determined by comparing it with the list. But
as we saw before, this does not mean that there had been no analysis in the earlier
chapters, or that analysis consists only in transforming arguments into canonical
form. Aristotle himself refers back to his proof of completeness in chapter A 23 as
a case of analysis (A 42, 50a5–10, cp. above p. 97, n. 7). And, indeed, the most
important case of analysis in the sense of identifying the logical structure of an
argument must have been—the discovery of the figures and moods themselves.
This was the analysis of ordinary language arguments with categorical premises
and conclusion, and it began with a new account—or, as one might as well say, a
new analysis—of simple propositions that distinguishes the Analytics from the de
Interpretatione. In the de Interpretatione, a declarative sentence is said—following
Plato—to consist of a name (ὄνομα) and a verb (ῥῆμα; literally just an utterance or
a saying, but adapted for the purpose of distinguishing the subject from the
predicate expression). Aristotle makes it clear that these are two distinct and
mutually exclusive kinds of expression, so that a name cannot function as a
predicate, nor a verb as a subject.
In the Analytics, the terminology of names and verbs has disappeared. Instead,
Aristotle now describes a proposition as ‘a sentence that affirms or denies some-
thing of something, and this is either universal or particular or indeterminate. By

¹¹ See Philoponus, In Aristotelis Analytica Posteriora commentaria, 334–5. Philoponus also claims
that the word ‘analysis’ is used in a different sense in the Posterior Analytics, but the various senses he
lists there are implausible, and the proposals were no doubt prompted by the absence of explicit
discussions of analysis in these books.
¹² For a recent discussion of this question, see J. Barnes, ‘Proof and the Syllogism’, in E. Berti (ed.),
Aristotle on Science: The ‘Posterior Analytics’, Padua1981, 17–59; and R. Smith, ‘The Relationship of
Aristotle’s Two Analytics’, Classical Quarterly 32, 1982, 327–35.
100    

“universal” I mean belonging to all or to none of something; by “particular”,

belonging to some, or not to some, or not to all’ (An.Pr. A 1.24a16–17).
These propositions consist of what he now calls terms: ‘I call a term that into
which a proposition is resolved (διαλύεται), that is, what is predicated and what it
is predicated of, with the addition of “to be” or “not to be” ’.¹³ Terms, unlike names
or verbs, can function as either subject or predicate. The phrases ‘belonging to all’,
‘belonging to none’ etc. replace the four ‘predicables’ distinguished in the Topics—
definition, genus, proprium, and accident. With these two changes, Aristotle has
introduced the terminology of his general theory of deductive argument, which
covers all types of predication and overrides the difference in function of names
and verbs. His new account of simple propositions allowed Aristotle to assign the
shortest forms of argument—those with two premises that have a term in
common—to exactly three ‘figures’ according to the position of the terms in the
premises: either AB/, or BA/BC, or AB/CB. And the switch to terms that can
function in both subject and predicate position also allowed him to introduce the
deduction-rules of conversion: if A belongs to no B, then B belongs to no A; if
A belongs to some B, then B belongs to some A; if A belongs to all B, then
B belongs to some A. Conversion was Aristotle’s favourite method of proving the
validity of moods beyond the first-figure ones, which he took to be evidently valid
given the meaning of the expressions ‘belongs to all’ and ‘belongs to none’. With a
complete list of possible two-premise arguments (in later terminology, the moods)
and the rules of conversion and reductio ad impossibile, he could establish the list
of valid moods that he later used as a standard of validity. Or, in anachronistic
words, he could show that the validity of an argument depends on what is now
called its logical form¹⁴—that is, in the case of syllogistic, the relations between the
terms, not the objects referred to or described by them.
In the Posterior Analytics, Aristotle used the system of syllogistic itself in the
investigation of deductive arguments from an epistemological perspective.
Assuming that all scientific proofs must be syllogistic in form, he could charac-
terize the indemonstrable first premises of demonstration as ‘immediate’ (ἅμεσα),
that is, such that no term could be found that might serve as a middle term in a
demonstration of the relevant proposition, and he then tried to prove that any
chain of syllogistic inferences must have a limit both in the upward sense (to
premises of greater generality) and in the downward sense, ending with indivisible
species.¹⁵ In fact, such epistemological uses of syllogistic occur already in the Prior
Analytics, for example, in book B, where Aristotle proves that though a false
proposition cannot be inferred from true premises, a true proposition can be

¹³ The Greek text seems to be corrupt at this point; for the difficulties, see my note ad loc. in
G. Striker, Aristotle, Prior Analytics Book I, Oxford 2009.
¹⁴ For the notion of logical form, see J. Barnes, ‘Logical Form and Logical Matter’ (1990); repr. in
J. Barnes, Logical Matters, Oxford 2012, 43–143.
¹⁵ For this proof, see J. Lear, Aristotle and Logical Theory, Cambridge 1980, chapter 2.
  ‘  ’   ’   101

derived from false premises (An.Pr. B 2–4), and in book I, where he shows that the
Platonic method of division cannot be used to prove a definition (A 31, to which
he refers back at An.Po. B 5.91b12–14). As these examples show, the Posterior
Analytics as we have them cannot be understood without the general syllogistic,
which allowed Aristotle to develop an account of scientific knowledge and dem-
onstration that is far more detailed and precise than the outlines one can find in
Plato. Aristotle had good reasons for treating the Prior Analytics and the Posterior
Analytics as a single work. And though we cannot be sure whether the ancient
commentators were right or wrong in thinking that the common title is derived
from the section on analysis in the first book of An.Pr., it is at least as plausible to
think that Aristotle chose his title with a view to the fundamental analysis that
made it possible to develop his formal system of syllogistic.¹⁶

¹⁶ I am gratefull to Alex Crager and Stephen Menn for discussing an earlier version of this paper
with me.
 8
A Note on the Ontology of Aristotle’s
Categories, Chapter 2

This chapter argues that the description of Aristotle’s ontological classes in chapter
2 of the Categories contains an implicit criticism of Plato by distinguishing
between the relations of belonging to a species or genus and having an attribute.
This distinction is then used in chapter 5 of the Categories to argue for the
ontological primacy of individual substances.

Let me begin with a few points about the Categories that seem to be agreed
among scholars: this fragmentary little treatise, which seems to be largely
concerned with ontology, nevertheless belongs to the Organon, the set of
works that make up what one might call Aristotle’s theory of argument, and
the Organon (all of it, I am inclined to think) was written while Aristotle was
still in the Academy. The traditional title Categories is probably late and the
treatise is clearly not a finished work, but the alternative title ‘Introduction to
the Topics’ (πρὸ τῶν τόπων) confirms that what we have is associated with the
Topics.¹ The Topics, in turn, is pretty clearly an Academic production—a
manual for dialectical exercises focused on definitions and their parts. But
the Topics also seems to have a thoroughly un-Platonic ontology, and so it
would not be surprising if Aristotle had provided a kind of metaphysical
preface—preferably one in which he offered some arguments for his departure
from the Platonic scheme of Forms vs. particulars. In fact, the reason why
many of us still find it convenient to begin a lecture course on Aristotle’s
metaphysics with the first five chapters of the Categories is precisely that these
chapters seem to present so clearly the contrast between Aristotle and his
teacher: Aristotle bestows the honorific label οὐσία—usually translated as
‘substance’ or ‘essence’ in Aristotle’s works, ‘being’ or ‘what really is’ in
Plato’s—on just those things that Plato tended to treat as less real, less
enduring etc. than his Forms.

¹ For these points, see M. Frede, ‘The Title, Unity, and Authenticity of the Aristotelian Categories’, in
M. Frede, Essays in Ancient Philosophy, Minneapolis 1987.

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0008
       ’  ,   103

Now, there is no explicit attack on Plato in the Categories, and chapter 5 begins
with a bold assertion announcing the reversal of priorities. Aristotle does, of
course, then go on to offer quite a series of arguments to defend his claim, using
the terminology he has introduced for this purpose in chapters 2 to 3. But this
might be seen as question-begging—the terminology seems to stack the cards in
favour of particular subjects as the only ‘independent’ kind of entities, and so one
might be tempted to think that the dispute about what counts as basic has no clear
answer. Gail Fine, for example, at the end of her meticulous examination of the
arguments from Aristotle’s lost treatise on the Platonic Forms (Περὶ ἰδεῶν),²
concludes that whether one should side with Plato or with Aristotle may often
be hard to decide and could be a matter of different intuitions. That may be true in
a way, though I would be inclined to say it is not so much because of different
intuitions as because the question of priority may itself be a little antiquated or
odd: why insist on singling out precisely one class of entities as basic when it
makes most sense to start with—at least—two? So I will not try to argue here that
Aristotle should be regarded as the winner in this particular debate. All I would
like to point out is that while there is indeed no explicit polemic against Plato, the
somewhat self-serving terminology of Categories chapter 2 can be seen as raising a
legitimate question about Plato’s two world—ontology that justifies at least
another look at Plato’s enthusiasm for the ontological priority of the Forms.
What Aristotle points out in chapter 2, and what a fellow member of the
Academy would probably have recognized, is that the technical term ‘participa-
tion’ (μέθεξις), and some of the others that Plato uses for the same purpose in his
dialogues, actually cover two distinct relations that do not coincide, so that Plato
should, in fact, have recognized four classes of entities instead of his
customary two.
Aristotle introduces his fourfold classification of ‘things there are’ as follows
(2.1a20–b6):

Of things there are,

some are said of a subject but are not in any subject;
man, for example, is said of a subject, the individual man, but is not in any
subject.
some are in a subject but are not said of any subject . . .
for example, the individual grammatical knowledge is in a subject, the soul, but is
not said of any subject
Some things are both said of a subject and in a subject—
knowledge, for example, is in a subject, the soul, and is also said of a subject,
grammatical knowledge

² See G. Fine, On Ideas, Oxford 1993, 28–9 and 241.

104    

Some things are neither in a subject nor said of a subject—

for example, the individual man or horse, for nothing of this sort is either in a
subject or said of a subject. (emphasis mine)

Both ‘being in a subject’ and ‘being said of a subject’ are defined more narrowly
here than in other contexts. Aristotle explains that something is said to be ‘in
something as a subject’ if it is ‘in something not as a part and cannot exist
separately from what it is in’; and something is ‘said of a subject’ in the sense
assumed here if ‘all things said of what is predicated will be said of the subject too’
(3.1b11–12). Commentators from ancient times on have been quick to translate
the cumbersome and unusual terminology of ‘in a subject’ and ‘said of a subject’
into Aristotle’s customary technical terms: the first class comprises species and
genera of substances, the second, individual attributes, the third, universal attri-
butes, and the fourth, individual substances. Ackrill³ points out that the two
phrases Aristotle uses here hardly occur as technical terms except in the
Categories—and there, one might add, mainly in chapter 5, where Aristotle
discusses substance. Some ancient commentators noticed that Aristotle’s termin-
ology was designed to focus on the two different relations, as shown by
Simplicius’s comment (In Cat. 45.19–23):

But why did Aristotle not use the customary expressions, speaking of ‘universal
substance’ or ‘particular attribute’ and so on? Perhaps because he gave an outline
of each in clearer words; for ‘being in a subject’ conveys the essence of attribute,
as ‘<said> of a subject’ conveys that of universal.⁴

To be ‘in’ something as a subject is to be an attribute, to be ‘said of something as a

subject’ is to be a universal. What Simplicius does not say, of course, is that this
might have anything to do with Platonic μέθεξις—but then Simplicius was a
Neoplatonist, who presumably assumed that Plato and Aristotle agreed in their
metaphysical doctrines. If we focus on the two relations of being a universal and
being an attribute, we get the following descriptions of the four classes:

(1) universals that are not attributes

(2) attributes that are not universals
(3) attributes that are universals
(4) individuals that are neither attributes nor universals

³ J.L. Ackrill, Aristotle’s Categories and De Interpretatione, Oxford 1963, 74.

⁴ Τί δήποτε δὲ μὴ κατὰ τὸ σύνηθες ἐχρήσατο τοῖς σημαινομένοις ὁ Ἀριστοτέλης, καθόλου οὐσίαν
λέγων καὶ μερικὸν συμβεβηκὸς καὶ τὰ λοιπά; ἢ ὅτι δι ̓ ἐμφανικωτέρων ὀνομάτων ὑπογραφὴν ἑκάστου
παρέδωκεν: τὸ γὰρ ἐν ὑποκειμένῳ εἶναι τὴν οὐσίαν τοῦ συμβεβηκότος παρίστησιν, ὥσπερ τὸ καθ
̓ ὑποκειμένου τὴν τοῦ καθόλου.
       ’  ,   105

The technical terms ‘universal’ (καθόλου) and ‘attribute’ (συμβεβηκός) are also
Aristotle’s, not Plato’s, but re-stating his classification in this way brings out the
point that he is introducing two relations where Plato apparently saw only one.
Both species-terms and attribute-terms are predicated of subjects, and Plato used
to describe the underlying fact or state of affairs as a case of a thing participating in
a Form. The order in which Aristotle introduces his four classes highlights his
criticism of the Platonic story. The first class, natural kinds or species of
Aristotelian primary substances, is a class of universals that are not attributes.
Now, the Phaedo seemed to introduce forms precisely as attributes—see Phd.
74a9–c10; and note the comments at 103a4–22: ‘Then, my friend, we were talking
about things that have the opposites, calling them by the names they take from
them; whereas now we are talking about the opposites themselves from whose
presence in them the things so called derive their names.’⁵ But if participation is
the subject–attribute relation, where do Forms like man (ἄνθρωπος) go? And if
Socrates is a man because he participates in humanness, what in the world is
Socrates, the subject, if not a man? In Plato’s Parmenides, Socrates confesses that
he is not sure whether he should assume that there are Forms of man, fire, water,
and so on (130c4–d8), but it is clear that Plato eventually decided that there must
be such Forms, given that Forms are supposed to be what is common to many
things that are described by the same word.⁶ Perhaps, then, one should say that
Forms are primarily universals—what many things called by the same name have
in common and the objects of definition; and perhaps μέθεξις should be seen as
the relation of particular to universal, so that the class of Forms should comprise
both attributes and kinds? Now look at the description of Aristotle’s second class:
there seem to be some attributes, such as individual instances of literacy, that are
not universals; so one cannot say that the particular–universal relation is funda-
mental and attributes are a subclass of universals.⁷ In other words, the two
distinctions Plato uses to mark the contrast between Forms and the things that
participate in them do not yield only the two classes Plato seems to recognize; he
should admit at least four. Aristotle’s last two classes—universal attributes and
particular subjects—are, in effect, the ones that Plato usually does recognize, and
hence, I suppose, they are put at the end of the list, with some special emphasis for
the last item provided by the terminology Aristotle has chosen to use. This

⁵ τότε μὲν γάρ, ὦ φίλε, περὶ τῶν ἐχόντων τὰ ἐναντία ἐλέγομεν, ἐπονομάζοντες
αὐτὰ τῇ ἐκείνων ἐπωνυμίᾳ, νῦν δὲ περὶ ἐκείνων αὐτῶν ὧν ἐνόντων ἔχει τὴν ἐπωνυμίαν τὰ ὀνομαζόμενα.
⁶ See, for example, Rep. 10,596A6–7: ‘For surely we usually posit one Form for each multitude of
things to which we apply the same name’ (εἶδος γάρ που τι ἓν ἕκαστον εἰώθαμεν τίθεσθαι περὶ ἕκαστα τὰ
πολλά, οἷς ταὐτὸν ὄνομα ἐπιφέρομεν).
⁷ I realize that there is a dispute among scholars as to whether individual attributes should be taken
to be instances of attributes in individual subjects or maximally specific properties that can occur in
several subjects. If, as I suggest, these entities were introduced in the context of Platonic division, it
would be most plausible to take them as instances in individual subjects. For a detailed discussion of the
controversy, see M. Wedin, Aristotle’s Theory of Substance, Oxford 2000, 38–66.
106    

terminology is, as it were, a mirror image of Plato’s, with the accent on the
participants: where Plato would say that particulars bear the names of the
Forms in which they participate or are ‘called after’ them (ἐπωμυμίαν ἔχει),
Aristotle prefers to say that attributes and kinds are either ‘in’ or ‘said of ’ subjects.
That Aristotle was thinking of Platonic ‘participation’ is confirmed by a defin-
ition of the verb ‘to participate’ (μετέχειν) he uses several times in the Topics
(4.121a11–12; cf. 5.132b35–133a11; 4.126a17–25). It shows, I think, that Aristotle
was not the first nor the only member of the Academy to point out the difference
between attributes and universals: μετέχειν is defined as ‘admitting the definition
of the thing in which something participates’,⁸ and that is exactly how Aristotle
uses the phrase ‘to be said of something as a subject’ (cf. Cat. 5,2a19–21: ‘It is clear
from what has been said that if something is said of a subject, it is necessary that
both its name and its definition be predicated of the subject’⁹). This definition of
participation should be Academic rather than one introduced by Aristotle himself,
since μετέχειν is not one of Aristotle’s own technical terms. But the definition rules
out cases of participation such as Socrates’ participating in similarity (Prm. 129A)
or in plurality (Prm. 129C), since the definitions of similarity or plurality clearly
cannot be predicated of a particular person. The definition seems to show that
Aristotle and some of his colleagues decided to opt for the particular–universal
relation rather than the subject–attribute relation in defining participation, and
this might, for example, be due to the emphasis on division of genera into species
(διαίρεσις), evident in Plato’s late dialogues. Aristotle’s class of individual attri-
butes would offer a plausible answer to a question that naturally arises from
Plato’s insistence in the Philebus (16C5–E2) that Forms or genera are both one
and indefinitely many. What, one might ask, are the many that participate, say, in
colour—things that have colour, or the instances of colour ‘in’ those things? If one
is thinking of division, it might be plausible to opt for individual instances of
colour rather than the things that have the colour, given that neither Plato nor
Aristotle seem to make a clear distinction between the relations of individuals to
their species and species to their genus. Plato himself, on the other hand, had
spoken of tallness and the like being ‘in’ persons in the Phaedo, so that it would
have appeared plausible to use the ‘in’-locution for attributes. But if one settles for
a definition of μετέχειν that excludes the subject–attribute relation, then Socrates’
‘participating’ in similarity becomes more complicated—we would now have to
say that he has in himself an item that participates in similarity. Now, a Platonist
might conceivably be willing to recognize a new class of particulars, but the
puzzles raised about participation in the Parmenides would no doubt have been
exacerbated by recognizing two different kinds of participation.

⁸ ὅρος δὲ τοῦ μετέχειν τὸ ἐπιδέχεσθαι τὸν τοῦ μετεχομένου λόγον.

⁹ φανερὸν δὲ ἐκ τῶν εἰρημένων ὅτι τῶν καθ ̓ ὑποκειμένου λεγομένων ἀναγκαῖον καὶ τοὔνομα καὶ τὸν λ
όγον κατηγορεῖσθαι τοῦ ὑποκειμένου.
       ’  ,   107

To return to the chapter of Aristotle’s Categories from which we started: the

Aristotelian twist, as it were, comes with Aristotle’s insistence that Socrates’
relation to the kind humanity cannot be explained by saying that he has an
instance of humanness in himself—presumably for the good reason that
Socrates already is a man. In other words, things’ being what they are cannot be
explained in the same way as their having certain distinguishable features, and
they must already be something determinate and identifiable in order to admit of
the kind of explanation that Plato offers in the Phaedo for ‘coming-to-be-and--
passing-away and being’. Attributes appear to be dependent on determinate
subjects, and the relation between these subjects and the species to which they
belong cannot be understood on the model of the subject–attribute relation.
This, I would like to suggest, is the point where a question of ontological
priority could legitimately be raised—and we know, of course, that Aristotle
wanted to decide it in favour of particular subjects. Let me now briefly turn to
chapter 5 of the Categories to see how far he gets.
Aristotle’s first and most general argument, from 2a34 to 2b6, ends with the
triumphant conclusion ‘ . . . all the other things are either said of the primary
substances as subject or are in them as subjects. So in the absence of primary
substances it is impossible for any of the other things to be’.¹⁰ This conclusion is
notoriously overstated. If we take it to mean, as one naturally would at first sight,
that all other things depend on primary substances (particular subjects) because
they cannot exist without them, but particular subjects can exist without the rest,
the argument seems to fail dismally. For not only would we have subjects without
attributes—‘bare particulars’, as they have been called by some brave philosophers
in the last century—but since the species and genera of substances are included in
‘all the others’, we would be left with what Locke aptly calls ‘something we know
not what’, since whatever it is that is ‘bare’ could not even be described as a
particular.¹¹ This cannot be what Aristotle had in mind, since his primary
substances, as we can see from his examples, are particulars of determinate
kinds, such as individual persons or horses, and those will necessarily have
some attributes, though not necessarily a fixed set. If Aristotle wanted to maintain
that in some sense everything else depends upon particular subjects that are
neither attributes nor universals, he had to be more specific about the relations
he had in mind. As far as existence goes, the dependence appears to be mutual.
Since we know that he continued to hold on to the primacy of his individual
substances, I propose that we look at other parts of this chapter to see whether
there might be some better grounds for his priority-claim. It seems to me that once

¹⁰ . . . τὰ ἄλλα πάντα ἤτοι καθ’ ὑποκειμένων τῶν πρώτων οὐσιῶν λέγεται ἢ ἐν ὑποκειμέναις αὐταῖς
ἐστιν. μὴ οὐσῶν οὖν τῶν πρώτων οὐσιῶν ἀδύνατον τῶν ἄλλων τι εἶναι.
¹¹ Locke, An Essay Concerning Human Understanding, book II, ch. 23, s. 2. Compare Aristotle’s own
thought experiment in Met. Z 3.1029a9–30: if we took away all attributes from a substance, we would be
left with nothing, except perhaps a matter that has no determinate properties of its own.
108    

again the distinction between attributes and kinds (genera and species) had a role
to play. I would suggest that Aristotle’s main reason for counting all attributes as
dependent entities shows up in the section about the proprium of substance, at
4a10–22, which is said to be the fact that only a (primary) substance is such that it
may ‘receive contraries’ while remaining one and the same thing.¹² This is not
presented as an argument for the primacy of particular subjects, but it points to a
fact that was crucial in Plato’s own first introduction of the distinction between
things and their attributes, namely that the same particular thing may appear to be
characterized by opposite attributes. Plato concluded—rightly—that the attributes
must be distinct from the things that have them, but then he went on to elevate the
attributes over their subjects, on the grounds that attributes (taken as universals)
are changeless and eternal (timeless), while subjects can and do change: when we
paint the green fence black, we do not thereby destroy the colour green. Aristotle,
on the other hand, seems to say: when we have painted the fence, the green colour
is gone, but the fence is still there, and for any colour to come to be, there has got
to be some body in which it can be instantiated. Now, this will not establish the
dependent status of all attributes, since some of them, notably the differentiae of a
species, the characteristics that distinguish them from all other species of the same
genus, are such that a subject cannot persist while losing them. But Aristotle’s
discussion of attributes seems to be concerned mainly with contingent ones—he
notes that there is a difficulty with differentiae and decides that they are not ‘in’ a
subject (Cat. 5.3a21–29). This is not a plausible move, since it works only because
Aristotle uses the adjectives ‘two-footed (sc. animal)’ (δίπουν) and ‘(animal) living
on land’ (πεζόν), which make it possible to predicate the definition of the subject
as well. But it would seem natural to say, for example, that rationality (λόγος)
rather than ‘being rational’ (λόγον ἔχον) is the differentia of humans,¹³ and clearly
the definition of rationality cannot be predicated of people. However the difficulty
with differentiae or essential attributes is to be solved, we might agree that
particular contingent attributes do depend on their subjects for their existence,
since it is possible for a subject to exist without having them, but not vice versa. So
while it is not possible for a thing to have no attributes at all, it is possible for a
subject to change its attributes while remaining the same, but not for a particular
attribute to change subject. If one thinks of attributes in the context of change, as
Aristotle evidently does in this passage, one can at least understand Aristotle’s
view about their dependent status. So much, then, for the status of attributes.
But what about natural kinds and their genera? Here I would be inclined to say
that the dependence is indeed mutual—you can’t have a thing that is one without
being one of a kind. Aristotle seems to recognize this to some extent—he grants

¹² Μάλιστα δὲ ἴδιον τῆς οὐσίας δοκεῖ εἶναι τὸ ταὐτὸν καὶ ἕν ἀριθμῷ ὂν τῶν ἐναντίων εἶναι δεκτικόν.
The phrase ἕν ἀριθμῷ (one in number) shows that he is speaking of primary substances only.
¹³ For this point, see An.Pr. A 34.47b40–48a15.
       ’  ,   109

the status of substance to the species and genera of individual subjects, but he also
relegates them to the status of secondary substances, no doubt because though they
may be seen as (generic) subjects of (generic) attributes, they are still universals, that
is, such as to be said of something as of a subject, and thus not ultimate subjects.
And this means that they fail to be ‘a this’ (τόδε τι), as Aristotle argues at 3b10–21.
This criterion for substance-hood remains constant throughout Aristotle’s
works, but it is interesting to see that it is here introduced as a commonly held
view (δοκεῖ, 3b10). Perhaps it also came up first in discussions in the Academy?
For example, in the Sophistici Elenchi. (178b36–179a10) Aristotle claims that the
Third Man puzzle arises because one has erroneously admitted that ‘what is
predicated of all [sc. men] in common’ is a this.¹⁴ Aristotle paraphrases the
expression τόδε τι by ‘individual (ἄτομον) and one in number’. Species and genera,
at least in this treatise, are not one in number.¹⁵ Words like ‘man’ or ‘horse’ cannot
be used by themselves to pick out an individual subject without some demonstra-
tive, and it is presumably true that you cannot point to a universal. Also, one
might say that while sensible particulars can be identified independently of their
species (saying, for example, ‘the white thing there’, cp. Top. A 7,103a30), it would
be difficult even to explain what a species is supposed to be without reference to its
members. However, Aristotle’s subsequent attempt to assimilate these species and
genera to the category of quality (or some other; see the Soph. El. passage) sounds
hesitant and half-hearted, and would surely be a bad idea, since it would obliterate
the distinction between attributes and species that seemed to be at the basis of
Aristotle’s revision of Platonic ontology.
What does not seem to occur to Aristotle, either here or elsewhere, is that species
and genera might be seen as classes or sets, although one should think that this is
actually suggested at least by the word γένος. But in the chapter on γένος in Met. Δ,
Aristotle explicitly distinguishes between the use of the word for a family or group
of animals of the same species, such as mankind, and its use as a generic term that
is the subject of the differentiae (Met. Δ 28.1024a29–b9). And, of course, so long as
one assumes that the genus can be predicated of the species, this interpretation is
ruled out—mankind is not an animal, nor would Aristotle have thought so.
So the status of species and genera of substances remains unclear in this
chapter,¹⁶ and their demotion to secondary status may simply be an early indica-
tion of Aristotle’s reluctance to accept any kind of universals as entities in their
own right.

¹⁴ Καὶ ὅτι ἔστι τις τρίτος ἄνθρωπος παρ ̓ αὐτὸν καὶ τοὺς καθ ̓ ἕκαστον· τὸ γὰρ ἄνθρωπος καὶ ἅπαν τὸ
κοινὸν οὐ τόδε τι ἀλλὰ τοιόνδε τι ἢ ποσόν ἢ πρός τι ἢ τῶν τοιούτων τι σήμαινει.
¹⁵ But see Top. A 7.103a23–31, where Aristotle cites the cases of coat and cloak, two-footed
terrestrial animal, and man as examples of unity in number.
¹⁶ For the ambiguous status of ‘secondary substances’ in the Categories, see most recently C. Perin,
‘Substantial Universals in Aristotle’s Categories’, Oxford Studies in Ancient Philosophy 33, 2007,
125–44.
110    

This is as far as Aristotle gets in the Categories. If I am right about the Platonic
inspiration of chapter 2, then the ontology of this treatise is thoroughly Academic
indeed—the sort of thing one could hope for from a critical Platonist who sees the
point of Plato’s theory, but who thinks it might need some modifications. Which
only goes to show that Aristotle was surely the best Platonist ever.
As for the dispute about priorities, I would be inclined to agree about the
dependent status of attributes: even if not all of them are accidental, one could
invoke the—admittedly epistemological—argument from conceptual priority that
Aristotle himself uses in Met. Z 1 (roughly: in order to define ‘walking’ or ‘healthy’
or indeed ‘two-footed’, you have to mention animals, but not vice versa;
Z 1,1028a34–36) to argue for some kind of secondary status, but this does not
happen in the Categories. Not so for natural kinds, though—and one might
suspect that the reason for the disappearance of the terminology of secondary
substances from Aristotle’s vocabulary was that he himself was not satisfied with
the arguments of Cat. 5. However, the ontology of chapter 2 is exactly what he uses
in the Topics, where species and genera have a much more important role than
concrete individuals. In fact, the distinction between primary and secondary
substances is not even mentioned there.
The question of ontological priority is reopened only in the central books of the
Metaphysics. At this point, it has become much more complicated due to the
introduction of the correlative concepts of matter and form, of which there is no
trace in the Categories. Without wishing to delve into the labyrinth of interpret-
ations of the ‘substance books’, I would like to draw attention to a statement that
appears in the ‘summary’ of chapters 1–2 of book H and also at De Anima 2.1
(412a6–9), and that seems to be a result, however reached, of those tortuous
investigations: Aristotle repeatedly declares that sensible substance is ‘in one sense
matter, in another form, and in a third, the thing that is constituted by these (τὸ ἐκ
τούτων)’. Matter is substance as underlying subject, though only potentially a
‘this’; form is a ‘this’, but separable only in thought; the third is the only thing that
comes to be and passes away, and that is separable without qualification (H
1.1042a26–31). Aristotle does not rank these—presumably for the good reason
that they exist, if and when they do, only together.
It seems to me that this line could in a way be read as a vindication of the basic
status of the concrete individuals designated as primary substances in Cat. 2. Once
matter and form have been identified as the ‘principles’ of natural substances, they
count as substances as well insofar as they are principles of the generally recog-
nized substances.¹⁷ And the relation between the form of a concrete particular and
the particular itself is not the same as that between a concrete particular and its

¹⁷ This may not be a plausible assumption from a modern point of view, since matter and form are
theoretical entities, but it seems to follow from Aristotle’s view that the ‘principles’ of a thing must be
‘prior’ to it. At least that is how I would understand a line like Met. Z 3,1029a30–32.
       ’  ,   111

attributes: it is either identity (if the form is seen as the actuality of the thing) or
that of a constituent inseparable from the individual itself. Separability or inde-
pendent existence did not come up explicitly in the Categories (implicitly perhaps
in the first argument for the primacy of primary substances), but in Met. Z it is
listed as one of the features of substance (Z 1,1028a33–34), and in H 1 it seems to
confirm the basic status of concrete individuals. But while the form–matter
analysis offers a clear distinction between the relations of subject to attribute on
the one hand, particular to form on the other, the individual forms of Met. Z and
H can no longer be identified with the species-universals of the Categories. In Z 13,
Aristotle argues at length that no universal can count as a substance. What
members of the same species have in common, it seems, is not a metaphysical
entity in which they might be said to participate, but simply the fact that the same
definition holds for all of them—a form of words that is synonymous with the
general term that signifies the form and is, therefore, truly predicable of all
corresponding individuals. We might say that these individuals are all tokens of
the same type—but the type itself is no longer seen as a metaphysical entity in its
own right. The genus, in turn, is no longer considered as a whole of which the
species are parts (as originally suggested by the word καθόλου); rather it is now
described as (analogous to) matter that underlies the differentiae (see e.g. Met. Δ
28, cited above). The analogy is perhaps understandable,¹⁸ but not very helpful,
since it does not leave room for the relations of inclusion, overlap, or exclusion
between classes that are so obviously represented in the divisions of a genus
into species—not to mention the four term-relations that underlie Aristotle’s
syllogistic.¹⁹ At this point, then, the Platonic-Academic conception of ‘participa-
tion’ has completely disappeared from Aristotle’s ontology, and it seems that he
never became interested in the ontology of classes. Whether this is to be regretted
or not, I think we still have reason to be grateful to the medieval philosophers who
bequeathed to us the distinctions between substance and essence, form and
species.

¹⁸ For this point, and the difference between the relations of form to matter and attributes to
subjects, see the classic article by J. Brunschwig, ‘La forme prédicat de la matiere’, in P. Aubenque (ed.),
Etudes sur la Métaphysique d’Aristote, Paris 1979, 131–66.
¹⁹ Interpreters tend to assume that the schematic letters in the syllogistic stand for class-terms. But
Aristotle might have considered them as names of kinds: thus P. Thom, for example, assumes that
Aristotelian kinds (εἴδη) are a special sort of sets (namely such that all their members fall necessarily or
essentially under the kind term); terms such as ‘white’ or ‘walking’ he calls quasi-Kind terms. See
P. Thom, The Logic of Essentialism, Dordrecht 1996, 316–17.
 9
Emotions in Context
Aristotle’s Treatment of the Passions in
the Rhetoric and His Moral Psychology

In the Rhetoric, Aristotle considers emotions in two different contexts: as

motives for wrongdoing in book I, and as influences on judgement in book II. In
the long chapter on pleasure in book I, he adopts a Platonic framework, treating
pleasure as the fulfilment of a painful lack. In book II, he describes emotions as
caused by impressions (φαντασίαι) of either good or bad things, suggesting a
theory that would classify them as either pleasures or pains. The Stoics developed
a theory of emotion along these lines, with the difference of insisting that
impressions that lead to emotions are always mistaken.

Aristotle’s Rhetoric contains the most detailed and extensive treatment in the
Aristotelian corpus of what the Greeks called the affections of the soul—the pathē,
‘passions’ or ‘emotions’.¹ This is a strange and remarkable fact, given the import-
ance of emotional dispositions in Aristotle’s theory of the moral virtues, and
indeed in his moral psychology in general. The nonrational part of the soul
whose virtues are the virtues of character (NE 1.13,1103a3–10) can be regarded
as primarily the seat of the emotions. Hence, one would expect to find a full
account of the emotions either in the treatises on ethics or in De Anima; and
indeed the first chapter of De Anima (403a29ff.) seems to promise a proper
scientific account of the subject. But it appears that Aristotle never got around
to this task. In the absence of a systematic treatment of this part of his psychology,
we may therefore turn to the Rhetoric to supplement what Aristotle says in other
places.
However, evidence from this work must be treated with caution. Quite a few
chapters in the Rhetoric are meant to provide the aspiring orator with materials for

¹ I will use these two terms interchangeably in what follows. The point of using both is, I think, not
just that they are both common translations of the same Greek word, but also that they seem to
emphasize two different aspects of the subject. While ‘emotion’ would seem preferable in an
Aristotelian context, ‘passion’ is more appropriate in the Stoic context. But it is important to realize
that Aristotle and the Hellenistic philosophers were talking about the same thing.

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0009
    113

speeches on topics of general relevance—for example, the virtues for speeches of

praise; human happiness or goods in general for giving advice; and so on. What
Aristotle offers in these passages ought to be regarded as a collection of
commonly acceptable theses that are likely to convince a not very educated
audience, rather than as Aristotle’s own considered views about the subject.
Now, the chapters about the emotions in book II of the Rhetoric are supposed
to tell the orator how to produce (or to avoid, as the case may be) emotions
or emotional attitudes in his audience. This means that they should in
principle be telling him the facts about psychology, not just what ordinary
people might be inclined to believe. Yet, what we find in these chapters does
not look like a treatise on psychology. Aristotle usually begins each chapter
with a definition and then tries to derive from it the answers to the questions
‘from what, regarding whom, and in what kind of state of mind’ an emotion
will arise. He offers no general definition of what a pathos is, and he usually
does not try to argue for the correctness of his definitions. Most commenta-
tors have therefore concluded that the exposition in the Rhetoric is an exercise
in dialectic, based on ‘reputable opinions’ (endoxa) rather than scientific
psychology.² But that would be odd, since these chapters are supposed to
tell the orator how to arouse or calm the passions, not how to talk about them.
What matters is not what people think about emotion but what is in fact the
case. John Cooper,³ who notices the difficulty, suggests an ingenious solution:
perhaps Aristotle thought that since rhetoric itself is not a science, and an
orator may well be ignorant of the principles of the scientific disciplines
relevant either to his craft or to the topics he discusses, an exposition of
psychological matters written for orators should not be based on the findings
of scientific disciplines, but proceed from plausible or respectable assump-
tions, as do the arguments of the orator himself. After all, such common or
respectable views also form the starting point of a serious philosophical
investigation, and they may be presumed to contain at least some truth,
even if not clearly expressed.
I do not find this consideration entirely persuasive. It seems to me that there is a
simpler, if somewhat trivial, solution to the puzzle. We should accept, I think, the
point that in order to arouse or to prevent emotions, the orator should in principle
know the truth about psychology, so far as it may be available, and not just some
plausible or popular views. But this does not imply that he needs to be an expert
psychologist himself. Much like present-day lawyers or judges, he may be content
with accepting the results of scientific research without worrying about their

² See, however, W.W. Fortenbaugh, ‘Aristotle’s Rhetoric on Emotions’, Archiv für Geschichte der
Philosophie 52, 1970, 40–70.
³ In his ‘Rhetoric, Dialectic, and the Passions’, Oxford Studies in Ancient Philosophy 11, 1993,
178–84.
114    

theoretical foundations or justifications—for example, judges and lawyers may

surely accept the expert opinions of engineers or architects regarding the con-
struction of a bridge without having to study engineering themselves, so
Aristotle’s orator should rely upon the results of scientific psychology, just as he
is also supposed to rely upon the results of Aristotle’s logical investigations in
Prior Analytics and Posterior Analytics. What we should expect to find in the
Rhetoric is not science or philosophical theory, but theory-based results. This is
what we do find where Aristotle refers to the Topics or the Analytics: he will
mention a theorem or result and then refer the reader to the technical treatises for
explanation or proof.⁴ There are no comparable references to a technical treat-
ment of the emotions—for the good reason, I take it, that Aristotle did not have a
proper scientific account, either his own or somebody else’s. Aristotle seems to
think (1.10,1369b31–32) that what he needs for his purposes in the Rhetoric is
definitions, and those may again have to be adapted so as to be ‘neither unclear
nor too precise’.⁵ The Rhetoric would obviously not be the place to develop a
theory of the emotions, so Aristotle turns to the best available substitute: theories
developed in the Academy, either by Plato or by Aristotle’s fellow students, no
doubt with some contributions from Aristotle himself. There is quite a bit of
evidence in the Topics to show that various definitions of particular emotions were
proposed and discussed in the Academy,⁶ and Plato’s Philebus contains something
like a sketch of a more general theory in the section on ‘mixed’ or ‘impure’
pleasures (47C–50E), but it does not seem that the definitions Aristotle uses
belong to an already established theory.⁷ Now, since philosophical views can
also figure as premises in dialectical debates, Aristotle’s definitions will not have
more than the status of respectable assumptions, but I think we can be fairly
confident that he chose those definitions or theses that he himself would have been

⁴ See, for example, 1.2,1356b9, 1.2,1356b12, 1.2,1357a30, 1.2,1357b22–25; 2.25,1403a4–5,

2.25,1403a12.
⁵ This passage should be compared with NE 1.13,1102a23–26, where Aristotle says that the student
of politics will have to study psychology as well, but only to the extent that it is needed for the purpose
at hand: it might be tedious to go into greater detail. An example of a definition that would be ‘too
precise’ might be one that mentions the physiological side of an emotion as well as its formal account,
such as the definition of anger discussed at DA 1.1,403a29ff.
⁶ For discussions about the definition of particular emotions in the Topics see, for example, on anger:
2.113a35–113b3, 4.125b28ff., 4.126a6–12 (also mentions fear and shame), 127b27–32, 6.151a14–19,
8.156a32–33; on calmness or mildness (praotēs): 4.125b20–27; on indignation and envy:
2.109b35–110a4 (this should be compared with the lengthy disquisition about their opposition to
pity at the beginning of Rhetoric 2.9).
⁷ Here I would disagree with Fortenbaugh (Aristotle on Emotion, New York 1975), who argues that
Aristotle is working with a ‘bipartite’ (human) psychology that replaces the Platonic ‘tripartite’ theory
in order to accommodate the emotions. Difficulties with the Platonic schema do surface in an
interesting passage in the Topics (4.5,126a6–12), but I doubt that this is sufficient to speak of a new
psychological theory. Plato’s distinctions are clearly based on types of desire, while the distinction
between the ‘reasoning part’ and the part that can listen to or obey reason is based on the beliefs or
judgements that can be attributed to the relevant ‘parts’. One does not have to see these ‘divisions’ as
necessarily incompatible.
    115

inclined to accept. As a technical handbook for orators, the Rhetoric is neither an

exercise in dialectical argument nor a piece of philosophical theorizing.
Given this kind of background, it is not surprising to find that Aristotle’s
treatment of the emotions in the Rhetoric does not proceed from a unified
theoretical framework. As a matter of fact, I think that his brief discussion of
emotions as motives, along with the long chapter on pleasure and ‘passionate
desire’⁸ (epithumia) in 1.10–11, is close to Plato’s theories in the Republic and the
Philebus, whereas the long section in 2.2–11, which deals with emotions as
influencing judgement, seems to go in a different direction—towards a theory
that invites comparison with the Stoic one. I do not have the space here to
substantiate these claims in detail, but I will try to indicate what I take to be the
main differences between the two treatments of the emotions by a few examples.
In the second part of the chapter, I will turn to a brief comparison with the Stoics,
which will inevitably take us out of the primarily psychological and morally
neutral context of the Rhetoric and into moral psychology. I think that this
comparison will also shed some light on the dispute between Stoics and
Aristotelians about the relation between virtue and the emotions.

Passions as Motives and as Influencing Judgement:

Books I and II of the Rhetoric

The pathē come up in two different contexts in the Rhetoric: first, in the section on
forensic speeches, they are mentioned as motives for wrongdoing, and one
particularly important motive, epithumia or ‘passionate desire’, is treated in
1.11—a long chapter about pleasure and what is pleasant. For anger, the other
motive explicitly mentioned—and presumably for any other passions, such as
jealousy, that might provide a motive for wrongdoing—the reader is referred to
book II. This second, and main, treatment of the emotions considers them, in
accordance with Aristotle’s plan outlined in 1.2,1356a1–20, insofar as they have an
influence on the attitudes and judgements of the listeners (2.1,13783a20ff.). As is
his custom in the Rhetoric, Aristotle avoids repetition by dealing with a subject
only once, even if it should be considered from different perspectives.⁹ So he does

⁸ I avoid the usual translation of epithumia as ‘appetite’, in this case, because Aristotle explicitly
distinguishes belief-based desires from the purely irrational cravings like hunger and thirst, which have
their origin in the body; see 1.11,1370a18–27.
⁹ The most striking example of this kind of economy is character as a ‘means of persuasion’: instead
of explaining how speakers may convince an audience of their moral wisdom and reliable character,
Aristotle simply refers the reader back to his collection of common views about the virtues for speeches
of praise or blame in 1.9, saying that ‘one would use the same things to establish one’s own good
character as one would for another person’ (2.1.1378a16–19). He can hardly mean that the best way to
present oneself as a morally good person consists in extolling one’s own virtues, and so the reader is left
with the task of adapting the materials given in 1.9 to a different purpose for which some further advice
116    

not pick up epithumia again in book II,¹⁰ and he does not emphasize the role of the
passions as motives for action in the longer section, leaving it to the reader to
figure out how passionate desire may influence judgement (2.1,1378a4) or how
envy or fear, for example, would lead to action. Apart from its prominence as a
motive for wrongdoing, there may be yet another reason for the separate treat-
ment of epithumia in book I, namely, that book II focuses exclusively on emotions
as relating to other people. Passionate desire or appetite may often be directed at
other things that one would like to have rather than at persons, and so it might not
fit very well into the schema that Aristotle uses in dealing with the other emotions.
Be that as it may, Aristotle’s division of labour between books I and II has the
interesting effect of separating, as it were, the desiderative from the cognitive
aspects of emotion. I will come back to this point later; let me first take a look at
the two ‘theoretical frameworks’ that seem to emerge from the different sections.
In 1.10,1368b32–1369a7, Aristotle introduces anger (orgē) and epithumia as
two kinds of irrational desire (alogos orexis) on a list of seven possible ‘causes
(aitiai) of action’. One might, at first, find it odd that Aristotle mentions these two
but not, say, the passions in general on a list that contains, as its other items,
chance or accident (tuchē), nature, force or compulsion (bia), habit, and calcula-
tion or reasoning (logismos). But then one notices that orgē (anger) is used as a
synonym of thumos (spirit), and thumos and epithumia (spirit and appetite) are,
of course, the names of two of Plato’s ‘parts of the soul’ in the Republic. Since
thumos as a synonym of orgē is normally used by Aristotle in a narrower sense
than in Plato, where it is associated with a desire for honour, not primarily
revenge, the two kinds of irrational desire still do not seem to be on the same
level of generality—anger is a particular emotion, while ‘passionate desire’ as 1.11
shows, seems to cover a number of different emotional states. However, at
1.10,1369a18, the expression di’alogon orexin (1369a2–3) is picked up by dia
pathos, and so one might conclude that all the passions are supposed to be covered
by ‘spirit and appetite’, and that they are all to be seen as irrational desires. The
Platonic psychology definitely seems to be in the background.
The same is true, I think, of the chapter on pleasure and the pleasant, which
opens with a definition of pleasure as ‘a movement by which the soul as a whole is
consciously brought into its normal state of being’ (tr. R. Roberts)—a definition
patently borrowed from Plato’s Philebus (31Dff.). Many commentators¹¹ have
taken this as a sure indication that what Aristotle presents here cannot be his own

might well have been helpful. The chapters 10–11 in book I probably also serve a double purpose: both
to instruct the orator about likely causes of wrongdoing and to tell him what motives he may plausibly
ascribe to wrongdoers.
¹⁰ Stephen Leighton (‘Aristotle and the Emotions’, Phronesis 27, 1982, 162–5) argues that Aristotle
probably intended to exclude epithumia from the realm of the passions because it is entirely irrational.
But he overlooks the crucial passage in 1.11 in which Aristotle distinguishes between emotional
epithumiai and bodily cravings, as well as the mention of epithumia as influencing judgement in 2.1.
¹¹ See J.C.B. Gosling and C.C.W. Taylor, The Greeks on Pleasure, Oxford 1982, 194–9.
    117

view: after all, Plato’s theory is explicitly rejected in Aristotle’s official discussions
of pleasure (NE 7 and 10). But this, I think, is a mistake. Pleasure is treated in the
Rhetoric as the correlative of epithumia. ‘Whatever is passionately desired is
pleasant, for epithumia is the desire for what is pleasant,’ says Aristotle
(1.11,1370a16–18). By contrast, Aristotle’s official theory in the Ethics makes
real pleasure independent of epithumia and lupē (NE 7.12,1152b36).
Nonetheless, in spite of Aristotle’s official doctrine, pleasure as desire fulfilment
is not absent from the Ethics either. The pleasures involved in continence and
weakness of will, or those adequately indulged by the temperate person, can
hardly be taken as pleasures of Aristotle’s preferred sort. Temperance is typically
concerned with those ‘animal pleasures’ (NE 2.10,1118a24–25) to do with food,
drink, and sex; and we do not hear of cases of weakness of will that consist in
playing the piano for an excessive amount of time, or neglecting one’s children for
the sake of solving yet another philosophical puzzle, although I would not object
to seeing these as genuine cases of akrasia. Aristotle might have avoided some
confusion in the Ethics if he had explicitly recognized at least two kinds of
pleasure—desire-fulfilment on the one hand, and enjoyment of activity on the
other. Instead, he decided to demote desire-fulfilment to the status of ‘accidental’
pleasure (kata sumbebēkos, NE 7.12,1152b33–1153a2). In a discussion of motives
for wrongdoing, however, the pleasures that correspond to passionate desire are
clearly more relevant than the more refined activity pleasures that take pride of
place in the Ethics. So we do not have to invoke chronological considerations—
such as the view that the Rhetoric is probably an early work—to argue that
Aristotle’s adoption of Plato’s definition should be taken seriously.
The survey of pleasures seems to be indebted to the Philebus in yet another way:
namely, in treating the emotions that fall under epithumia, such as, for example,
erotic love, as involving both pleasure and pain. The parallels between Plato’s
suggestive sketch of an account of emotional states as mixtures of pleasure and
pain (Philebus 47C–50C) and Aristotle’s chapter on pleasure in the Rhetoric have
been investigated in detail by Dorothea Frede,¹² to whom I am indebted for this
point. To the extent that Aristotle is following Plato here, then, one might expect
him to develop a view that treats emotions as ‘mixed feelings’—appetite (epithu-
mia), as Aristotle casually says in NE 3.11,1119a4, goes with pain; on the other
hand, the prospect of fulfilment will be pleasant. But as we will see, this is not what
actually happens in book II. Already in 1.11, the fit between some of Aristotle’s
examples and the definition of pleasure as a conscious or perceptible return to the
natural state is somewhat tenuous. When he speaks of the pleasure that comes
from victory or from the impression that one is superior to others (phantasia
huperochēs 1370b33–34)—I suppose we might call this a kind of pride?—it is

¹² See her ‘Mixed Feelings in Aristotle’s Rhetoric’, in A.O. Rorty (ed.), Essays on Aristotle’s Rhetoric,
London 1996, 258–85.
118    

difficult to regard the desire to excel as a perceived lack or disturbance of the

natural state, unless, of course, one takes it that being superior to others is
everybody’s natural condition. And while feeling inferior might count as distress-
ing, leading to a desire to surpass or at least be even with one’s peers or
competitors, such feelings need not inevitably precede or accompany the pleasure
of feeling superior: some people clearly enjoy competition and feel quite confident
that they will succeed. Obviously, we might declare passionate desire to be always
painful by definition, but these examples would seem to throw doubt upon such
an assumption.
Whether or not Aristotle endorsed an account of the passions as mixed feelings
at some point of his career, we cannot tell. However, he tends to use the phrase
‘accompanied by pain and (or) pleasure’ (hois hepetai lupē kai hēdonē,
2.1.1378a21; hois hepetai hēdonē ē lupē, NE 3.1105b23; cp. EE 2.2.1220b13) to
characterize the class of pathē in general, after giving a few examples, and this may
be an indication that he started out from something like the Platonic account.
But the theoretical framework that seems to emerge from the longer treatment
in book II is different. For now, only three of the emotions are defined as forms of
desire (anger, love or friendly feeling, and hate¹³), while no less than six out of ten
are explicitly defined as forms of distress or pain (lupē). While some of the
‘painful’ emotions may be associated with certain pleasures (e.g. fear with the
hope of salvation), it seems clear that these pleasures are not seen as a necessary
ingredient in ‘mixed feelings’. More important, although most emotions classified
as forms of distress will evidently lead to desires—such as the desire to escape an
imminent danger, in the case of fear—those desires do not form a part of their
definition. What serves as the defining feature—the differentia specifica, as it
were—is in almost all cases an impression or appearance (phantasia)—that a
terrible evil is near, that someone has suffered an undeserved misfortune, that one
has been treated with disrespect, and so on—which causes the pain or disturbance.
It is evident that Aristotle is deliberately using the term ‘impression’ rather than,
say, ‘belief ’ (doxa) in his definitions in order to make the point that these
impressions are not to be confused with rational judgements. Emotions are caused
by the way things appear to one unreflectively, and one may experience an

¹³ The treatment of the last two is exceptional in several ways: philia (and presumably also, by
implication, its opposite) is defined in terms of ‘wishing’ (boulesthai), a verb that is elsewhere reserved
for rational desire; and Aristotle seems to contrast hate with anger as being ‘without distress’,
unemotional rather than emotional, as it seems. It may be that the awkwardness arises simply from
the fact that Aristotle is using definitions formulated for a different purpose. One way in which one
might wish to set these two off from the rest would be to say that they seem to be emotional dispositions
rather than occurrent emotions—hexeis rather than kinēseis, in Aristotle’s terms. Still, whether or not a
person is seen as a friend or an enemy will no doubt have an influence on the audience’s emotional
attitude, and so the treatment of these dispositions would seem to be appropriate in its place. But it is
perhaps not surprising that friendship or goodwill and hate do not quite fit into the pattern used for the
other emotions.
    119

emotion even if one realizes that the impression that triggered it is, in fact,
mistaken.
An emotion is classified as a form of distress if the impression is negative—that
something bad has happened or is about to happen—and one would expect that a
positive impression, that something good has happened or is about to happen,
should lead to the classification of the corresponding emotion as a form of delight
or pleasure. This is not, in fact, the case for any of the explicitly discussed emotions
in book II, although one might have thought that the ‘opposites’ of anger, pity, and
fear mentioned at the beginning (2.1,1378a20–23) would be the pleasant feelings
contrasting with these forms of pain or distress. But most of the alleged ‘opposites’
of the emotions treated in these chapters are not the emotional states that might
correspond to the impression that some good thing has happened or will happen,
but rather the state of mind corresponding to the absence of the specific distress.
So calmness is opposed to anger, confidence (rather than relief) to fear, shameless-
ness (rather than, say, pride) to shame, while pity, rather than being opposed to
something like taking pleasure in other people’s good fortune, is opposed to two
different forms of distress: righteous indignation (nemesis) and envy or jealousy
(phthonos). The system of ‘opposites’, as these examples already show, does not
work very well. No doubt Aristotle introduces it because an orator may be out to
prevent or restrain an emotion just as much as to provoke one, but it does not seem
to be true that preventing or restraining a given emotion will automatically lead one
to produce an opposite emotion. Another reason for the absence of pleasant
emotions in book II may lie in the fact that some of them show up, although not
explicitly described as emotions, in the chapter on pleasures, 1.11. I have already
mentioned the example of pride; others might be the feelings of lightheartedness
and freedom from care (1370a14–15), or the pleasures of anticipation as well as
relief (1370a32–1370b10).¹⁴ The best candidate for a pleasant emotion in book II
might seem to be gratitude (charin echein, 1385a18), but here Aristotle works with a
definition of charis (the favour for which gratitude would be felt) instead of
describing or defining the actual emotion.¹⁵

¹⁴ It is worth noting that chapter 1.11 already shows the characteristic use of phainesthai or
phantasia to indicate the impression that causes an emotion; see 1370b8, b13, b33; 1371a9, a19,
a23–24. A definition of epithumia parallel to that of orgē can perhaps be inferred from
1.10,1369b15–16 together with 1.11,1370a17–18: orexis meta lupēs phainomenou hēdeos.
¹⁵ Translators and commentators alike take chapter 2.7 to be dealing with ‘kindliness’ or benevo-
lence, the feelings of people who do favours to others, rather than the gratitude of those who see
themselves as receiving a favour. But the Greek phrase charin echein normally means ‘to be grateful’,
and once we realize that this is the topic of the chapter, it is also clear that Aristotle is following his usual
schema, dealing with people ‘towards whom’, ‘on account of what’, and ‘in what state of mind’ gratitude
is felt. (See Cooper, ‘Rhetoric, Dialectic, and the Passions’, 1993, 184–5, for the difficulties involved in
reading the chapter in the usual way.) Aristotle presumably resorts to a definition of favour (charis)
because no formal definition of gratitude (charin echein) was available. His advice about the ‘opposite’
is clearly meant to show how one can persuade people that they have no reason to be grateful. Hence,
I propose to translate the opening sentences of 2.7 as follows: ‘To whom people feel grateful, on account
120    

It would be an exaggeration to say that Aristotle develops a systematic theor-

etical framework that classifies emotions as forms of pleasure or pain according to
the positive or negative evaluative impressions that cause them, but he seems to be
going in that direction, and at the same time moving away from the—possibly
Platonic—view that treated the passions as irrational desires involving a mixture
of pleasure and pain. If pleasure and pain are to serve as genera, then their species
can hardly be treated as mixtures of both. This does not rule out mixed feelings, of
course; but instead of describing each emotion as itself a mixture of pleasure and
pain, one would now speak—quite plausibly—of mixtures of emotions: hope and
fear, for example.
It seems to me to be no accident that this apparent change in theoretical
outlook comes with the change of perspective from passions as motives for
wrongdoing to passions as influences on judgement. An orator who seeks to
produce a certain attitude in his audience need not, and in most cases does not,
intend to get his listeners to proceed to immediate action. His aim is to make them
see the case at hand in a certain way: to look favourably upon his client or upon
the orator himself, or to take a negative view of his opponent. Which attitude they
have will depend crucially, as Aristotle says, on the emotions they feel towards the
persons involved in the case. And so his focus on judgement leads Aristotle to
distinguish the emotion itself from the desire that may be caused by it. As I have
already indicated, I take this to be a valid and important distinction. Even though
nobody would wish to deny that emotions often lead to desires and hence
motivate actions, most emotions are not only, or not essentially, desires or motives
for action. Some emotions, such as pity and shame, can be present without leading
to any specific desire at all, as when one feels pity for the hero of a tragedy or
embarrassed about some foolish behaviour of one’s own. (I take it that the wish
that things had turned out otherwise or that we had not behaved in such a stupid
manner does not count as desire in the Aristotelian sense of orexis.) Fear will lead
to a desire to escape the imminent danger, if that is possible, but it should still not
be identified with that desire, for the desire may be there without the emotion, and
perhaps a brave person will experience fear, but—if facing the danger is the
courageous thing to do—still not even desire to run away. Finally, positive
emotions like pride and relief would seem to be, or be based on, desire-fulfilments,
not themselves desires.
I assume, then, that it makes sense to distinguish in principle between emotions
and desires (notwithstanding the fact that a few emotions are indeed defined by
Aristotle as desires, and that epithumia is, of course, a form of desire), and that the

of what, and in what state of mind, will be clear when we have defined favour. Let favour, then—that
with regard to which the person who has received it is said to be grateful—be a service rendered to
someone in need, not in return for something, but . . . ’.
    121

essential feature common to all the passions lies in the fact that they are caused by
an evaluative impression or appearance.
To show how the Rhetoric’s account of the emotions fits into the moral
psychology of Aristotle’s Ethics, I turn now to a comparison with the Stoics,
whose theory of the passions can in a way be seen as an elaboration of
Aristotle’s views, but who also arrived at diametrically opposed conclusions with
regard to the relation of the emotions to moral virtue.

Emotions in Moral Psychology: Aristotle and the Stoics

Aristotle’s discussions of the passions in the Rhetoric stay mainly on the level of
straightforward psychology and away from moral evaluations, for the obvious
reason that the orator has to be able to argue either side of a case. His purpose is to
persuade and influence judgement, not to educate or, for that matter, to corrupt
the audience. Hence, Aristotle explains what will make people feel angry or
embarrassed, grateful or jealous, rather than when or whether they ought to feel
angry or grateful. The doctrine of the mean is conspicuously absent. For the Stoics,
however, there could hardly have been such a thing as a morally neutral psych-
ology of the emotions, since they considered emotions not as a normal and natural
part of human experience but as disturbances or illnesses of the rational soul.
Nonetheless, their account of the passions has much in common with Aristotle’s,
and in order to see how the theory adumbrated in book II of the Rhetoric might
have been developed into a systematic account of the emotions, including the
positive ones and not limited to emotions relating to other people, it is instructive
to take a look at the Stoic classification of the passions.¹⁶ If we disregard for the
moment the—admittedly crucial—difference between mere impression and full-
fledged belief, we can see the Stoic schema as a plausible elaboration of Aristotle’s
sketch. What is common to all emotions is that they are based on (or, according to
some sources, actually consist in) an evaluative belief. ‘Pleasure’ and ‘pain’ are
used as generic terms for emotions, due to the belief that something good or bad is
present, while ‘passionate desire’ (epithumia) and ‘fear’ (phobos) are generic labels
for emotions based on the belief that some good or bad thing is imminent or near.
The elevation of passionate desire and fear to the generic level, where Aristotle
seemed to operate only with pleasure and pain, seems to me to be a plausible
move, and one which alerts one to some peculiarities of Aristotle’s less systematic
treatment. We have already seen that the chapter on pleasure (1.11) seemed in fact

¹⁶ For the general outline of the Stoic classification, see the testimonia in H. von Arnim, Stoicorum
Veterum Fragmenta III, 1903; reprint, Stuttgart 1964, 377–420. I have offered a more detailed account
of the Stoic position in ‘Following Nature’, Oxford Studies in Ancient Philosophy 9, 1991, 1–71,
especially 61ff. (reprinted in Striker, Essays on Hellenistic epistemology and Ethics, Cambridge
University Press 1996).
122    

to cover a number of different emotions under the headings of epithumia and

hēdonē, and it is surely plausible to suggest that longing for one’s beloved, for
example, is different from the desire to win, just as the pleasure of seeing one’s
beloved is different from pleasure taken in victory. In the case of fear, it appears
that Aristotle consciously opted for a rather narrow notion, linking fear exclu-
sively with the prospect of great physical harm. He acknowledges elsewhere (NE
3.6,1115a10ff.; cp. EE 3.1,1229a31–1229b22) that people are afraid of other things
as well, such as embarrassment, and he occasionally uses what appears to be an
already current definition of shame (aidōs) as ‘fear of disrepute’ (phobos adoxias,
NE 4.9,1128b12–13). But he may have preferred the narrower definition in order
to keep fear as the emotion relevant to courage. Once again, I find the Stoic version
plausible. Fear of physical destruction may be the most prominent case, but it still
seems more accurate to describe other forms of anxiety as types of fear rather than
simply as forms of distress.
The Stoic system, based as it is on types of evaluative belief, also shows up the
distance between Plato’s distinctions of types of motivation in the Republic, which
seemed to underlie Aristotle’s treatment of anger or ‘spirit’ and passionate desire
as motives, and a theory of the emotions taken by themselves. In the Stoic
classification, all forms of passionate desire (including, notably, anger) are
grouped together under epithumia. The other three kinds, by implication, should
presumably not be understood as forms of desire, even though the general Stoic
definition describes passion or emotion (pathos) as a kind of impulse (hormē).
What is indicated by this term would seem to be rather a kind of concern,
inclination or aversion as the case may be, that may, but need not, lead to
immediate action. For the Stoics, as for Aristotle in a few passages, the emotions
are primarily a kind of disturbance.
So far I have emphasized what I take to be striking similarities between the
Aristotelian and the Stoic account of the emotions. But as is well known, Stoics
and Aristotelians assigned very different places to the emotions in their respective
moral psychologies. In terms of psychological theory, the decisive difference lies in
the view the Stoics take of the impressions or beliefs that underlie the emotions.
While Aristotle seems to regard emotion as a spontaneous and natural response to
the way things appear to us,¹⁷ the Stoics maintain that passion arises only when an
impression is assented to and accepted as a truth. Because emotion is tied in this
way to reason’s ability to accept or reject an impression, the Stoics deny that young
children are even capable of emotion. The evaluative responses and the desires of
children are accounted for by a natural instinct for self-preservation and for
concern for other human beings, which regulates their lives until reason can

¹⁷ See NE 2.5,1105b19–1106a4 for the points that the capacity for emotion is natural and that
emotions are involuntary.
    123

take over. In this psychological picture, the emotions seem to be superfluous: the
action-guiding role they are said to have on the Aristotelian theory is filled first by
natural instinct, then by reason. According to the Stoics, there is no need to
postulate yet another natural capacity or source of motivation in the human soul
in order to explain human behaviour and action, including morally right action. If
the emotions were natural, one would have to assume that they somehow con-
tribute to the human good—but the Stoics insisted that far from being natural, the
passions were a deplorably common aberration of reason,¹⁸ an illness of the soul
that arises from the error of ascribing real value or disvalue to the external objects
that we are made to seek out or avoid by our instinctive inclinations and aversions.
Now it is true, of course, that on the Aristotelian theory, one and the same
action or belief may be explained by reference to two or three different human
capacities; and virtuous action in particular is explicitly said to require that reason
and emotion be in agreement about the best way to act. But this is not because
Aristotle gratuitously invokes two motives where one might be sufficient. It seems
clear that Aristotle is not thinking in terms of theoretical economy. He is offering
an account of what he takes to be observed facts, including the fact that normal
human beings experience emotions; and he follows Plato in distinguishing emo-
tions from both bodily urges and rational decisions or judgements. This seems to
be the commonsensical line to take: given that emotions appear to be normal in
the sense that every human being experiences them, Aristotle assumes that they
are a part of our natural endowment and tries to make sense of them on the basis
of that assumption.
The Stoic theory, on the other hand, seems to be motivated at least as much by
moral as by psychological considerations. The Stoics apparently thought that both
moral badness and human unhappiness have their origin in an excessive concern
for things beyond a person’s control—either passionate attachments to other
people or strong desires and aversions towards those things that are relevant to
self-preservation, and that humans as well as other animals will pursue or avoid
under the guidance of their natural instincts. Human beings should indeed
continue to follow the pattern of pursuit and avoidance set by nature, but a
truly rational human being would eventually come to see that success in obtaining
those things has no weight at all compared to the real goodness that is to be found
in perfect rationality, as present in human virtue and in the divine order of the
universe. A person who had reached this insight would attain both peace of mind
and perfect moral virtue. What prevents most people from attaining this goal,

¹⁸ The Stoics realized, of course, that it was paradoxical to maintain both that humans are made by
nature to achieve perfect virtue and that just about every human being in fact turns out to be morally
bad. It is not clear that they offered a convincing solution to the problem of the origin of evil. For some
source materials on this question, see A.A. Long and D.N. Sedley, The Hellenistic Philosophers,
Cambridge 1987, vol. 1, s. 65, 410–23.
124    

according to the Stoics, is precisely their liability to passion—that is to say, their

tendency to ascribe real goodness or badness to external things. Once reason has
assented to the impression that one of the objects of our natural instinct is truly
good, the effect, according to the Stoics, will be devastating. The erroneous belief
will inevitably have an influence on subsequent impressions and judgements, and
it will eventually lead to overwhelming and insatiable desires as well as inner
turmoil caused by the frustration of those desires. If the passions are thus to be
regarded as the source of all evils in human life, they can hardly be considered as
part of nature’s plan; and since, as we saw before, human action could in principle
be accounted for in terms of reason and instinct alone, the Stoics felt justified in
advocating the complete extirpation of the passions.
One might have some doubts about the Stoics’ sweeping indictment of passion
as the source of all the evils of human life, pointing out that cold-blooded crime
seems to be at least as widespread, and arguably more horrifying, than the more
spectacular variety of crimes of passion. It is not very plausible to suggest, for
example, that the crimes of concentration camp doctors were motivated by an
obsessive attachment to external goods of any sort. But in ancient times, defenders
of the Aristotelian position seem to have focused rather on the claim that passion
or emotion is unnatural and superfluous. Plutarch, for example, argues that the
Stoic psychology is mistaken, as evidenced by the common phenomenon of inner
conflicts between reason and emotion (De Virtute Morali 445a–447c), and that,
furthermore, the moderate, allegedly unemotional impulses recognized as legit-
imate by the Stoics are really nothing but what Aristotelians would call a moderate
degree of emotion (De Virtute Morali 449a–449b). The Stoics had convincing
replies to both of these objections: they pointed out that ‘inner conflicts’ can be
construed as reason’s wavering between different options rather than as a battle
between opposed forces inside the soul, and that emotion is not needed to account
for ordinary, everyday action. The debate between the advocates of Peripatetic
metriopatheia (a moderate degree of emotion), on the one side, and of Stoic
apatheia (freedom from emotion), on the other, as described by Plutarch, looks
like something of a stalemate.¹⁹ While the one side insists that emotion is natural
and has to be accommodated even though it may occasionally lead to excess, the
other side persists in its view that the passions, once admitted, will inexorably lead
to disaster and should, therefore, be avoided. Both sides agree that virtue and right
action must be explained in terms of reason’s correct judgements and decisions.
So long as the capacity to experience emotions is seen as just one more source of
motivation that may either support or hinder reason, it seems unclear, at the least,
why it should be needed for virtue, regardless of whether or not it can be avoided.

¹⁹ For a detailed discussion of this debate as presented in other ancient authors, see Martha Craven
Nussbaum, The Therapy of Desire, Princeton 1994, chapters 10–12.
    125

But once we realize that Aristotle, if not all of his later followers, distinguishes
between the emotion as an affection of the soul caused by an evaluative impression
and the desire, if any, that may arise from it—the cognitive and the motivational
aspect of emotion—we can perhaps see more clearly what underlies his notorious
thesis (NE 6.12,1144a29–1144b1) that moral virtue is required for the intellectual
virtue of practical wisdom or intelligence (phronēsis). As the Rhetoric makes clear,
emotions will have an influence on the way we see and judge other people and
their actions as well as our own future prospects. Now, if emotional dispositions
are what underlies virtue of character, the influence of emotion on judgement
cannot be regarded as merely distorting, a distraction, as it were, from rational
thought. An orator’s attempt to arouse or dispel emotions should also not be seen
as mere manipulation, or as an attempt to produce conviction by illegitimate
means.²⁰ If morally good people can be expected to have certain characteristic
emotional responses, then the influence of emotion may sometimes be what is
needed to see things in the right way. For example, it may be perfectly appropriate
for a speaker to remind the people in the audience of services rendered to the
community by his client, so as to make them feel grateful; and it may be equally
legitimate to arouse pity for the victim of an undeserved misfortune. If the
audience were impervious to such feelings, it might well arrive at an unfair or
overly harsh verdict. Since emotion will have an influence on how we see and
judge people and their actions, the right kind of emotional disposition may be
what enables us to see things in the right moral perspective. If we had not been
brought up to detest cruelty, for example, we might not even notice it as a bad
thing and so would act cruelly or let cruelty pass without interfering. Again, if we
had not learned to feel distress at the suffering or unfair treatment of others, we
might overlook it or fail to be moved by compassion or indignation to try to
prevent or alleviate suffering or to redress an injustice. What is at issue here is not
the ability to figure out that cruelty or undeserved suffering are bad or undesirable
in general. The Stoics would, of course, have insisted that their ‘wise man’ was well
aware of such things and would respond appropriately. The role of emotion as
described by Aristotle seems rather to be that of directing one’s attention to the
practically or morally relevant features of a situation: ‘Fear makes people good at
deliberation,’ as he puts it in Rhetoric 2.5,1383a6–7.
The person of practical wisdom, as Aristotle says in a famous passage (NE
6.12,1144a31–36), must start reasoning from the premise that such-and-such,
whatever it may be, is ‘the end and the best’. But unless she is a (morally) good
person, this will not appear to her (touto d’ ei mē tōi agathōi, ou phainetai).

²⁰ On the ‘appeal to emotion’ as a legitimate part of the orator’s performance, see Alan Brinton,
‘Pathos and the “Appeal to Emotion”: An Aristotelian Analysis’, History of Philosophy Quarterly 5,
1988, 207–19.
126    

It seems to me significant that the verb translated as ‘appear’ here is phainesthai—

precisely the one that occurred (either as a verb or in its cognate noun phantasia)
in the definitions of the emotions in book II of the Rhetoric. Aristotle’s concise
remark can be understood in two ways, both of which may well be intended. First,
a bad person might fail to notice what would be the best course of action in a given
situation. For example, she might simply not perceive her neighbour as a person
who has suffered an undeserved misfortune and who is, therefore, in need of help.
Second, even if it were pointed out to her that help was required, she might not see
that as the best course to take—it may not appear attractive to her, even if she ends
up doing it. The first interpretation brings out the cognitive element; the second,
the motivational role of emotional dispositions for moral character. It is the right
kind of emotional disposition that enables the morally virtuous person to see or
recognize what is the best in any situation. A bad person, by contrast, might be
described as morally blind: not only will she fail to notice the relevant aspects of a
situation but even if someone told her ‘what is best’, she might fail to recognize it
as the best. If this is correct, then the Aristotelians have a stronger argument to use
against those who would rid us of our passions than just that emotions seem to be
a natural phenomenon. They could claim that our capacity for emotional
responses is what alerts us to morally important features of situations that we
need to take into account in attempting to arrive at a rational judgement.
In the context of deliberation, an emotional response can be described as one
that is in proportion to the good or evil perceived, and hence adequate, if the
initial impression holds up under reflective scrutiny. Aristotle no doubt expects a
virtuous person’s emotions to be in line with ‘what the person of practical wisdom
would determine’ (NE 2.6,1107a1) because her spontaneous reactions will not be
independent of her education, her previous experience, and the advice of others.²¹
The fact that emotions are not the result of rational decision does not mean that
they are mere reflexes—after all, Aristotle quite rightly distinguishes them from
completely irrational appetites like hunger and thirst. Emotions are not based on
reasoning, but, as Aristotle puts it, they (that is, the ‘part of soul’ to which they
belong) can be persuaded by reason (see NE 1.13,1102b25–1103a3). This means
not only that we can sometimes be talked out of an emotion by arguments to show
that our first impression was mistaken; it also means that our reason-based beliefs
and convictions will make us disposed to be impressed in certain ways and to have
the corresponding emotional responses. If Aristotle is right on this point, then the
Stoic ‘argument from excess’ loses much of its force. Aristotle would be the last to
deny that passions can lead to excess—excess is, after all, what is essential to one
sort of vice, like cowardice or greed. What Aristotle does deny is that emotion

²¹ For a fuller discussion of Aristotle’s views on the education of the emotions, see Nancy Sherman,
‘The Role of Emotions in Aristotelian Virtue’, Proceedings of the Boston Area Colloquium in Ancient
Philosophy 9, 1993, 1–33.
    127

must lead to excess. Given Aristotle’s view of the interaction between reasoning
(both one’s own and that of others) and emotion, I do not see why this should be
regarded as unfounded optimism. It is true that Aristotle does not offer a recipe
for avoiding the dangers of excess once and for all. But, as I have tried to show, he
offers us some good reasons to think that in getting rid of the passions, we might
lose more than we can hope to gain.²²

²² In writing this chapter, I have incurred debts to many people: first and foremost, to the graduate
students in my spring 1994 seminar on book II of the Rhetoric; to Dorothea Frede, for a lively and
illuminating discussion of her own paper in that seminar; to the Discussion Club at Cornell, for a
stimulating debate about a half-baked version of this chapter; and to Mitzi Lee, Nancy Sherman, and
John Cooper for helpful comments and criticism of the penultimate version.
 10
Aristotle’s Ethics as Political Science

Aristotle sees his Ethics as a part of political science—the part that sets out
an account of the best human life, and so the goal at which a good ruler would
aim for all of his citizens. This chapter argues that the practical reasoning of
virtuous agents will be guided by the rules of what Aristotle calls (in EN V 1)
justice in the sense of obedience to law, and also describes as complete virtue in
relation to others. Those rules should be found in the laws of a well-ordered
community, and hence readers of the Ethics will be expected to read the Politics
as well.

Anyone who has read the first few chapters of the Nicomachean Ethics will know
that Aristotle considers his treatise on ethics as a contribution to the science of
politics. The claim is stated prominently at the beginning (e.g. 1.2,1094b10–11;
cp. 1.13,1102a7–13); we are reminded of it in the discussion of practical wisdom in
book VI (e.g. 6.8,1141b23ff.), and the last chapter of the EN announces the
transition to the inquiry into politics proper—forms of government and the ideal
state. But few scholars who have written on the Ethics¹ pay more than passing
attention to these statements. This is understandable: not only has the treatise on
politics come down to us as a separate work; politics also plays hardly any role in the
treatises on ethics and, after all, the two disciplines have been treated as distinct
since ancient times. Furthermore, ethics has stubbornly remained a part of phil-
osophy, while political science now officially purports to be a social science—one
which seems, however, to be largely a combination of history and philosophical
theory (much like Aristotle’s Politics, by the way). Besides, Aristotle’s Politics may
come as something of a shock for readers who begin with the Ethics. It is much more
obviously a product of its time in history and advocates views that are now
thoroughly unacceptable: it defends slavery, assumes the natural inferiority of
women to men, and argues for a rather elitist form of government. Finally, it is

¹ I use Ethics to refer to Aristotle’s ethical treatises in general, that is, both the Eudemian Ethics (EE)
and the Nicomachean Ethics (EN). Unless otherwise noted, translations are from the Revised Oxford
Translation of Aristotle’s Complete Works (J. Barnes (ed.), Princeton 1984), with occasional minor
modifications.

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0010
  ’      129

certainly a disappointment to discover that Aristotle thought his plausible and

attractive model of the best human life could be realized only in politics or in pure
research.² We can see why Aristotle thinks that politicians should study his theory of
the human good, since the aim of a ruler should be to provide the best life for the
lucky few who are capable of living it and (therefore) have the right to participate in
government. This also explains why Aristotle insists on the necessity of legal rules for
education (as already announced in EN 10.9,1179b34). But beyond this, a modern
reader might be forgiven if she decides that Aristotle’s conclusions in the Politics do
not follow from the premises laid out in the Ethics. The Politics contains a large
number of false empirical assumptions that a reader of the Ethics may safely ignore.
I sympathize with this attitude, even though, on closer inspection, the Politics
turns out to be less outrageous than one might think. But while Aristotle’s political
theory does indeed play little or no role in the Ethics, one should perhaps keep
in mind that Aristotle’s plan for two interconnected treatises may have led
to some gaps in the first one that may seem hard to understand if one limits
oneself to the Ethics. I think, in fact, that some recent scholarly controversies may
be due to the tendency to consider the Ethics as a work that is complete by itself. It
seems to me that some of the difficulties that have led to these controversies can be
resolved—whether or not one is satisfied with the solution—if one considers the
Ethics as only a part of the larger enterprise.

Ethics as an Aristotelian Science

But before I return to these questions, let me first take a look at Aristotle’s claim
that ethics as a branch of politics is a science in his sense of the word.
One might have some doubts about this on the ground that Aristotle contrasts
both ethics and politics with exact sciences like geometry in a famous passage (EN
1.3,1094b11–27; see also 2.2,1103b34–1104a11) in which he reminds the reader
that the subject matter of politics—what is noble and just—presents much vari-
ation and diversity; so much so that some people have maintained that it is only a
matter of convention or custom. The same is true of goods—the subject matter of
ethics. Therefore, Aristotle says, one should be content with an investigation that
indicates the truth only roughly and in outline: the subject is such that general-
izations will hold only for the most part, and so will the corresponding conclu-
sions. Hence an educated person will not expect strict demonstrations in ethics.
Indeed, to ask a public speaker for formal proofs would be just as silly as to accept
merely persuasive arguments from a mathematician.

² It might be worth noting that Cicero still seems to think in the same way (De Off. I 69–73).
Choosing a profession was presumably not a part of the leisured classes’ lifestyle, since they did not
usually have to earn a living.
130    

It is unclear whether this verdict should be taken to cover absolutely everything

that Aristotle has to say later. I find it hard to believe, for instance, that he takes his
definition of eudaimonia as a rough and ready generalization. On the other hand,
the notorious ‘doctrine of the mean’ is a thesis that offers guidance, if any, only in
a vague and general way. I think, in fact, that Aristotle is running together two
distinct points in these passages: first, he wants to say that certain generalizations
admit of exceptions (such as that wealth is a good thing), second, he also wants to
point out that the general theses he argues for are so general that they need to be
complemented with a lot of detail to provide any actual guidance for action. This
is obviously true of the thesis about the mean, but it could be said also of the
definition of the human good. But even if we insisted that all of Aristotle’s theory
is to be understood in this way, it would not follow that the investigation he is
about to begin could not be what he calls a science: a theory that explains a set of
more or less agreed facts by finding general principles that will help one under-
stand why the facts are as they are. And a few paragraphs further down
(1.4,1095a31–b8) Aristotle describes his procedure as following the familiar pat-
tern of scientific investigation by starting out from the things ‘better known to us’
and leading up to principles that are ‘better known in themselves’.
One difference between a purely theoretical science like mathematics and a
science concerned with action lies in the fact that the second is—in our, not
Aristotle’s terminology—an applied science that deals with individual cases. Here
the analogy with medicine which Aristotle frequently invokes, notably in EN VI,
can be helpful. Doctors treat individual patients whose condition varies so much
that any textbook knowledge will have to be supplemented with experience. But
this does not mean, of course, that the general theory of health and illness or of the
human body that informs the doctors’ diagnoses and therapies is not a science.
There remains a gap between the scientific theory and the treatment of actual
cases that only extensive experience can bridge—and the same holds for ethics and
practical wisdom.
But the most important difference between ethics and the theoretical sciences
lies in a point that Aristotle mentions just after the remarks about its lack of
precision—namely that the aim of the investigation is not just knowledge, but
action (1095a5–6). Here the contrast between ethics and medicine is more sig-
nificant than their similarity. Both ethics and medicine have an aim that goes
beyond finding the truth. But while the good to be achieved by medicine—
health—is uncontroversial and largely accessible to observation, as are the indi-
vidual facts and low-level generalizations collected by the doctors, the facts to be
considered in ethics can only be securely grasped by persons who have received
the proper education (1.4,1095b4–7). Aristotle devotes the first book of each of his
treatises on ethics to the establishment of a definition of the human good. But he
realizes that such a definition will not be sufficient to lead an agent to adopt the
proposed conception of the good, and to act accordingly. For, unlike an
  ’      131

anthropologist who wants to learn about the moral code of a tribe she does not
know well—for example, mafiosi or golf players—the student who is to benefit
from lectures in political science or ethics must have learned to value the things
she recognizes as good or noble and see them as desirable ends of action, so that
she may use the explanations offered by the science not only as explanations, but
also as justifications of her own way of acting. The anthropologist need not adopt
the moral code she is studying, while the student of ethics is trying to arrive at a
better understanding of her own moral judgements.
The education Aristotle has in mind here has to do, as he explains in book II
(3,1104b8–27; see also 10.9,1179b23–31), with pleasures and pains. That is to say,
children are trained by praise and blame, rewards and punishments to like and
dislike the right things, as Plato memorably put it. This will lead them to see the
right things as good and bad, and to aim at things that are good because they
are good.
The requirement for a good education marks the difference between a science
of action (πρᾶξις in Aristotle’s strict sense of that word) and a productive science,
or for that matter a theoretical science as well. Both doctors and botanists have to
be keen observers, but they can collect the information they need regardless of
their moral character, and there is no principled reason why a child might not
begin to study botanics by the age of eight. Quite the contrary—he may become a
better scientist the earlier he starts. In ethics, however, the facts do not become
clear through mere observation of people’s behaviour. Aristotle is surely right in
thinking that systematic reflection on the values one has been taught to appreciate
will not make one a better agent unless one also endorses those values. Ethics,
then, is not what we would call an empirical science. While both ethics and
empirical science start from a collection of facts that one tries to understand,
the peculiar nature of the facts one needs to be familiar with in order to study
ethics also explains why one’s initial collection of putative facts may turn out to be
an incoherent set. Reflection on moral questions may sometimes lead one to revise
one’s initial judgement, while the observations of doctors and biologists are not
usually expected to need revision. I emphasize this because Aristotle has some-
times been accused, on account of his method, of being hopelessly conservative.
Yet he himself points out early on that judgements about what is good or noble
may be controversial—not, I presume, because he shares the view of those who
maintain that there are no facts to be found in this field, but because it may take
some reflection to sort out the controversies.
To continue with the outline of the proposed investigation: the plan of the
Nicomachean Ethics can be compared, for example, with that of the De Anima.
Aristotle begins by establishing a definition of the subject—the human good—and
shows that it is supported by the common beliefs (or intuitions, as they are now
called) that we already have about it. As with the (second) definition of the soul in
the DA (2.413b10–13; 414b19–415a13), he then goes on to analyse the elements
132    

that go into the definition: in the case of the soul, the various capacities that define
it; in the case of the human good, virtue and action. Since virtue of character is
defined as a disposition to make the right choices in accordance with reason, we
also get a discussion of choice and deliberation. One would then expect the book
on the intellectual virtues, EN VI, to shed further light on the notion of ‘right
reason’—which is indeed what Aristotle promises at its beginning (1,1138b32–34).
After the virtues comes an investigation of the less desirable states of character—
self-control, weakness of will, and viciousness. The rest of the plan is less clearly
tied to the definition, but can be connected to the conditions Aristotle had set out
in the first book for any candidate for the human good: pleasure is not only one of
the traditional candidates for the human good, it must also be a part of the best life
because, as Plato had pointed out in the Philebus (21D6–E5), no one would be
content with a life devoid of all pleasure. Friendship is relevant because Aristotle
had also noted that the good life for an individual will have to include the well-
being of her family, friends, and fellow citizens as well (EN 1.7,1097b8–11). The
final chapters of the EN return to a question left open in book I (7,1098a16–18)—
should we say that the best life consists in activity in accordance with all the
virtues, or only with the best among them?—and apparently answers it in favour
of the second option: the best virtue, namely excellence in pure theoretical
thought, is what makes for the best life. Aristotle notes that his present project
has been completed (10.9,1179a33–35), at least as far as writing can contribute to
it, and proposes to proceed to the investigation of legislation and forms of
government.
However, in terms of the avowed purpose of making us better agents, most
readers might be forgiven for feeling somewhat disappointed. Barring the radical
conclusion that we should all try to become philosophers or mathematicians, we
seem to be left with very little indeed in terms of advice for acting as we should.
This is, of course, because book VI, in spite of its initial promise, has little or
nothing specific to say about the content of our deliberations. Aristotle talks a lot
about the differences between theoretical and practical reasoning, which is exer-
cised in deliberation. Its end is, as he puts it, ‘truth in accordance with right desire’
(2,1139a29–31), reminding us of the requirement for a good upbringing; and the
general aim of the agent is ‘acting well’ (εὐπραξία). What he does not tell us is just
how we determine what constitutes ‘acting well’ in each particular situation. So it
is understandable that many of Aristotle’s readers have argued that the apparent
gap must be filled by what we already have.

Practical Reasoning: Rules or No Rules?

With apologies for oversimplification, let me outline the solutions that have been
proposed in the last few decades. They fall roughly into two types: the first and
  ’      133

most widely accepted view is that the practically wise person’s deliberations and
decisions will be guided by her conception of the human good. There is an
ongoing controversy over whether this means that she must have studied the
Ethics to acquire the relevant clear and specific conception of the best human life,
or whether this conception is largely implicit, arising initially from her (good)
upbringing and becoming more articulate over the course of her life³—a process
that would no doubt be helped along by a study of the Ethics, but that would not
presuppose it. If we assume the more plausible second version of this interpret-
ation, the suggestion is that, according to Aristotle, a good education would lead a
young man ([sic] that is, after all, what he is talking about) to opt for a life of
virtue—as opposed to, say, honour and glory, or pleasure (see EN 1.5)—and to
have acquired an attachment to the relevant values, so that he takes pleasure in
virtuous activity and detests viciousness. Such a person would have a fairly good
idea of what counts as virtuous or evil action in his society, and this would enable
him to see, in each particular situation, ‘what the practically wise man would
choose’. So he would set himself the right aims—say, defending his country,
contributing to public enterprises, helping the poor—and then figure out the
right way of achieving them. This, then, is what Aristotle has in mind when he
claims that the good and wise person will act ‘for the sake of the noble’.
The second proposal comes from interpreters who tend to see Aristotle’s ethics
as a kind of polar opposite of modern theories of morality, and Kant’s theory in
particular.⁴ Their view is that there is no gap in Aristotle’s account of practical
wisdom. Aristotle does not think that the good person’s deliberations or decisions
rely on any kind of rule, nor a blueprint of the kind provided by his own account
of the human good. Though one might still say that the good person’s decisions
express her conception of the best life, this conception is not explicitly invoked in
any deliberation, being entirely the result of her good upbringing. After all,
Aristotle compares the wise person’s deliberative ability to perception informed
by experience (see EN 6.8,1142a23–30; 11,1143a35–b14). Decisions, just like
perceptions, concern particulars, and this is taken to mean that the virtuous
man will simply ‘see’ what needs to be done, without appeal to any rules. So,
defenders of this anti-Kantian view insist that Aristotle offers no rules or stand-
ards of rightness for decisions because he thinks there can be no such thing.
It seems to me that neither party to this dispute can be entirely right. For the
first view is open to another objection, less frequently mentioned in this context: it
is not at all clear how an Aristotelian conception of the best life—whether

³ For those different interpretations, see: J.M. Cooper, Reason and Human Good in Aristotle,
Indianapolis 1975; repr. 1986; S. Broadie, Ethics with Aristotle, New York 1991; R. Kraut, ‘In Defense
of the Grand End’, Ethics 103, 1993, 361–74.
⁴ The most vigorous exponent of this view is J. McDowell—see, for example, his ‘Deliberation and
Moral Development in Aristotle’s Ethics’, in S. Engstrom and J. Whiting, Aristotle, Kant, and the Stoics,
Oxford 1996, 19–35.
134    

articulate or implicit—could help one in figuring out what should be done.

Aristotle’s version of eudaimonism is not a theory like utilitarianism in which
we can use the conception of the good to find out what is right, simply because
Aristotle’s conception of the human good includes moral virtue. The agent who
wants to do the right thing will hardly be helped by the advice that she should do
what the virtuous person would do, since what she wants to know is precisely what
it is that the virtuous person would do. If Aristotle’s account of the human good
only tells her to do what virtue requires, this is not much better than being told ‘in
case of doubt, do the right thing’.—But before we conclude that, therefore, the
opposite party must be right, we should remember that Aristotle also keeps
referring to the wise person’s knowledge of ‘the universal’ (that is, as his examples
indicate, the universal premise of a practical syllogism). He tends to illustrate this
by his woefully trivial examples of practical syllogisms, typically taken from
medicine, like this one: All heavy liquids are bad; this is a heavy liquid . . .
(6.8,1142a20–23). Presumably the conclusion is that I should not drink this liquid.
To use an example that looks more like moral reasoning: suppose I see that
someone has been injured in a car accident. I reason like this: People who are
injured need help; in particular, they need to see a doctor; so, let me take this
person to a hospital. It might plausibly be argued that most of us would indeed
‘see’ immediately that the accident victim should be taken to a hospital. But one
might equally plausibly retort that this is because we know, quite generally, that
accident victims should be assisted, and that they need medical help. And this
kind of general rule is surely what an agent would appeal to if asked to justify or
explain her action. It is difficult to understand how, on the ‘no rules’ view, there
could be any meaningful controversy about the rightness or wrongness of specific
ways of acting. I should hesitate to ascribe such a view to an author who clearly
thinks that questions about what is good or bad, just or unjust, can not only be
discussed but also, in most cases, correctly answered.
What I want to suggest instead is that Aristotle does not spell out the universal
premises of practical deliberation because he takes it to be obvious what they must
be—namely rules of justice in the wide sense in which, as Aristotle tells us in EN V,
justice is ‘complete virtue in relation to others’. Aristotle believed that most, if not
all, of those rules can be found in the laws of a well-ordered city, and the Greeks
traditionally understood justice in this wide sense as obedience to law (τὸ νόμιμον,
1129a34, b11–25).
Here is how Aristotle introduces his distinction between general and specific
justice (EN 5.1,1129a26–1130a9):

Now ‘justice’ and ‘injustice’ seem to be ambiguous, but because the homonymy is
close, it escapes notice . . . Let us then ascertain the different ways in which a man
may be said to be unjust. Both the lawless man and the grasping and unfair man
are thought to be unjust, so that evidently both the law-abiding and the fair man
  ’      135

will be just. The just, then is the lawful and the fair, the unjust the unlawful and
the unfair . . .
Since the lawless man was seen to be unjust and the law-abiding man just,
evidently all lawful acts are in a sense just acts; for the acts laid down by the
legislative are lawful, and each of these, we say, is just. Now the laws in their
enactments on all subjects aim at the common advantage either of all or of the
best or of those who hold power, or something of the sort; so that in one sense we
call just those acts that tend to produce and preserve happiness and its compo-
nents for the political society.
And the law bids us do both the acts of the brave man (e.g. not to desert our post
or to take flight and throw away our arms) and those of the temperate man . . .
and similarly with regard to the other virtues and forms of wickedness, com-
manding some acts and forbidding others; and the rightly framed law does this
rightly, and the hastily conceived one less well.
This form of justice, then is complete virtue—not absolutely, but in relation to
others . . . And it is the most complete form of virtue, because it is the actual
exercise of complete virtue . . . Justice in this sense, then is not a part of virtue but
virtue in its entirety . . .

Aristotle’s distinction between general justice in this wide sense and the specific
virtue he discusses in EN V—what we might call fairness as opposed to selfish
greed (πλεονεξία)—has not attracted a lot of scholarly attention in recent debates,
and one has to admit that Aristotle himself does not make much use of it. Indeed,
one might argue that the lines between general and specific justice get blurred even
in book V itself, once Aristotle turns to questions of what he calls ‘political justice’
in chapter 10; and as far as I know, the Politics does not even mention the
distinction. The opening chapter of EN V may well be one of many instances of
implicit criticism of Plato: Aristotle notes that while greed is traditionally
described as the contrary of justice,⁵ it is actually not the case that every instance
of injustice is a case of greed. Plato sets out the traditional contrast between
(allegedly natural) selfishness and justice at the opening of Rep. II (358B–367A),
but what he then defends in the rest of the book is not only fairness, but moral
virtue as a whole—righteousness, as one might call it with a slightly antiquated
term. Remember, for example, that justice is defined in book IV as each part (of
society as well as of the soul) doing its proper job, which shows that justice
presupposes the other virtues. What is at issue in the Republic is indeed justice
as lawfulness or obedience to the law—the notion of justice attacked by
Thrasymachus in book I. But justice in this wide sense is a matter of politics, as

⁵ Before Plato, see the fragment On Truth by the sophist Antipho (DK 87B44); for Plato, Gorgias and
Rep. I–II.
136    

is the law. As Aristotle puts it succinctly in Pol. 1.2, ‘justice is a political thing.⁶ For
it is the order of the political community; and a judge’s verdict is a determination
of what is just’ (transl. GS). So in the Ethics Aristotle limits himself to a discussion
of the virtue of fairness, leaving the discussion of justice as determined by the law
for the treatise on politics. But what the law (if good) prescribes is precisely
morally good conduct, as the lines preceding this passage show: ‘For man, if
perfected, is the best of animals, but when separated from law and justice, the
worst of all; since armed injustice is the most dangerous, and he is equipped at
birth with the arms of intelligence and virtue, which can be used for the best as
well as the worst ends.’ I suggest that if there is a term in the Greek vocabulary that
corresponds to our ‘morality’, it is δικαιοσύνη, not ἀρετή (virtue) or even καλόν⁷
(noble or fine). It is, after all, generally acknowledged that the Greek word ἀρετή
covers more than what we would call moral virtue—I suppose this lies behind the
recent tendency to translate it as ‘excellence’. The same is true of καλόν, translated
as ‘noble’ or ‘fine’ in the context of moral philosophy, ‘beautiful’ in the context of
aesthetic evaluation. Aristotle agrees with Plato’s view that the aim of the true
politician is to make the citizens ‘good and obedient to the law’ (EN I.13,
1102a7–10). He tells us that the law prescribes conduct in accordance with all
the virtues (5.1,1129b19–25, quoted above; also 1130b22–24)—not, of course, by
telling the citizens to act temperately or courageously, but by prescribing ways of
acting that require courage or temperance. Courage, for instance, consists not
simply in facing dangers, but in facing them for a noble reason such as defending
one’s city. Temperance consists not simply in refraining from pleasure, but in not
pursuing pleasure when doing so would cause harm—in the case of adultery, the
harm would be done to another person, in cases like drunkenness, to ourselves by
making us unable to function as reasonable and responsible agents. In this way,
the law also indicates what counts as ‘the mean’ or what one ought to do (ὡς δεῖ)
in certain types of situation. In other words, if there are principles of morality to be
found in Aristotle’s works, they should be looked for under the heading of justice,
not practical wisdom.
Now, legislation is clearly a part of politics proper, and it can be done well or
badly. As Aristotle says in Pol. III 4 (1277b7–32), in the best city—and only
there—the virtue of the good man will coincide with that of the good citizen.
Hence, the place for a discussion of moral principles is presumably not the Ethics,
but the Politics.

⁶ ἡ δὲ δικαιοσύνη πολιτικόν· ἡ γὰρ δίκη πολιτικῆς κοινωνίας τάξις ἐστίν, ἡ δὲ δίκη τοῦ δικαίου κρί
σις. The text is difficult, and some editors (notably Ross in his OCT) have emended it. My translation
is an attempt to capture what I take to be the point Aristotle wants to make.
⁷ Here I am indebted to T. Irwin’s study ‘Aristotle’s Conception of Morality’ (Proc. of the Boston
Coll. in Ancient Phil. I, 1986, 115–43), though I want to go beyond Irwin in claiming that ‘justice’ in the
sense of obedience to law actually is Aristotle’s term for what we now call morality. It seems
unfortunate to me that in making his case against Bernard Williams’ claim that there is no such notion
in Aristotle, Irwin draws almost exclusively on the Rhetoric rather than the Politics.
  ’      137

I cannot here enter into a detailed discussion of what Aristotle has to say about
justice in the Politics, but what he says about the aim of good legislators in both the
Ethics and the Politics seems to me to confirm the view that he sees morality as a
matter of the order of society and the relations between citizens. In the opening
chapter of EN V, Aristotle asserts that legislators prescribe action in accordance
with all the virtues. In the Politics, this claim is supported by the argument that the
end of the civic community is ‘living well’ (Pol. 1.2)—and we already know from
the Ethics that this means leading a life of virtue. This point is taken up again in III
9 (1280b7–12), where Aristotle rejects the idea of law as a mere contract aimed at
guaranteeing protection from mutual harm, but not concerned with making the
citizens good and just people. As I have already suggested, legislators or rulers will
try to achieve this aim by giving their citizens reasons to act for noble ends. For
example, a citizen would be expected to contribute to the common good—not just
by paying taxes, but also (in ancient Athens) by financing dramatical perform-
ances or erecting public monuments, thus exercising the virtues of generosity or
magnificence. Again, citizens will have to serve in the army and go to battle to
defend their country, which requires courage and bravery. Their ability for
practical reasoning will be developed in an education that teaches them both to
be ruled and to rule—and ruling over equals is seen by Aristotle as the highest
exercise of practical wisdom. In this way, then, engaging in the activities of a
citizen enables the members of a community to lead an active life in accordance
with virtue—as laid down in Aristotle’s definition of the human good. It is true
that in EN VI (8,1141b29–1142a10), after having said that political expertise and
practical wisdom are the same disposition (though not the same ‘in being’, i.e. in
definition), Aristotle mentions that the label ‘practical wisdom’ is usually thought
to apply only to the wisdom exercised in looking after one’s own individual affairs,
because politicians tend to be seen as busybodies. But he immediately adds: ‘but
no doubt one cannot run one’s own life well without some expertise in household
management or without some form of government’, and notes that how one
should manage one’s own life is an open question that should be investigated
(1142a10–11). This may be a rare concession to the view that one could also lead a
good life without being involved in the political life of the city, but in the Politics
Aristotle clearly thinks that acting on a larger scale is more admirable. The only
exception is the intellectual virtue of pure theoretical thought, which requires little
or no cooperation. But even here Aristotle says explicitly that the rulers must
provide room for it (EN 6.13,1145a6–9).
The aim of making citizens virtuous is also implicit in the requirement that
legislators must seek to establish (civic) friendship among the citizens. This
friendship manifests itself in intermarriages between families and shared activities
of citizens such as religious ceremonies, which show the citizens’ decision to lead
their lives in common (3.9,1280b35–39).
138    

The relation between justice and friendship is perhaps best illustrated in a

slightly paradoxical passage from EN VIII (1,1155a22–28):

Friendship would also seem to be what holds states together, and legislators to be
concerned about it even more than about justice. For unanimity seems to be
something like friendship, and this is their main aim, while they fend off civil
conflict most of all as being a kind of hatred. And when people are friends they
have no need of justice, while when they are just they need friendship as well; and
the highest form of justice is thought to hold between friends.

This passage might seem at first to speak against the identification of morality with
justice, but I think the contrast which Aristotle appears to draw first between
justice and friendship, then to deny again by saying that friendship is the highest
form of justice is actually meant to be the distinction between morally correct
conduct—which is all that the law can prescribe—and the moral attitude of a
person who accepts justice as a good and hence goes beyond what the law requires.
Friendship includes, but goes beyond mere moral obligation, one might say,
remembering that for Aristotle the truly virtuous person is the one whose inclin-
ations are in perfect harmony with her sense of duty.
One might protest at this point that morality has a much larger scope than
legislation (and vice versa, of course). Aristotle does occasionally acknowledge
this, as, for example, in EN X (9,1180a34–b2) where he says that it makes no
difference whether the laws are written or unwritten. But the insouciance of this
remark should remind us of two historical facts: first, that the Greek word νόμος
means both ‘law’ and ‘custom’, and second, that the discipline of jurisprudence
either did not exist or was, at best, in its infancy when Aristotle wrote. It is no
accident that ‘jurisprudence’ is a Latin word. Even though in speaking of legisla-
tion Aristotle must surely have been thinking of officially enacted laws, he may not
have spent much time asking himself what kinds of things should or could be
prescribed or prohibited by law, and what should—whether for reasons of prin-
ciple or of practicality—be beyond the scope of legislation. Like Plato, Aristotle
considers the law as a means of moral education—education, that is, for those who
are capable of seeing the point of the rules, coercion for those who are allegedly
unable to control their base impulses (EN 10.9,1180a4–5). Law and custom, then,
will provide some rules of morality. But neither of these offer rules that can be
applied ‘mechanically’, as is sometimes suggested by anti-Kantians. Again like
Plato, Aristotle is acutely aware of the fact that legal rules are generalizations that
will not always fit all the relevant features of a particular situation. Legislators aim
at the common good and at harmony among the citizens, and they formulate their
laws with these aims in mind. But they cannot provide for all possible circum-
stances in advance, and so the application of the law– as any judge or lawyer
knows very well—requires more than a reading of books. This is why Aristotle
  ’      139

insists on the need for experience. What he has in mind is not the kind of
experience—systematic empirical observation—that forms the basis of skills and
theories and can be passed on to others, but rather personal experience that will
help one to find the appropriate response in a given situation. This is what is
needed to act ‘at the right time, on the right occasions, in relation to the right
persons, for the right reason, in the right way’ (EN 2.6,1106b21–22)—all the
qualifications Aristotle attaches to his formulation of the ‘mean’. The analogy
with medicine holds here, too: a well-educated young person who begins to be an
independent agent is in a situation similar to that of a student of medicine who
begins his internship. This kind of experience is indeed not codifiable. Each
person has to acquire it for herself—not in the absence of rules, but because it is
needed to apply the rules in the right way. Young people who have been brought
up in the right way will have acquired some knowledge of what counts as good and
right, and they will also have formed emotional attachments to these values that
will enable them to notice the morally significant features of a given situation.
When Aristotle says, as he does both with respect to medicine and to ethics, that
experience without theory is better than theory without experience, he is clearly
thinking of personal experience. (For medicine, see Met. A 1, 981a12–24; for
ethics, EN 6.7,1141b14–22.) It is because of their personal experience that we
‘ought to pay attention to the undemonstrated sayings and opinions of experi-
enced and older people or of people of practical wisdom no less than to demon-
strations’, as Aristotle says, ‘for having acquired an eye from experience, they see
aright’ (EN 6.11,1143b13–14). One might be surprised to find unargued experi-
ence placed on the same level as scientific demonstration, but Aristotle is acknow-
ledging a familiar fact: namely that although experience may rank below scientific
knowledge in the order of understanding, it also vastly exceeds the scope of what
we can explain, and even what we can clearly articulate. The comparison of
experience in this sense with perception is very much to the point. We do not
study psychology to make some sort of sense of what others do, and we do not
study decision theory to weigh the consequences of different ways of acting. One
result of such experience is what Aristotle calls ἐπιείκεια (equity or decency) in
book V—the ability to see when and why strict adherence to the letter of the law
would be a mistake. In other contexts we might call it tact, or patience, or
psychological astuteness—abilities that depend on a sense of the character of a
person or the special features of a situation, and that go beyond what is captured
by general rules. So practical wisdom, like medicine, will be guided by general
rules, but—also as in medicine—these rules have to be supplemented by personal
experience. Even if we consult others—experts or friends, as the case may be—our
decisions will in the end depend on our overall judgement of the circumstances
and the persons affected by our actions, and this part cannot be covered by any
textbook.
140    

To return now to the controversies I mentioned above: it seems to me that those

who maintain that Aristotle deliberately offers no rules at all are concentrating
exclusively on the part played by personal experience, while those who think that
deliberation must be guided by the agent’s conception of the final good, overlook
the crucial role of justice in moral deliberation. What distinguishes Aristotle—
and, for that matter, other Greek philosophers—from modern authors like Kant is
not that he thinks there can be no rules of morality, but that he does not clearly
distinguish between moral and legal rules, and therefore definitely sees moral rules
as rules of society. One might perhaps find a precursor of the modern distinction
in the difference between written and unwritten laws, and an even clearer
approximation in the Stoic distinction between the laws of universal reason that
apply to all human beings as opposed to the specific legal codes of actual states.
But it took many more centuries before morality could be described as autonomy
in the modern sense—giving laws to oneself.
If I am right about the place of general justice in the larger scheme of
Aristotelian political science, it should be clear that the aim of becoming better
agents will not be achieved by reading only the Ethics. This also means that
individual morality, according to Aristotle, cannot be studied in isolation from
the political framework in which we live as agents. Only in the best city will the
virtue of good men in general coincide with the virtue of the best citizens, and
hence the best life can only be realized in the best social order. It follows that it is
probably a misunderstanding to see the Ethics as a book that is meant to advise its
readers about their individual ‘rational life plans’, to use Rawls’ phrase. One might
be led to think so because of the remarks in the opening chapters of both the EE
and the EN about the usefulness of knowing what eudaimonia consists in, and in
particular by the line in the EE (1.2,1214b10–11) which says that ‘it is a sign of
great folly not to have one’s life organized with a view to some end’. It is not very
clear from the context in which these remarks occur what kind of end Aristotle is
thinking of, but I suspect he may mean not much more than that one should have
a clear idea of what one values most in life—virtue or pleasure, as the traditional
alternatives tended to be presented by the Greeks, with the Platonic addition of
pure theoretical thought.
Aristotle’s general account of the best life for humans can hardly provide much
in terms of individual decisions. What Aristotle sets out in both his treatises could
be better described as the conditions that make a good human life possible. This
general theory is of importance primarily for the politician and legislator who can
have a direct influence on the organization of the lives of his fellow citizens. Now,
Aristotle may very well have thought of his audience as consisting of potential
legislators, but even apart from that, he is no doubt right in thinking that
systematic reflection on this topic may make every one of us better as moral
agents. Even if one does not need to refer to the general theory in every deliber-
ation, both one’s actions and one’s life plan (if one has one) will be influenced by
  ’      141

this kind of reflection. Furthermore, even a good education will probably not
provide one with a lot of insight about how the various things one has learned to
value hang together, or how they should be ranked. Understanding this may well
serve to strengthen the moral character of people who are already inclined to think
and act in the right way. Education and the law provide, as it were, the facts which
an Aristotelian science sets out to understand and explain, and I think that both
the Ethics and the Politics serve precisely this purpose. By showing us why we have
good reason to think that moral virtue is essential to a good human life or why
good rulers should act in the interest of the common good and try to foster a spirit
of friendship among the citizens, and how they might go about doing that, we gain
a better understanding of what might be good or bad in our own lives and
communities, and that, one hopes, might also contribute to making us better
agents.
So a prospective student of Aristotle would be well advised to attend both
semesters of the course on what Aristotle calls, at the end of the EN, ‘the
philosophy of human affairs’ (ἡ περὶ τὰ ἀνθρώπεια φιλοσοφία, 10.9,1181b15).⁸

⁸ Earlier versions of this chapter were presented at the annual Ancient Philosophy conference in
Chicago, at the Political Philosophy workshop at Johns Hopkins University in Baltimore, and at the
University of California, Los Angeles. I am grateful to Paula Gottlieb for her detailed comments on a
very rough first draft, and to the audiences at Johns Hopkins and UCLA for helpful criticism and
suggestions for improvement. Dorothea Frede also read an early draft and encouraged me to continue.
 11
Two Kinds of Deliberation:
Aristotle and the Stoics

Both Aristotle and the Stoics developed accounts of deliberation; but while
Aristotle focused almost exclusively on finding the best way to reach a given
goal, the Stoics were above all interested in deliberation about right and wrong.
The first part of this chapter discusses the various meanings of the word
προαίρεσις and argues that the result of deliberation is often a plan for action.
By distinguishing between theoretical and practical reasoning, Aristotle also
showed that there are two different intellectual virtues. The Stoics did not take
over this distinction, including practical wisdom in wisdom in general, and
describing it as concerned with what ought or ought not to be done. Stoic
deliberation aims at determining what it is appropriate to do (or not to do)
according to both human and universal nature, but since the success of an
action does not affect its moral character, they did not find effective means–
ends reasoning necessary for virtue.

Deliberation is a kind of reasoning in which one engages to answer the question

what should I do? This question can be understood in at least two different ways,
depending on the situation. It may mean: what should I do to reach a certain goal?
Or it may mean: should I do this, or not? Both kinds of deliberation are forms of
practical reasoning, distinguished from theoretical reasoning by the fact that it
should lead to action, whereas theoretical reasoning aims at the discovery or
demonstration of truth.
The distinction between practical and theoretical reasoning appears first in
Aristotle’s Ethics. Plato does not yet have it, though examples of both kinds of
deliberation occur in his dialogues. He seems to have thought that an expert in
either a science or a technical skill will understand his subject well enough not
only to give an account of it, but also to produce it if it is something that can be
produced. A doctor, for instance, would know what health is, and this would
enable him as far as possible to restore or to preserve the health of his patients.
The most conspicuous example of this conception of expert knowledge appears in
the Republic, where Plato assumes that the philosophers who have attained

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0011
    :     143

knowledge of the Good will be the best rulers, able to make the city and its
citizens good.
As the examples show, this is a questionable view, and it might well be one of
the things that led Aristotle to his distinction between practical and theoretical
thought. For, unlike Plato, Aristotle did not think that philosophers would be the
best politicians. He agreed that good rulers should indeed have quite a lot of
knowledge (what he calls ‘political science’) but insisted that even more important
for political expertise is experience, since politicians have to deal with particular
persons and actions. Their primary virtue is practical, not theoretical wisdom, and
the exercise of practical wisdom consists in good deliberation. So we find in both
of Aristotle’s Ethics explicit discussions of deliberation, and also the distinction
between the two intellectual virtues—theoretical and practical wisdom (σοφία and
φρόνησις, respectively).
After Aristotle, the Stoics also developed a theory of deliberation, of which we
find a relatively late and probably somewhat unorthodox version in Cicero’s De
Officiis, based on a work by the Stoic Panaetius (second century ). Comparing
the two theories, one notices that, while Aristotle’s official account of deliberation
in the Ethics deals only with deliberation about how to reach given ends,¹ the
Stoics focused almost exclusively on questions about whether or not one should
do a proposed action.
Why this difference? I will argue that it reflects different conceptions of a
successful life of action.

Aristotle—Deliberation as Planning

Aristotle describes practical deliberation in chapters III 2 and 3 of the

Nicomachean Ethics (EN) and chapter II 10 of the Eudemian Ethics (EE), where
he distinguishes merely voluntary action from action on the basis of reasoning. In
order to characterize action based on reasoning, he introduces the concept of
προαίρεσις. The Greek word is usually translated as either ‘choice’ or ‘decision’,
but I will leave it untranslated for the moment. Aristotle explains that προαίρεσις is
‘thought to be most closely bound up with virtue, and to discriminate characters
better than actions do’(EN 3.2,1111b5–6).
As Gauthier and Jolif note in their commentary on the EN (II,189–90),
προαίρεσις was a novelty in the vocabulary of philosophy and even to some extent
in ordinary Greek at the time, but perhaps because Aristotle goes on to discuss the
views of other philosophers about προαίρεσις, they do not ask what might have
motivated the innovation. Aristotle goes through a number of possible candidates

¹ For a detailed defence of this point in a different context, see K.M. Nielsen, ‘Deliberation as
Inquiry’, Phil. Review 120, 2011, 383–421.
144    

among psychical phenomena that could perhaps be treated as representing or

including προαίρεσις, namely all three forms of desire recognized by Plato and his
followers and also judgement or belief (δόξα). In the EE he describes this proced-
ure as trying to determine the genus, which could be either desire or judgement. In
each case, Aristotle argues that these things could not be identified with
προαίρεσις. It seems that προαίρεσις did not easily fit into the psychological
vocabulary of the Academy, and so Aristotle ends up with the somewhat awkward
solution that it must be a combination of desire and thought—either a deliberate
desire for something in our own power, as he puts it in EN III 3 (1113a11; cp. VI 2,
1139a23) and EE II 10 (1226b17), or desiderative thought (EN VI 2, 1139b5),
leaving it open whether desire or thought should be considered as the genus.
Aristotle has more to say about this in EN VI 2, but the formula remains the same.
προαίρεσις is also described as the final point of deliberation and the starting point
of action (EN III 3.1113a3–7), but again this does not clearly show what it is
supposed to be. A mental act, perhaps, as suggested by the English words ‘choice’
and ‘decision’? But then why would Aristotle assume without argument that it
must be preceded by deliberation?
We do not actually have a text from a philosopher before, or contemporary
with, Aristotle that argues for any one of the alleged alternative views. There is no
discussion of προαίρεσις in Plato. According to des Places (Lexique), the word
occurs only once, namely at Prm. 143C2 (also cited by LSJ), in a passage that has
nothing to do with deliberation or decision. The word just means ‘picking up’ or
selecting from a list of items, with no implication of preference. There are a few
more occurrences of the corresponding verb, mostly in the same neutral sense,
though sometimes in the sense of preferential choice. Given that we have very little
of the philosophical writings of Aristotle’s contemporaries, it is, of course, possible
that others—presumably members of the Academy, since Aristotle’s list of can-
didates looks thoroughly Academic—had offered accounts of προαίρεσις. But
again, this does not seem very likely: there is no mention of προαίρεσις as opposed,
for example, to ἐπιθυμία, βούλησις, or δόξα in the Topics, which often reflects the
interests of Aristotle’s early colleagues. It is at least as plausible to think that here,
as elsewhere, Aristotle is simply reporting the views of others in his own termin-
ology, saying no more than that others dealt with what he would call προαίρεσις in
connection with one or the other form of desire, or with judgement. Hence it
makes sense to ask why Aristotle thought it important to find a place for
προαίρεσις in the vocabulary of moral psychology.
Now, since the word seems to mean either simply choice or preferential choice,
one might at first suppose that Aristotle intended to mark a gap between desire
and action, or between thought and action, as the Stoics did with their notion of
assent. It is one thing to desire something, another to act on the desire, and also to
think that one should do something and to actually do it. And in the case of
rational animals, one might say that the step from desire or thought to action
    :     145

consists in choice or decision. But this cannot have been Aristotle’s reason for
introducing the notion of προαίρεσις, since he makes it perfectly clear that people
can and often do act from appetite or emotion without προαίρεσις. On the other
hand, Aristotle assumes throughout and without further argument that προαίρεσις
requires deliberation and reasoning, which is why he holds that children and non-
human animals do not act from it. Why would he take it to be so obvious that
choice or decision requires deliberation?
This brings us to the often-noted point that neither ‘choice’ nor ‘decision’ seem
to be entirely adequate (English) translations of the Greek term. After all, even
young children do make preferential choices—for example, preferring chocolate
ice cream to vanilla; and they also, I would say, make decisions—for example,
about what to wear on a cold day—that may lead to protracted discussions with
their parents. Some scholars have, therefore, resorted to translations such as
‘deliberate choice’ or the like to indicate what Aristotle seems to have in mind.
But why did he take the word προαίρεσις to express this? Anthony Kenny² has
recently proposed ‘resolution’, taking into account cases in which proairesis does
not result in a particular action, as, for example, the resolution of self-controlled
persons not to give in to certain kinds of temptation. Yet none of these translations
explains why Aristotle takes it for granted that a proairesis must be based on
deliberation: choices and even decisions can be made by children who are not yet
capable of deliberation, and New Year’s resolutions usually are not the result of
lengthy reflection, though one may indeed assume that they have been preceded
by some relevant thought.
At this point it helps, I think, to return to the dictionary. LSJ list quite a few
different meanings of προαίρεσις, though they begin with (preferential) choice.
The most relevant ones, I think, are these:

(1) choosing one thing before another; resolution

(2) purpose, plan, scope of action
(3) (in political language) deliberate course of action, policy
(5b) sect or school of music or philosophy

As Gauthier and Jolif note in their commentary,³ the word προαίρεσις has the
process/product ambiguity familiar from other Greek nouns ending in -σις:
αἴσθησις may mean both the act of perceiving and the impression received;
διήγησις may mean both the telling of the story and the story itself, and notori-
ously, αἵρεσις means not just a choice, but also what is chosen—namely a sect or
party (as in ‘heresy’). Similarly, then, προαίρεσις may also mean both the choice

² ‘Practical Truth in Aristotle’ in Episteme, etc., ed. B. Morison/K. Ierodiakonou, Oxford 2011,
277–84.
³ R.E. Gauthier and J.Y. Jolif, L’Éthique a Nicomaque, 2nd ed., Louvain/Paris 1970, vol. I, 189–90.
146    

and what is chosen. And Gauthier/Jolif point out that the second sense is actually
more common in the fourth century orators than the first.
Aristotle himself also uses the word in several of these senses, as a few
examples show:

EN X 9.1179a33–35:
If these matters and the virtues, and also friendship and pleasure, have been dealt
with sufficiently, are we to suppose that our project [i.e. the present investigation]
has reached its end?
Ἆρ’ οὖν εἰ περί τε τούτων καὶ τῶν ἀρετῶν, ἔτι δὲ καὶ φιλίας καὶ ἡδονῆς, ἱκανῶς
εἴρηται τοῖς τύποις, τέλος ἔχειν οἰητέον τὴν προαίρεσιν;
Rhet. I 1,1355b17–21:
. . . for sophistry lies not in the capacity, but in the purpose, . . . and in dialectic a
man is a sophist with respect to the purpose, but a dialectician not with respect to
the purpose, but with respect to the capacity.
(i.e. sophist and dialectician use the same skill, but differ in their purpose or plan)
σοφιστὴς μὲν κατὰ τὴν προαίρεσιν, διαλεκτικὸς δὲ οὐ κατὰ τὴν προαίρεσιν ἀλλὰ
κατὰ τὴν δύναμιν.
See also Met. Γ 2,1004b24–25: sophistry differs from dialectic τοῦ βίου τῇ
προαιρέσει [life plan]
For the connection with deliberation, see EN V 8,1135b19–27:
When someone acts with knowledge but not after deliberation, it is an act of
injustice—e.g. the acts due to anger or to other passions necessary or natural to
man . . . But when a man acts from a plan, he is an unjust man and a vicious man.
Hence acts proceeding from anger are rightly judged not to be done of malice
aforethought . . .
ὅταν δὲ εἰδὼς μὲν μὴ προβουλεύσας δέ, ἀδίκημα, οἷον ὅσα τε διὰ θυμὸν καὶ ἄλλα
πάθη . . . οὐ μέντοι πω ἄδικοι διὰ ταῦτα οὐδὲ πονηροί· . . . ὅταν δ’ ἐκ προαιρέσεως,
ἄδικος καὶ μοχθηρός. Διὸ καλῶς τὰ ἐκ θυμοῦ οὐκ ἐκ προνοίας κρίνεται· οὐ γὰρ ἄρχει
ὁ θυμῷ ποιῶν, ἀλλ’ ὁ ὀργίσας.
(cp. EN V 8,1136a1; Rhet. I 13,1374a9–16)

When deliberation consists, as Aristotle describes it, in searching for a way to

bring about a certain end, finishing when one has reached the kind of action that
can be done without further preparation, this means, not necessarily that one will
immediately take the first step, but simply that one’s plan is now complete and can
be put into practice when the occasion arises. And what is chosen or decided upon
at the end of deliberation should be, as the above examples indicate, a plan or
policy rather than a particular action. Deliberation ends, as Aristotle says, when
one has reached something that the agent can do without further preparation.
    :     147

This will then be the first step in the realization of the plan; but that is not to say
that deliberation must lead to immediate action. As has often been noted, in the
case of the acratic and the self-controlled person, their resolution, say, to resist
temptations to inappropriate pleasures will have been made before such an
occasion arises—it is a choice of policy.
If we take the act of προαίρεσις to be the adoption of an action plan or policy, it
is easier to understand what Aristotle means when he says such things as that ‘the
virtues are a sort of προαίρεσις, or not without προαίρεσις’ (EN II 5,1106a2–4: Ἔτι
ὀργιζόμεθα μὲν καὶ φοβούμεθα ἀπροαιρέτως, αἱ δ᾽ ἀρεταὶ προαιρέσεις τινὲς ἢ οὐκ
ἄνευ προαιρέσεως): the conduct of a morally virtuous person could plausibly be
understood as being based on the adoption of a set of action policies or moral
principles. Or ‘the προαίρεσις of living together is friendship’: people who have
chosen a policy of living together are friends (Pol. III 9, 1280b38 ἡ γὰρ τοῦ συζῆν
προαίρεσις φιλία). Again, an acratic person is acting against her usual policy
(perhaps even in choosing to make an exception), whereas the self-controlled
person sticks to her policy in the face of temptation.
So far I have said nothing about the so-called ‘practical syllogisms’⁴ (not
Aristotle’s own term) that are sometimes taken to be typical examples of practical
deliberation. Aristotle himself says that the conclusion of this kind of reasoning is
an action, as in the following example: I need a drink; here is a drink—at once he
drinks (de Motu Animalium 7,701a32–33). However, such quasi-syllogisms obvi-
ously do not represent a search for means and ways to reach a given end, as
Aristotle describes deliberation in the Ethics; rather, they serve to illustrate the
transition from a plan or a desire to action. This is the question Aristotle addresses
in the chapter of the de Motu in which he compares practical thinking to a
theoretical syllogism (701a7–12): how is it that thinking sometimes leads to
action, sometimes not, and that an animal sometimes moves and sometimes
does not move? In the case of humans, the transition still involves practical
reasoning, but this is not strictly speaking a part of deliberation; the second
premise in the ‘syllogism’ consists in noticing an occasion where a policy should
be applied, or the presence of available means for taking action. In EN III 3
(1112b33–1113a2), Aristotle says that just as ends are not the subject of deliber-
ation, so neither are the particulars that belong to perception. He emphasizes the
role of perception in his discussion of practical wisdom in EN VI: the ability to
recognize what can and should be done in a particular situation is a matter of
experience, not knowledge, and this kind of recognition is compared to the role of
νοῦς (insight) in the theoretical sciences, which recognizes the truth of first
principles that must be known without demonstration (EN VI 8,1142a14–30;

⁴ For this interpretation of practical syllogisms, compare K. Corcilius, ‘Two Jobs for Aristotle’s
Practical Syllogisms?’, Logical Analysis and History of Philosophy 11, 2008, 163–84; and P. Fernandez,
‘Reasoning and the Unity of Aristotle’s Account of Animal Motion’, Oxford Studies in Ancient
Philosophy 47, 2014, 151–204.
148    

11,1143a32–b5 for νοῦς). προαίρεσις provides only the first premise in the ‘syllo-
gism’; the second must come from perception or experience.
Aristotle’s examples illustrate action more broadly, including but not limited to
the implementation of an action plan. They show how the thought of a desirable
type of object together with the perception of a token of the type may lead to
action: the agent is thirsty and wants to drink; she sees a glass of water and
proceeds to drink it. Hence Aristotle’s analogy can also be generalized to animal
action: just as a thirsty person who sees a glass of water will drink it, so a hungry
lion who sees a gazelle will go for it.
A plan or a resolution is the starting point of action in the sense of leading to a
psychological state in which an animal—human or otherwise—will take action
once it perceives an opportunity to fulfil a desire or an occasion for putting a
policy into action. This is why Aristotle can say that the courageous man will
typically be ready to respond quickly to sudden and unforeseen events without
having to engage in reasoning (EN III 8,1117a17–22). He is still acting from
proairesis, because he has already adopted the right principles, helped by educa-
tion and the law, so that he will act immediately once he recognizes a situation
where the relevant principle applies.
The interpretation of deliberation as resulting in a plan for action is confirmed,
I think, by Aristotle’s treatment of practical wisdom in EN VI. Aristotle begins the
chapter that leads to the definition of φρόνησις by considering what kind of person
one would describe as φρόνιμος: the man who is good at deliberating about what is
good or beneficial for him with respect to living well in general (VI 5,1140a25–28).
Having defined practical wisdom as ‘the disposition to act, based on true reason-
ing, with regard to what is good or bad for humans’,⁵ Aristotle gives his one and
only concrete example of such a person in that book—he says that ‘this is why we
think that people like Pericles are practically wise, because they are capable of
considering (θεωρεῖν) what is good for themselves and for humans in general. And
we take these people to be managers of households or politicians’ (VI
5,1140b7–11). Later (VI 8,1141b23ff), Aristotle declares that practical wisdom
and political expertise are the same, though not in definition, and that this holds
despite the common tendency to think that practical wisdom belongs only to
private individuals because politicians tend to be busybodies who are out for their
own advantage. Aristotle’s extremely simple examples of practical syllogisms tend
to obscure the fact that deliberation, and especially political deliberation, is
actually quite difficult and requires a lot of factual background knowledge.
A survey of the requirements for politicians can be found in the section on
deliberative speeches in Rhet. I 4,1359b20ff. They include the state’s financial
resources and all its expenses; its actual and potential military power; what wars it

⁵ EN VI 1140b4–5: λείπεται ἄρ’ αὐτὴν εἶναι ἕξιν ἀληθοῦς μετὰ λόγου πρακτικὴν περὶ τὰ ἀνθρώπῳ
ἀγαθὰ καὶ κακά; reading ἀληθοῦς instead of ἀληθὴ with Susemihl.
    :     149

has fought and how, and the same for neighbouring states and possible enemies;
next, one needs to know what the defences of the country are and of what kind and
size; furthermore, how much is needed for food to support the city, what is
produced at home and what has to be imported or exported; and finally, one
needs to know about legislation—that is, the forms of government and what is
expedient for each of those. If this reads a bit like a table of content for Aristotle’s
Politics, that is, of course, no accident. Aristotle concludes this chapter of the
Rhetoric by pointing out that all of these topics properly belong to political science
rather than to rhetoric (I 4,1360a37). The speaker needs to know the particular
facts about the given situation; he is not expected to have the more general
knowledge about economics, or military strategy, or constitutional arrangements
that the trained politician should have. But I think that this knowledge of
particular facts is the kind of experience that Aristotle has in mind when he says
that ‘one should listen to the unproven pronouncements of experienced, older and
wise persons no less than to demonstrations, since experience has given them an
eye so that they see aright’ (EN VI 11,1143b11–14).
I am emphasizing the large scope of experience here in order to show why
Aristotle regards practical wisdom as equal in importance and difficulty to
theoretical knowledge. This could hardly be the case if it consisted just in making
the right decisions in everyday life suggested by Aristotle’s concrete examples. In
fact, the examples in EN VI typically serve to point out possible errors or
deficiencies in practical reasoning at the point of acting, not to illustrate the
deliberations of political rulers or even of private citizens. The achievement of a
wise ruler lies in devising a plan for action that will be beneficial to the commu-
nity, feasible under the circumstances, and successful. This is why practical
wisdom counts as an intellectual virtue, attained only by the best individuals. In
the Politics Aristotle even goes so far as to claim that φρόνησις is the one virtue that
properly belongs (ἴδιος) only to rulers—their subjects may just need true belief
(Pol. III 4,1277b25–30). In the Ethics, he concedes that there is deliberation also
about private affairs, but he sets that aside for the time being, saying that ‘it is
unclear how one ought to manage one’s own affairs, and this should be investi-
gated’ (VI 8,1142a10). It is one of the promises that he apparently did not fulfil.
Now, planning—looking for the best available way to reach a goal—is just the
kind of reasoning that Aristotle describes as a kind of search or inquiry, by
contrast with scientific or demonstrative reasoning. Ends do not imply specific
means, nor do means imply the end. Practical inquiry is concerned with causal
relations and in this respect also different from mathematical analysis, with which
Aristotle occasionally compares it⁶ (EN III 3,1112b20–25). The mathematician is
trying to find a way of deriving a theorem from another theorem or axiom and will

⁶ EN III 3.1112b21–23, where Aristotle also says that not every inquiry is a deliberation; not, for
instance, mathematical inquiries. Cp. EN VI 9,1142a312.
150    

then be able to construct a deductive proof. The deliberator typically considers

several possible ways of reaching an end and then tries to find out what will be the
best under the circumstances; and the result is, of course, not a proof, but a plan
that may or may not turn out to work.
What, then about the other kind of deliberation, where a person has to decide
whether or not to take a particular proposed action? I do not want to suggest that
Aristotle has simply overlooked this, or that he would not call it deliberation.
Quite the contrary: in the Rhetoric (I ,1358b8–10), Aristotle tells us that delibera-
tive speeches will consist mainly in arguing for and against—presumably given
proposals in the assembly; but these proposals are judged as possible means or
methods to reach the overarching aim of benefiting the community. Questions of
just or unjust, noble or shameful may also come up, but, as Aristotle drily remarks,
only in a minor role (1358b20–25). So the weighing of alternatives is treated as a
part of the process of deliberation that leads to the adoption of an action plan.
A public deliberation will end in a vote on something that can be done (EN VI
8,1141b27). What is accepted by such a vote is usually a policy or a plan for some
action that is expected to contribute to the common good. The final decision may
indeed be made by a preferential choice, but it is a choice between better or worse
plans, not whether or not to take a particular action. Again, in the deliberations of
individuals, several means or ways of attaining the goal, if available, will no doubt
be considered, but the final decision will consist in accepting what seems to be the
best plan. So Aristotle was quite right in saying at the end of EN III 2 (1112a16)
that the very word προαίρεσις seems to indicate that it is something chosen in
advance, ‘before other things’.
In his official discussions of deliberation in EN III and VI, Aristotle restricts his
use of the word to a specific kind of deliberation, which does not mean that he
could not use it in a wider sense in other contexts.⁷ In fact, he does this in the very
next book of the Ethics, in his treatment of self-control and weakness of will.
The resolutions of the self-controlled or the weak-willed, and for that matter also
of the unrestrained pleasure-seeker, can hardly be based on deliberation in the
sense introduced in book III: one does not need to look for ways and means to
avoid adultery or drunkenness. It is sufficient to understand the reasons for these
policies, namely that adultery and so on would be harmful, to others or to oneself.
Yet these resolutions still count as προαίρεσις because they are consciously—or, as
we also say, deliberately—chosen, adopted before the particular occasions in
which they are applied.
In books III and VI of the EN, Aristotle describes as deliberation only the type
of reasoning that he takes to be most characteristic of the practically wise person.

⁷ The word φρόνησις is a similar case: Aristotle uses it in the restricted sense of practical wisdom in
the discussion of the intellectual virtues, but elsewhere also in the wider sense of intelligence or thought
(e.g. Met. Γ 5, 1009b12: διὰ τὸ ὑπολαμβάνειν φρόνησιν μὲν τὴν αἴσθησιν . . . ).
    :     151

Decisions about whether or not to do something have to be made by everybody

almost every day, and even if the decision was between right and wrong, Aristotle
may have thought that it would not be very difficult for people with the right kind
of education and character. As Protagoras remarks in his great speech in Plato’s
dialogue (327E–328A): there are no teachers of justice because children learn it
from everybody around them, just as they learn their language.
But there is also a more general reason for neglecting this second kind of
deliberation—namely that only planning involves a kind of reasoning that is
clearly distinct from theoretical thought. Questions that arise in cases of deciding
whether or not to act in a certain way would have to be answered by arguments for
and against doing the action, not by looking for ways to do it. The agent is seeking
a justification for one or the other option. She will do what she recognizes as right,
or beneficial, or healthy, and so her action depends on the correctness of her
judgement. A person who argues for or against a certain way of acting is trying to
establish the truth or falsity of a particular judgement (in ancient times, it had not
yet occurred to philosophers to think that value-judgements might be neither true
nor false). The skill needed for this kind of deliberation would be what Aristotle
calls dialectic—the ability to find and evaluate arguments on both sides of a
question. This is what orators contribute to public deliberation or court cases
where a jury has to weigh the evidence for and against a verdict of guilty or not
guilty. Dialectic is a skill that can be used in many different kinds of situation,
including philosophy; it is not specifically concerned with decisions about action.
Being good at dialectic does not count as a virtue. The virtue involved in decisions
for or against a kind of action is virtue of character, not intellectual virtue: the
greedy person will choose what is more profitable even if it is not fair, the good
person will do what is fair even if it is not profitable. If there is one type of
reasoning that is specifically directed at action, it is finding ways and means to
given goals, or planning. Now, the ability to make plans effectively and success-
fully is not dialectic, but what Aristotle calls δεινότης—cleverness or cunning
(EN VI 12,1144a23–b1). Cleverness is included in good deliberation, but being
good at planning is not identical with the virtue of practical wisdom. As Aristotle
says—‘if the aim is noble, the ability is praiseworthy, but if the aim is bad, it is
mere cunning’. Moral virtue makes the end right in both kinds of deliberation. But
Aristotle may well have been right in singling out planning as a special kind of
reasoning, and the ability to do this successfully for the right ends as an intellectual
virtue.

The Stoics: Moral Problems

If we now turn to the Stoics, we find a very different picture. The Stoics seem to be
exclusively concerned with questions about what one should or should not do,
152    

both in general and in particular situations. Unlike Aristotle, they could not rely
on education, and even less on pleasure and pain as means of inculcating the right
moral attitudes. Quite the contrary—the Stoics thought that children would
almost inevitably be corrupted by their social environment, so as to develop
attachments to the wrong things and then fall into the kind of mental disturbances
they called passions or emotions.⁸ For a Stoic, the path to virtue would begin, in
most cases, with the painful realization that most of one’s beliefs and the corres-
ponding conduct were thoroughly misguided. Only then could a proper education
to virtue begin—and for the Stoics as self-declared followers of Socrates, all human
virtues were kinds of knowledge. So Stoic education would consist mainly in
instruction, not habituation. In one of their many definitions of the good, it is
defined as the perfection of rational beings insofar as they are rational—that is, the
virtue of rational beings (presumably including the gods; see DL VII 94, cp. also
Cicero, Fin. 3. 33). Human wisdom, which comprises all other virtues, consists in
the knowledge of things divine and human, and the kind of knowledge that
pertains to action is knowledge of what to do and what not to do. The wise person
will be living in agreement with universal nature, the divine reason that organizes
and administers the world, and agreement with nature is what the Stoics regarded
as the highest human good, that is, εὐδαιμονία.⁹
The reason that governs the universe is also described as a law ‘which com-
mands what ought to be done and prohibits its opposite’ (Cicero Lgg. I 18). But
how does one find out what those commands are? The Stoic answer was: by
observing human nature, that is, the way Nature has designed human beings to
behave (DL VII 87, Chrysippus). Natural behaviour, for humans as for other
animals, is based on two original impulses or instincts. There is first the instinct
for self-preservation, which enables animals to recognize and pursue what they
need to keep themselves alive and healthy, and to avoid what threatens to harm or
to destroy them. Second, many animals, and humans in particular, also have a
natural impulse towards sociability, caring for others of the same species and
wanting to live together with them. We can assume, then, that Nature intends for
us to act according to these two basic tendencies. But recognizing these two
natural impulses does not provide much guidance for everyday action. We need
to know exactly what we should do or avoid doing in pursuit of self-preservation,
and also how exactly we should or should not behave towards others. In Stoic
terms, we need to know what counts as appropriate action (καθῆκον). The Stoics
defined appropriate action as ‘an action that has a reasonable justification’ (DL VII
107; cp. Cicero, Off. 1.8 and Fin. 3.58). A reasonable justification should consist in

⁸ See DL VII 89 and the chapter ‘de perversione rationis’ in SVF III, 53–6.
⁹ This is, of course, a very rough sketch of the Stoic doctrine. For more detail, see, for example, A.
A. Long, ‘Stoic Eudaimonism (1989)’; repr. in Long, Stoic Studies, Cambridge 1996, 179–201or
G. Striker, ‘Following Nature’ (1991); repr. in Striker, Essays on Hellenistic Epistemology and Ethics,
Cambridge 1996, 221–80.
    :     153

showing that one’s action conforms to one or the other of our original natural
impulses—and that, it turns out, is not always easy. Stoic deliberation, therefore,
consists in finding the right justification for whatever one might consider doing,
and since the justification is not always obvious, we should not be surprised to find
the title ‘On appropriate action’ on the lists of works for all the major Stoics from
Zeno on.
One problem with finding a justification was that philosophers (not to mention
ordinary people) had different conceptions of what belongs to human nature.
Zeno, no doubt influenced by his Cynic teacher Crates, seems to have held a very
restrictive view of what is needed for self-preservation: the bare necessities of life,
such as food, drink, and shelter, would be sufficient. Also like the Cynics, he
rejected many venerable customs as mere conventions, and hence not based on
nature. So, for instance, in Zeno’s ideal city, marriage would be superfluous, and
men and women would be free to have sex with whomever they liked. Money and
political institutions would not be needed, since communities living in the natural
way would be held together by mutual friendship.¹⁰ Chrysippus apparently still
shocked his audience by arguing that he could not find anything wrong with incest
(Sextus Empiricus, PH III 246), and so on. Later Stoics would argue for a more
expansive notion of what is natural for humans. After all, the arts and crafts that
had led to a more comfortable lifestyle were due to the natural intelligence and
inventiveness of humans, and hence could be accepted as natural too. In the end,
these Stoics would accept most of the actual institutions of human states—
government, marriage, and commercial relations included. Panaetius and, follow-
ing him, Cicero explicitly rejected the Cynic lifestyle (Off. 1,128, 148), but for other
Stoics the ascetic life of a Cynic remained an option—either as a ‘shortcut to
virtue’ (Apollodorus of Seleucia ap. DL VII 121, perhaps commenting on Zeno)
or, as Epictetus describes it in his portrait of the ideal Cynic (Diss. 3. 21), as a kind
of prophet or divine messenger, reminding people of the errors of their ways and
setting an example of a truly natural life.
A second difficulty arose from the fact that general rules of conduct would have
exceptions. For instance, the rule that one should not physically harm others
would not hold for self-defence or the punishment of criminals, and the rule that
one should keep oneself healthy might be overridden in situations where it was to
one’s advantage not to be healthy—as in draft evasion. (But the example comes
from Zeno’s pupil Aristo of Chios, see SE, M XI 66.) Perhaps the most remarkable
exception of this kind was the Stoic doctrine of ‘reasonable exit’, that is, suicide, in
situations in which there was no longer any possibility of acting appropriately in
other ways (see e.g. Cicero Fin. III 60–61).

¹⁰ For Zeno, see M. Schofield, The Stoic Idea of the City, Cambridge 1991, chapters 1 and 2.
154    

Third, and perhaps most menacing, it might happen that the two basic ten-
dencies were in conflict, self-preservation pulling in one direction, concern for
others in the other.
For all these reasons, advice about appropriate ways of acting would be helpful
for those who had not reached the blessed state of wisdom—and that means
almost everybody.
In order to provide this sort of advice in a systematic way, one would have to
show how certain ways of acting followed from one or the other of the basic
natural impulses and could, therefore, be considered as in accordance with human
nature, and one should also indicate what might justify an exception from those
normal ways of acting.
We no longer have any of the early Stoic treatises on appropriate action, but
Cicero’s De Officiis I and II, adapted from a book by the Stoic Panaetius, may at
least provide an impression of the contents of such treatises. This is not entirely
unproblematic, because Panaetius, who was writing for readers like the educated
Roman aristocrats of his time, avoided the complicated technical terminology of
his school, which would no doubt have seemed awkward and pedantic to such
readers. As we know from Cicero’s own exposition in de Finibus III and many
other sources, the Stoics distinguished between appropriate action (καθῆκον) and
right or virtuous action (κατόρθωμα).
Appropriate action—the subject of deliberation—was for them correct action,
but not necessarily morally virtuous action. Virtue requires not only doing the
right thing, but also doing it from the right motive, namely the desire to live in
agreement with nature, and hence ordinary people who lacked virtue could also
act correctly following their natural impulses without counting as virtuous.
Orthodox Stoics would describe deliberation about appropriate ways of acting
as concerned with ‘selection (as opposed to choice) of what is in accordance with
nature and rejection of what goes against it’. The objects of this selection—that is,
whatever contributes to self-preservation or self- development, but also what
belongs to social life—were called ‘indifferent’, that is, neither good nor bad, in
contrast to virtue and vice as the only real goods or evils. Yet those ‘indifferents’
did have different values—they were called ‘preferred’ or ‘dispreferred’, depending
on whether they were in accord with human nature or not. Strictly speaking, then,
Stoic deliberation was not concerned with what is morally right or wrong, or even
what is really useful or not, but simply with what is or is not in accord with human
nature, both with respect to self-preservation and self-development and with
respect to concern for others and sociability.
However, judging from Cicero, Panaetius seems to have set aside the termino-
logical distinctions between the two kinds of value. Instead, he had complicated the
schema for the case of rational animals by introducing specifically human natural
impulses for each of the four cardinal virtues—in his version, wisdom, justice,
magnanimity including courage, and decorum or temperance (Off. I.15)—based
    :     155

respectively on the desire for knowledge, the desire for social life, the desire for
freedom and great achievements, and the uniquely human desire for beauty. Those
four were treated under the label of the honourable, while the impulse for self-
preservation, which in other expositions of Stoic doctrine (including Cicero’s de
Fin. III) would include at least the desire for knowledge, was limited in Panaetius’s
version to the ‘comforts of life’ such as health, wealth, and power, and treated
under utility. As a result, Panaetius’s classification of problems did not match the
old distinction between the two basic natural impulses towards concern for oneself
and concern for others, since the desire for knowledge (and probably also the desire
for leadership and freedom, the motive leading to magnanimity; I 13), discussed in
the section on the honourable, would normally fall under concern for oneself.¹¹ But
this allowed Panaetius to discuss deliberation according to the old contrast between
morality and material self-interest, and to represent possible conflicts between
concern for oneself and concern for others as conflicts between morality and
self-interest.
According to Panaetius (Off. 3.7; cp. 1.9), deliberation about appropriate action
was concerned with three kinds of questions: first, is the action under consider-
ation honourable¹² or dishonourable? Second, is it useful or useless? And third,
how should one decide when what appears to be honourable conflicts with what
appears to be useful? Panaetius’s work was organized according to this list. It dealt
first with questions of what is honourable or not, devoting a chapter each to
wisdom, justice and beneficence, magnanimity and courage, and decorum or
‘fittingness’—virtues that were based on one of the four natural human impulses.
Next came questions of utility, limited—presumably with a view to Panaetius’s
and Cicero’s expected audience—to the most important factor, namely the affec-
tion and support of other people. The last book should have dealt with conflicts
between morality and utility; but though Panaetius himself had promised this, as
Cicero insists (Off. 3.7–10), for reasons that left his successors guessing, Panaetius
never wrote the final book. Cicero tried to fill the gap, partly by his own efforts and
partly by using books by other Stoics later than Panaetius.
For the purposes of this chapter it will be sufficient to consider Cicero’s
chapters on the social virtues of justice and beneficence, meant to show what
kind of action corresponds to the human impulse towards concern for others and

¹¹ It is probably no accident that knowledge (scientia, not sapientia) is treated rather perfunctorily in
two short paragraphs (I 18–19), where Cicero limits himself to the discussion of theoretical knowledge
rather than wisdom in the all-embracing sense in which it would be equivalent to virtue. We do not
know, of course, whether this was the same in Panaetius’s book. The section on magnanimity (I 61–91)
is longer because it also covers courage and contains an extended discussion of the relative merits of
military versus civil achievements.
¹² honestum—I prefer the translation ‘honourable’ to ‘morally good’, since the honestum includes
more than moral virtue—for instance, theoretical wisdom and ‘great deeds’. Unless otherwise indi-
cated, in this as in most quotations in the text I follow the translation by E.M. Atkins in M.I. Griffin and
E.M. Atkins, Cicero, On Duties, Cambridge UP 1991, with occasional modifications.
156    

living in a community with them, and his treatment of the concern for ‘the
comforts of life’ (commoda vitae, 2.9) in book II, which clearly belongs to the
impulse for self-preservation, and a few examples of conflicts from book III.
Cicero begins his exposition on justice with two basic principles: ‘Of justice, the
first task (munus) is that no man should harm another unless provoked by
injustice; the next that one should treat common goods as common, and private
ones as one’s own’ (Off. I 20). The first principle can easily be accepted as a
necessary requirement of communal living, but what about the second? Here
Cicero has to add an argument: ‘No property is private by nature, but rather by
long occupation (as when men moved into some empty territory in the past) or by
victory (when they acquired it in war), or by law, by settlement, by agreement, or
by lot . . . The distribution of private property is similar to this. Consequently, since
what becomes each man’s own comes from what had in nature been common,
each man should hold on to whatever has fallen to him. If anyone should seek any
of it for himself, he will be violating the law of human fellowship’ (Off. I 21). The
reasoning is less than convincing, especially with respect to things acquired by
victory—and one might add that the naturalness of private property is still a
matter of debate today. Cicero is clearly speaking from concerns of his own at the
time of civil war.
The other main principle introduced under justice is good faith in promise-
keeping, with the usual exceptions (promises made under the threat of violence
are not valid, etc.), and finally the rules of war, where Cicero treats wars over
territory between empires as a kind of competition among rivals which should be
regulated by rules of fairness (Off. I 38–41). Under beneficence, he introduces and
explains three rules: ‘Next I must do as I proposed and speak about beneficence
and liberality. Nothing is more suited to human nature than this, but there are
many caveats. First, one must see that kindness harms neither the people whom
one seems to be treating kindly, nor others; next, that one’s kindness does not
exceed one’s capabilities; and then, that what is bestowed on each person is
according to his worth. Indeed, that is fundamental to justice, to which all these
things ought to be referred’ (Off. I 42). The third rule actually echoes a formula
that is often cited as a definition of justice: to each according to desert or worth
(ἀξία). The connection of these rules with human nature is not made entirely
clear, but Panaetius’s exposition was no doubt much more thorough. However,
the references to human nature at the beginning of each chapter and occasionally
throughout the discussion show what was intended: Panaetius (or Cicero) was
offering these rules of justice and beneficence as interpretations of the very general
theses that human beings care for each other and want to live in a community, in
order to show what it means in practice to live in accordance with the basic natural
impulses of humans. If the interpretations are sometimes questionable and their
justification unconvincing, this just goes to show that the command to follow
    :     157

human nature in one’s dealings with others leaves ample room for interpretation,
and therefore also for deliberation.
This point is perhaps even more evident in book III, where Cicero discusses
cases of apparent conflict between what is honourable and what is useful or
advantageous. Since Panaetius never wrote this part of his work, Cicero had
looked at several books by other Stoics to fill the gap, which shows that there
was a lively debate among Stoics about such problem cases. One famous example
is a dilemma that may well go back to Carneades’ criticism of Stoic ethics:¹³
imagine two men after a shipwreck, trying to hold on to a plank that will support
only one. What should the wise man do? Pushing the other man off seems to go
clearly against justice, but if both cede the plank to the other, they will both perish,
as much as if they both hold on. Cicero mentions several responses from a book by
the Stoic Hecato (a pupil of Panaetius), who tried to suggest solutions based on the
relative merits of the two persons, ending with the proposal that something like
the lot should decide if both men were equally wise and important to their
communities (Off. III 89–90). In another case, Cicero reports that the eminent
Stoics Antipater of Tarsus and Diogenes of Babylon disagreed about the solution:
‘suppose that a good man had brought a large quantity of corn from Alexandria to
Rhodes at a time when corn was extremely expensive among the Rhodians
because of shortage and famine. If he also knew that several more merchants
had set sail from Alexandria, and had seen their boats on the way, laden with corn
and heading for Rhodes, would he tell the Rhodians? Or would he keep silent and
sell his corn at as high a price as possible?’ (Off. III 50–53). The question is clearly
about fair pricing, and Cicero may be too eager to agree with Antipater, who
appealed to ‘principles of human nature’ to support the view that the merchant
should ‘not conceal from men the advantages and resources that are available to
them’.¹⁴ One example that Cicero cites from Hecato’s book (Off. III 90) inciden-
tally shows that conflicts could also arise between different moral obligations:
should a son whose father is stealing from a temple or digging a tunnel to the
treasury denounce the father to the magistrates? The first answer is that this would
be impious, since citizens should revere their parents; but when the story is
changed to the father trying to impose a tyranny or betraying his country, the
answer is that the claims of the community should take precedence over filial
piety. Again, these cases demonstrate that deliberation about right and wrong may
often be necessary, and that even eminent philosophers could disagree about what
constitutes a reasonable justification.

¹³ It appeared in Cicero’s report of Carneades’ famous speech against justice (de rep. III 51,
Lactantius) and also in the Anonymous Commentary on Plato’s Theaetetus, probably written in the
first century .
¹⁴ On this debate, see J. Annas, ‘Cicero on Stoic Moral Philosophy and Private Property’, in
M. Griffin and J. Barnes (eds.), Philosophia Togata, Oxford 1989.
158    

One might think that the other kind of deliberation—looking for ways and
means to reach a goal, or planning—might find its place in the second book of De
Officiis that deals with obtaining the ‘comforts of life’. Panaetius, who was clearly
writing for upper-class gentlemen, limited his discussion in this field to the most
important safeguard of human life—the support of other people, setting aside
questions about health care and economics, which had obviously been treated as
belonging to the subject by other Stoics (Off. II 86). The topic of that book, then, is
how to win the hearts and minds of others (II 19), but still the relevant questions
are repeated by Cicero in the same form as before: what is useful or not useful,
with the additional question: which among useful things are more so than others,
which are the most useful (II 1). Ways of acting are evaluated as possible means to
the end of a safe and comfortable life, as in the public deliberations that Aristotle
considers in the Rhetoric; with the difference that Cicero always insists that
immoral means are ruled out—presumably, injustice can never be justified. So
the emphasis is, as before, on justification, not on planning. This is actually not
surprising in a book about appropriate action. The advice given to readers has to
be general, that is, concern types or ways of acting and their effectiveness or
legitimacy, not the particular situations emphasized by Aristotle, where all one can
do is to list the various aspects that have to be taken into consideration. The ability
of planning ahead and making the best decisions in particular circumstances is
mentioned only once—as a means of winning the trust of others. This passage is
worth quoting in full, since it presents an interesting parallel to Aristotle’s
discussion of practical wisdom in EN VI:

That people have trust in us can be achieved by two things—namely, if we are

considered to possess practical sense together with justice. For we trust those
whom we take to be more knowledgeable than ourselves; whom we believe to
have foresight and to be able, when matters have reached a critical point, to act
effectively and to make the decision that suits the circumstances. For this is what
men think of as truly useful practical sense.
Just people—that is, good men—are trusted insofar as there is no suspicion of
deceit or injustice about them. Hence, we think that these are the persons to
whom our safety, our fortunes, and our children may be most rightly entrusted.
Of these two, justice has more power to instil trust, given that it has sufficient
authority even without practical sense, while practical sense without justice has
no way of inspiring trust. For, the more shrewd and cunning a man is, the more
odious and suspect will he be once his reputation for integrity (probitas) is gone’.
(Off. II 33–34; my translation)

The Latin word I have translated by ‘practical sense’ is prudentia—the term

normally used to translate Greek φρόνησις. It is here obviously used in a morally
neutral sense, but Cicero hastens to add that he does not mean to suggest that the
    :     159

virtues could exist separately; he is following Panaetius’s example in speaking with

the vulgar (II 35).
In EN VI 12,1144a23–29, the same capacity to get things done efficiently is
described by Aristotle as δεινότης—cleverness or shrewdness. ‘There is a capacity
people call cleverness, and this is the ability to do what leads to the aim one has set
oneself, and to reach it. Now, if the aim is noble, the ability is praiseworthy, but if
the aim is bad, it is mere cunning. This is why we call even the practically wise
clever and cunning. Practical wisdom is not this capacity, but it does not come
without it.’
But whereas Aristotle sees such cleverness as indispensable for good deliber-
ation, Cicero mentions it merely as a helpful means of winning trust.
Looking back from the Stoics to Aristotle, one might say that once you start
thinking about a justification for your actions, it seems that a good education and
good laws may not be enough. Some of the difficulties that arise in questions of
fairness or justice are discussed in EN V, and the Rhetoric shows that Aristotle was
quite aware that the evaluation of means to political ends requires serious discus-
sion, but it seems that he saw the devising of successful plans or policies as the
main achievement of practical wisdom. The Stoics’ advice to ordinary people
about applying rules of justice, whether by appealing to nature or to some other
fundamental principles, was certainly not superfluous and had a long-lasting
influence on the later legal tradition. On the other hand, the Stoics apparently
paid little attention to the kind of skill needed for successful planning.
But the contrast between Aristotle and the Stoics is not just a matter of
emphasis—what underlies it is ultimately the difference in their respective theor-
ies of the human good and their very different conceptions of what counts as
success in action.
Aristotle accepted the traditional triad of goods: of the soul, of the body, and
‘external’ goods. He defined the human good as a life of rational activity according
to virtue, thus clearly ranking the good of the soul above the others, but he also
acknowledged that the best human life requires at least some of the other goods as
well—either as means or prerequisites of virtuous activity or because the absence
of health, beauty, or friends would mean that a life was not complete or perfect
and could still be improved. As Aristotle points out in the Rhetoric, virtuous action
is possible on many different levels. At the minimum, an action may count as
καλόν—‘noble’ would be too strong a word here—simply because not doing it, or
doing the opposite, would be shameful (Rhet. I 9,1367a6–8). But the most noble
and admirable actions are those that are difficult or risky and that also lead to
beneficial results for others—the more so, the greater the number of people
affected (cf. Rhet. I 9.1366b3–7). This is why Aristotle holds that the greatest
opportunities for virtuous action come with the role of ruler, and why he some-
times even describes practical wisdom as a virtue of rulers only. It is not that those
virtuous actions are noble only because of their results. What makes them noble
160    

is—for example—the selfless effort and willingness to take risks on behalf of

others; but clearly a community will want to have political leaders who have not
only noble aims, but also the ability to reach them. Good rulers are praised not
only for their admirable intentions, but also for their ability to put them into
practice.
Aristotle realized that the achievements of great statesmen are not based on
scientific or theoretical thought alone, and this probably led him to the distinction
between practical and theoretical reasoning, and between deliberation and explan-
ation or justification. Just as theoretical thinking should result in insight (νοῦς),
but then also in explanation and deductive proof, so good deliberation should
result not only in a plan, but also in reaching the goal one has set oneself, and that
means that ‘cleverness’, the ability for effective planning, is a necessary prerequis-
ite for practical wisdom.
The Stoics, on the other hand, declared all bodily and external advantages and
disadvantages as indifferent with respect to the human good and could therefore
not consider the results of virtuous actions as good in their own right. All that
counts as good is doing the right thing for the right reason, regardless of effort or
outcome. In fact, the Stoics held that there could be no degrees of goodness, just as
there are no degrees of rightness. This reflects, of course, the difference between
strictly moral judgement and a wider evaluation that also takes into account the
effort and achievement of the agent. It is true that the swimmer who tries and fails
to save the drowning child has acted just as nobly as the swimmer who succeeds in
saving the child. And sometimes a noble failure may be even more admirable than
a relatively easy success—so, to use a Greek example, in the case of the Spartan
soldiers at the Thermopylae. Yet it is odd to deny that the survival of a child is a
good thing, or to maintain that the dead Spartans at the Thermopylae were in the
end as happy as the surviving Athenians after the victory at Marathon. However,
the very word κατόρθωμα, which means both right action and success,¹⁵ indicates
that success, for the Stoics, meant only doing the right thing. So the skill of
effective planning, while possibly helpful in everyday life, was not seen as making
a difference to the virtue of an agent. The Stoics did not take over Aristotle’s
distinction between practical and theoretical reasoning. Practical wisdom
(φρόνησις) was for them a branch of wisdom in general, distinguished from
other branches of knowledge only by its subject matter. This is understandable,
given that in their view, both theoretical reasoning and practical deliberation were
aimed at establishing the truth.

¹⁵ LSJ list ‘right action’ as a special sense of the word, which in fact usually means success. Literally,
of course, it means ‘getting things straight’, and it seems obvious that the Stoics exploited the etymology
of the word in making it a technical term.
    :     161

They would probably not have denied that strategic planning requires deliber-
ation,¹⁶ but as far as we can see, the kind of deliberation that they discussed as
philosophers had to do with justification, not with strategy.
This reminds us of the fact that Aristotle and the Stoics discussed deliberation
in the context of ethics. Each theory is focused on the kind of deliberation most
relevant for their conception of the human good. For Aristotle, the distinction
between two different kinds of reasoning supports the claim that there are two
distinct intellectual virtues. For the Stoics, only questions of right and wrong were
relevant to the success of any kind of action. And so the two theories reflect two
very different conceptions of human happiness.

¹⁶ In fact, the first book of the De Officiis even contains some advice about career planning—in the
particular case of a young Roman embarking on a political career, so probably due to Cicero, not
Panaetius.
 12
Academics Fighting Academics

The arguments of the former Academic Sceptic Antiochus against the

‘Roman books’ of his teacher Philo of Larissa, reported in Cicero’s Lucullus,
indicate that his understanding of Stoic epistemology was influenced by his
Academic background, and in particular by the arguments of Carneades. These
arguments also allow a plausible guess at the new position Philo defended in the
Roman books.

When I read Cicero’s Lucullus for the first time many years ago, I read it as a
straightforward account of the epistemological controversy between the Stoics and
the Academic sceptics. That is, after all, what Cicero promises, although he makes
it clear that his own information comes from a late stage in the debate, when the
positions of both schools had presumably evolved and shifted in response to
objections from the other side. But Cicero explicitly distinguishes between
several different positions within the sceptical Academy, and in the Lucullus
he sets aside the most recent among them, that of Philo’s ‘Roman books’, in
order to concentrate on the arguments of Arcesilaus and Carneades (Luc. 12).
Antiochus, on the other hand, is introduced as an orthodox representative of
the Stoic view, which he apparently adopted lock, stock, and barrel as a ‘correction’
of the doctrines of the early Academy (Ac. 1.35). The somewhat petty dispute
about Academic pedigrees reported in the Varro looked like a less important
side issue that could be safely ignored if one was mainly interested in the philo-
sophical arguments.
Over the last twenty years, however—thanks above all to the books by
J. Glucker¹ and H. Tarrant²—the figures of Antiochus of Ascalon and Philo of
Larissa began to take on more definite features, and I also realized that the
doctrine of Philo’s Roman books—pace Cicero—was of great interest not just
historically (which has been doubted), but also from a twentieth-century point of

¹ ‘Antiochus and The Late Academy’, Hypomnemata 56, Göttingen 1978.

² Scepticism or Platonism?, Cambridge University Press 1985.

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
?DOI: 10.1093/oso/9780198868385.003.0012
    163

view. It appears that Philo was the first to formulate a modest conception of
knowledge that can be compared to contemporary versions of fallibilism, drop-
ping the requirement of certainty that has bedevilled so many epistemological
debates both before and after his time. It is a great pity that the Roman books did
not survive, not even in the summary version of the report that Cicero apparently
gave in the lost Catulus.
We cannot hope to recover Philo’s own arguments, but I think that the
indications offered by Cicero about the main epistemological issue in the dispute
with Antiochus, and about Philo’s previous position, allow us at least to make an
educated guess at the reasoning behind Philo’s revolutionary move.
One factor that has so far received less attention than it deserves seems to me to
be the fact that the Sosus affair was a dispute among Academics—not just with
respect to the tradition of their school, but also, I think, in terms of the epistemo-
logical arguments of both sides. Given Cicero’s emphasis on the novelty of the
views expressed in Philo’s Roman books, all commentators have, of course, been
careful to distinguish Philo’s Roman views from earlier versions of Academic
scepticism. But most have also followed Cicero in taking Antiochus’s version of
Stoic epistemology at face value—and this may actually be rather imprudent.
Antiochus never became a member of the Stoa, and while he probably did get
first-hand information from his Stoic contemporaries in Athens (Mnesarchus and
Dardanus, according to Cicero, Luc. 69), it is likely that his understanding of
Stoicism was strongly influenced by his education as a member of the sceptical
Academy. We may therefore expect him to rely on an Academic interpretation of
Stoic doctrine, whether he realized it or not, and we should not assume without
further ado that this was exactly the interpretation the Stoics themselves would
have wished to defend.
Having raised some suspicions about Antiochus’s Stoic credentials, let me begin
by listing the information Cicero provides in the Lucullus about the main point at
issue between Philo and Antiochus.

(1) In Luc. 18, the speaker for Antiochus briefly refers to Philo’s innovations.
He mentions two objections: first, Philo is said to tell ‘manifest lies’, for which he
had been taken to task by Catulus the father (that is, in the lost first book); second,
Antiochus showed that he (Philo) found himself in exactly the position that he
feared: . . . et aperte mentitur ut est reprehensus a patre Catulo, et ut docuit
Antiochus in id ipsum se induit quod timebat. The ‘manifest lie’ had been men-
tioned before (12): it consisted in denying that the Academics had ever held the
view that was defended by Catulus the day before, namely (as we can safely
assume) the claim that nothing can be known (ἀκαταληψία).³ The second point

³ A note on translation: in this chapter, I translate the Greek verb καταλαμβάνειν (literally ‘to grasp’)
and its derivatives (κατάληψις, καταληπτόν) as ‘to know’, ‘knowledge’ and so on, except for the
164    

is explained in the following lines. According to Antiochus, when Philo attacked

Zeno’s definition of the cognitive impression (καταληπτικὴ φαντασία), he did
away with the criterion of what is known and unknown, from which it followed
that nothing can be known—the position that Philo was trying to avoid. So, Philo
rejected the Zenonian definition of the cognitive impression, and he did not wish
to accept the thesis, argued by his Academic predecessors, that nothing can be
known. The connection between these two points is brought out by Sextus
Empiricus: Philo said that as far as the Stoic criterion is concerned, nothing can
be known; but as far as the nature of things themselves is concerned, they can
indeed be known⁴ (PH I 235 ὅσον μὲν ἐπὶ τᾦ Στωικῷ κριτηρίῳ τουτέστι τῇ
καταληπτικῇ φαντασίᾳ, ἀκατάληπτα εἶναι τὰ πράγματα, ὄσον δὲ ἐπὶ τῃ φύσει τῶν
πραγμάτων, καταληπτά)
(2) Philo’s new doctrine was one that had not been held by any Academic
before, including Philo himself at an earlier stage. Philo, of course, maintained that
everybody from Plato on had held his view, but Antiochus’s claim to the contrary
was apparently confirmed by Heraclitus of Tyre, another Academic said to have
been present at the debate in Alexandria (Luc. 11).
(3) Before the Roman books, Philo held a position of mitigated scepticism or
probabilism shared by other Academics, notably Metrodorus (Luc. 78).⁵ These

adjective καταληπτική, which is translated as ‘cognitive’. It seems to me that this is the best way of
making it clear what these philosophers were arguing about. I am, of course, aware of the fact that the
Stoics reserved the ordinary Greek verb for ‘knowing’, ἐπίστασθαι, for the intellectual state of the expert
or the sage who has a systematic understanding of a given field, while using καταλαμβάνειν and its
cognates for the act of coming to know something or for an individual piece of knowledge. But English
does not have a convenient pair of terms such as the German ‘Erkenntnis’ and ‘Wissen’ to mark this
distinction, and the term of art ‘cognition’, used in English translations of Kant, has not become part of
everyday philosophical language. The English equivalent of ‘Erkenntnistheorie’ is ‘theory of know-
ledge’, not ‘theory of cognition’.
⁴ This interpretation was first proposed by R. Hirzel (Untersuchungen zu Cicero’s philosophischen
Schriften, III, Leipzig 1883, 196–200) and accepted by most later scholars, most recently, for example,
Barnes (‘Antiochus of Ascalon’, in J. Barnes and M. Griffin, Philosophia Togata, Oxford 1989, 51–96);
Lévy (C. Lévy, Cicero Academicus, Rome 1992); and Bächli and Graeser (notes in C. Schäublin,
A. Graeser, and A. Bächli, Marcus Tullius Cicero, Akademische Abhandlungen: Lucullus, Hamburg
1995). I see no reason to read the second clause of this statement as the strong claim that the nature of
things can be known (so R.J. Hankinson 1995, 119). The thought is, I believe, simply that if there is an
impediment to knowledge, it lies in the Stoic conception, not in the nature of things themselves. This
does amount to the claim that we may have objective knowledge of things (because they are not of such
a nature as to be unknowable), but not that we can come to know their nature. A different interpret-
ation has been suggested by David Sedley (‘The End of the Academy’, Phronesis XXVI, 1981, 72): Philo
said that the (sensible) world is unknowable to us because of the lack of an infallible criterion such
as the Stoics postulate, but that the truth about the sensible world is accessible to god. It seems to me
that this cannot be read into Sextus’s statement. The phrase ὄσον ἐπί indicates a limitation of the claim
that things cannot be known to those who accept the Stoic criterion, not to humans in general.
⁵ I take it that when Cicero says in the dedicatory epistle to Varro (Ep. ad familiares 9.8) that he
himself is playing the part of Philo, he is referring to the position Philo held before the innovations of
the Roman books. This would agree with the stance Cicero adopts explicitly in his last speech in the
Luc. (112–113): he endorses the Stoic definition of the cognitive impression as well as the claim that
nothing can be known, and concludes that he (unlike the wise man) will hold opinions. The
disagreement with Philo and Metrodorus about Carneades mentioned at Luc. 78 seems to concern
    165

philosophers thought that the Stoic definition of the cognitive impression was
correct, but that its conditions could not be met. So, they endorsed the thesis that
nothing can be known; but they also maintained—following, as they saw it, the
great Carneades—that it was legitimate, even for the wise man, to hold probable or
plausible opinions, both about everyday matters and concerning philosophical
investigations and arguments.
Given Philo’s own claim of continuity within the Academic tradition, it is not
likely that he came up with a radically new epistemology.⁶ The most economical
assumption about his ‘revolutionary’ claims would seem to be that he decided that
he himself and his fellow Academics had been too modest and overly impressed
by the Stoic doctrine. The thesis that nothing can be known was a consequence
only of the excessively high standards set by the Stoics. Once one sees this point,
one can claim not just reasonable opinion, but knowledge, though not the
infallibility that the Stoics demanded for their wise man. What Philo rejected,
then, was the requirement of certainty contained in the last clause of the Stoic
definition of cognitive impression: an impression that provides knowledge must
be such that it could not arise from what is not so.
If this hypothesis about Philo’s innovations is correct, then we have at least one
set of arguments directed against Philo’s position insofar as it was similar to the
view of those Academics who held that the Stoic criterion was not needed for the
conduct of life or for philosophical argument, since one could rely on probable
impressions. These people were offering a substitute for Stoic cognitive impres-
sions, and the difference between Philo’s earlier and later epistemological position
would have consisted simply in the claim that the substitute he offered was good
enough to justify claims to knowledge. In Luc. 32–36, Lucullus/Antiochus argues
that the alleged substitute will not do. These arguments must be either Antiochus’
own or those of contemporary Stoics, since they attack the views of Carneades’
successors. The arguments are not very impressive, but they reveal, I think, what
Antiochus took the Stoic position to be.
As far as I can see, Antiochus offers two⁷ main arguments:

(a) those who claim that true and false impressions cannot be distinguished
cannot even claim to have the notions of truth and falsehood (Luc. 32–33). Let us
disregard the polemical formulation of the Academic thesis: they did not hold that

only the point that even the sage will hold opinions. But Cicero also tells us in the Varro (Ac. 1.13) that
he heard Philo expound the doctrines of the Roman books, and I think one can see some traces of their
influence in this speech, esp. at 146 (see below n. 17).
⁶ Such as the doctrine ascribed to him by Tarrant, who introduces two new elements that are found
nowhere in Cicero—nor for that matter in other sources, I think. See my review of H. Tarrant,
Scepticism or Platonism? Cambridge 1985, in Ancient Philosophy 11, 1991, 202–6.
⁷ I skip the argument that the wise man will need to trust his own convictions and hence must have a
reliable ‘mark of truth’, briefly alluded to in 36. This point seems to me to belong to the moral side of
the dispute; see Luc. 23–24.
166    

true and false impressions could not be distinguished at all, but only that no true
impression was such that an exactly similar one could not, in principle, be false.
Antiochus may have thought that this was tantamount to the statement that there
was no difference at all. But there is no reason to agree with him that one cannot
form a concept unless one is able to decide with certainty whether it has been
correctly applied. Faced with the question whether Antiochus is a man or a
warship, anyone would no doubt assert that he was a man, and that it is false to
say that he was a warship. Unless we want to go for rampant self-contradiction, we
will take what we believe to be true and its contradictory to be false. But this does
not imply that we have a general method for deciding what is true and what is false
(hence the frequent disagreements among ordinary people as well as
philosophers).
Cicero’s attitude to this argument is a little ambiguous (Luc. 111). On the one
hand he calls it a particularly admirable point by which—according to
Antiochus(!)—Philo was greatly disturbed. On the other hand, he rejects it with
a somewhat perfunctory remark: ‘we recognize both truths and falsehoods, but
this only goes as far as approving⁸—we do not have a sign of knowledge.’ Perhaps
he was sufficiently impressed by Antiochus’s boasting that he did not see exactly
how the argument could be refuted; but his brief remark seems to show that other
Academics were not greatly disturbed.
(b) Antiochus’ second argument is based on the assumption that in order to
distinguish between true and false impressions, one must have a mark (nota) or
sign (signum) of truth.⁹ There seem to be two versions of such an argument in Luc.
32–36, but the first may be intended to be the same as the second. The first version

⁸ The text is: nam tam vera quam falsa cernimus. sed probandi species est, percipiendi signum nullum
habemus. The phrase probandi species est is difficult, and my rendering is tentative. Reid (followed by
Rackham) claims that species is a translation of φαντασία—as it may, but need not be (see n. 321, 276 in
Schäublin, Bächli and Graeser 1995). The sentence would then say, according to Reid’s paraphrase, that
an impression may lead to approval, but will not warrant a knowledge claim. But how would this be a
reply to the objection? Cicero should say something to the effect that one can make the distinction
without having an infallible criterion. Given that species may also be used to translate εἶδος (see e.g. Luc.
58, discussed below), I would suggest that this sentence should be closely connected with the preceding
one (reading a comma after cernimus), and understood as saying ‘this (i.e., our recognition of truths
and falsehoods) is only a kind of approval, (since) we do not have a sign for knowledge’. I do not know
what sense to make of Hankinson’s literal translation (R.J. Hankinson, The Sceptics, London/New York
1995, 117) ‘the impression is of approval’, but it underlines the fact that the usual rendering is not very
close to the Latin.
⁹ Plasberg lists both nota and signum as translations of the Greek σημεῖον in his index, and I think he
is right. It is tempting to suppose that nota must be a translation of ἰδίωμα or χαρακτήρ, words used by
Sextux Empiricus in his report on the epistemology of the later Stoics for the alleged peculiar feature
that characterizes cognitive impressions (so e.g. Schäublin, Bächli and Graeser 1995, n. 92, 216), while
only signum translates σημεῖον. But a look at the relevant passages shows, I think, that this will not do.
In several passages the nota must be the Stoic criterion, that is, the cognitive impression itself (e.g. 84:
eius modi notam quae falsa esse non possit, and non ea nota iudicabis qua dicis oportere ut non possit
esse eiusdem modi falsa), and Cicero seems to switch freely between nota and signum in 33–36. In Ac.
1.32, nota is combined with argumentum just as is signum in Luc. 36. There is only one passage where it
seems better to take nota to refer to a feature of the impression rather than the impression itself, Luc. 58
    167

appears in Luc. 33–34: if a cognitive impression has its features in common with a
false one, there will be no criterion, because a peculiar fact cannot be indicated by
a common sign (in eo autem si erit communitas cum falso, nullum erit iudicium,
quia proprium in communi signo notari non potest). What this seems to say is that
if anything is a sign both of x and y, then it cannot be used as a sign of either
separately. For example, if fever is a sign of both pneumonia and meningitis, it
cannot be used to diagnose either of these diseases. The point about ‘common
signs’ is, of course, correct, but it does not seem to apply to the situation it is
supposed to illustrate. An impression that p, even if it is compatible with the truth
of not-p, will still be an indication (if at all) of p only, not of not-p. So perhaps
Antiochus is using the terminology known from Philodemus (De Signis I 1–17),
according to which a common sign is one that may be present even if the thing
signified is not. This is the point made at Luc. 36: ‘what could be more absurd than
their way of speaking: “this is indeed a sign or argument of such-and-such a thing,
and that is why I follow it; yet it is possible that what is indicated be either false or
nothing at all”.’ (quid autem tam absurde dici potest quam cum ita locuntur: ‘est
hoc quidem illius rei signum aut argumentum, et ea re id sequor, sed fieri potest ut
id quod significatur aut falsum sit aut nihil sit omnino’.) But what ‘they’ say is not
absurd at all. In the absence of any evidence for not-p, I may take an impression
that p as evidence for p, and act accordingly, even though it is (logically) possible
that p is false.¹⁰
So much for Antiochus’s arguments against the probable impression as a
substitute for the Stoic criterion. What is most remarkable about the last argument
is that Antiochus describes the Stoic criterion as a ‘mark’ or ‘sign’ of truth. I very
much doubt that any real Stoic would have used such language. For cognitive
impressions were supposed to provide precisely those basic truths that are not
established by inference—the evident starting points of proofs, from which one
could then infer further non-evident truths. Cicero’s concise rendering of the
argument shows, I think, that the choice of terms here is no accident: ‘peculiar
feature’ (proprium, Greek ἴδιον) and ‘common sign’ (commune signum, κοινὸν
σημεῖον) were technical terms in the debate about sign-inferences reported, for
example, by Philodemus and Sextus Empiricus. Whatever notion of a sign is
involved here—and there is no further explanation in Cicero’s text—its use
implies at least that the ‘sign’ can be observed independently of the thing signified,
and this is a point that no Stoic worth his salt should have admitted about
cognitive impressions. On the other hand, the conception of impressions as
signs or evidence for underlying facts fits very well with the role ascribed to

(discussed below): quasi vero non specie visa iudicentur; quae fidem nullam habebunt sublata veri et falsi
nota. But there the special character of cognitive impressions seems to be indicated by the word species
rather than nota.
¹⁰ For a brief discussion of the argument, see Barnes (n. 2 above), 84f.
168    

impressions in the Carneadean ‘theory of criteria’ in Sextus’s report

(M 7.159–189).¹¹ Although the technical language of sign-inference is absent,
impressions are clearly treated as evidence—for example, they are compared to
messengers (163) or witnesses (184) who may or may not be reliable. In order to
be infallible guides to the truth, cognitive impressions, on this model, would have
to be conclusive evidence, and if this is what Antiochus took them to be, then he
could quite correctly describe them as proper signs of the facts they represent.
But this conception does not fit the—presumably Stoic¹²—comparison of sense
impressions with light, according to which sense impressions show us the objects
of perception in the way in which light reveals both itself and the objects of sight.
I do not infer that there is a table in front of me from the fact that I can see one in
broad daylight. The presence of daylight may make me confident that I can trust
my eyes, but it is not a sign of the presence of tables.
Nonetheless, the light-analogy appears side by side with Carneades’ theory in a
revealing passage in Sextus (M 7.163–164):

Just as light shows itself and all that is in it, so the impression, being the starting
point of knowledge in the living creature, must, like light, reveal itself and also be
indicative of the evident thing that produced it. But since it does not always
indicate the truth, but often deceives and disagrees with the things that sent it
forth, the way bad messengers do, it followed necessarily that one could not accept
every impression as a criterion of truth, but only, if any, the true one. Again, since
there is no true impression of such a sort that it could not be false, . . . the criterion
will be found in an impression that is common to the true and the false.

By switching from the light-analogy to the messenger-analogy, this passage puts a

distance between the impression and the object or fact that caused it. What the
observer receives is the impression, which may or may not correctly represent its
object. While the light-analogy seemed to stress the point that the impression and
its object are grasped simultaneously, the messenger-analogy suggests that the
impression offers only a report and does not provide direct access to the object.
The description of the criterion as a sign obviously fits the second analogy, but not
the first.
There is, then, some reason to think that Antiochus’s interpretation of Stoic
epistemology was Academic rather than genuinely Stoic. Yet, since Antiochus
himself professed his acceptance of the Stoic doctrine of cognitive impressions, we

¹¹ It is not necessary for my argument to assume that Antiochus himself is the source of Sextus’s
report, as has been argued by several scholars (first by Hirzel 1883, III 493–524; more recently, Tarrant
1985, 89–114; Sedley (‘Sextus Empiricus and the Atomist criteria of truth’, in Elenchos 13, 1992, 19–56;
contra, see Barnes 1989, 65), but this would, of course, offer a convenient explanation of the similarities
between Luc. 33–34 and M 7.163–164.
¹² Ps.—Galen ascribes it to Chrysippus, cf. SVF II 54; Diels, DG 401.11ff.
    169

must presume that he took himself to be advocating the Stoic view. In order to see
how Antiochus or his Academic predecessors arrived at this interpretation of Stoic
doctrine, we should go back, I think, to the beginnings of the debate with Zeno and
Arcesilaus. The story of this debate has been told in greater detail elsewhere, so
I will mention only what I take to be directly related to Antiochus’s interpretation.

II

In an anecdote that is no doubt too good to be true, Cicero tells us (Luc. 77) that
the debate began when Arcesilaus, impressed by Zeno’s novel claim that the wise
man would hold no mere opinions, inquired of Zeno how this would be possible if
the wise man could not attain any knowledge. Zeno confidently replied that the
sage would not need to have any opinions because he could indeed attain
knowledge: namely of things grasped by cognitive impressions. He then produced
the following definition of the cognitive impression: ‘an impression that arises
from what is, imprinted and sealed and rendered exactly as the thing is’. Arcesilaus
next asked whether this would still be the case if the true impression were such
that it might also be false. Zeno saw the point of the question and added another
clause to his definition: ‘such as could not arise from what is not that existing
thing.’ From then on, the Academics produced a series of arguments to show that
there cannot be a true impression of such a sort that there could not be a false one
of exactly the same sort.
This story puts the long debate in a nutshell, which makes it unlikely to be
historical. But apart from the issue whether one could live without assent, which
Cicero rightly sets aside in Luc. 78, the epistemological discussion between the two
schools was indeed centred around the third clause of the Stoic definition. Now,
even if it is true, as the story suggests, that this clause was added to forestall
Academic objections (so also SE, M 7. 252), it is likely that Zeno or his successors
would have thought of the addition only as a clarification, not as a new and
logically independent requirement. As the subsequent debate shows, the third
clause can be taken in either of these ways, which I will call, for brevity’s sake, the
weaker and the stronger interpretation.¹³
On the weaker interpretation, the third clause makes explicit what is already
implied by the second, namely that the cognitive impression will represent its
object with such precision and accuracy that it could not come from any other
thing. In other words, Zeno assumed that a sufficiently precise and accurate
impression of any given object would capture all its peculiar features in such a

¹³ M. Frede (‘Stoics and Skeptics on Clear and Distinct Impressions’, 1983; repr. in Frede, Essays in
Ancient Philosophy, Minneapolis 1987, 151–76) notes the ambiguity (p. 165), but then settles for the
stronger interpretation.
170    

way that it was necessarily different from any impression that might have come
from a different object.
On the strong interpretation, however, the third clause stipulates that a cogni-
tive impression must be different in kind from any non-cognitive one, so that the
reason it could not arise from any other thing would lie in the fact that it is not the
kind of impression that could misrepresent the object that caused it. Hence,
cognitive impressions would have to be of a type that distinguishes them from
all others, regardless of their content.
If one looks at the Stoics’ replies to the various kinds of counterexamples and
objections put forward by their Academic opponents, one can see, I think, that
they tried to rely on the weaker interpretation, but that they were pushed by the
Academic objections to adopt the stronger one. It is this stronger interpretation
that eventually led to the version we find in Antiochus.
The examples introduced by the Academics to support their notorious thesis of
ἀπαραλλαξία—the claim that for any true impression one can describe conditions
such that a false one would be exactly alike and hence indistinguishable—were
roughly speaking of three different kinds:

(a) Kinds of similar objects: eggs, coins, identical twins, impressions from the
same seal (cf. Luc. 54, 56–57, 84–86). The Academics suggested that these objects
were so similar to one another that it would be impossible to distinguish them by
any perceptible features. The Stoics replied by appealing to the metaphysical
principle that individuals of any given kind will necessarily have peculiar features
that distinguish them from all other individuals of the same kind (Luc. 50, 85).
They supported this claim by examples of experts like the famous Delian poultry
farmer who was allegedly able to tell which egg came from which hen. If one
accepts the general principle, then this would go to show that any two individuals,
however similar, must in principle be distinguishable, though it may take consid-
erable experience and skill in some cases to make the correct distinctions. The
Stoic sage, who would not have to be a poultry farmer, would presumably refrain
from making assertions about the identity of individual eggs or coins, considering
that he lacks the requisite expertise (cf. Luc. 57). So, error can be avoided either by
exercising due caution or by acquiring the relevant skill.
(b) Dreamers and madmen (Luc. 48, 51–53, 88–90). When dreaming or in a
state of mental derangement, people may have less than accurate or even wildly
false impressions, but they will not be able to realize that this is the case, and hence
be unable to distinguish cognitive from non-cognitive impressions. In fact, as the
Academics urged, such people may be quite confident that their sense impressions
are accurate; so, it would seem that error could not be avoided in this kind of
situation. Here the Stoics replied, first, that the impressions of madmen and
dreamers are, in fact, unlike clear impressions received in the waking state, as
witness the hesitation and bewilderment expressed sometimes by people on the
    171

verge of derangement, as well as the readiness with which one rejects a dream
impression once one wakes up (Luc. 51–52). Second, a state of lethargy, drunk-
enness, or madness may indeed be such that the normal capacity to exercise due
caution is temporarily lost. Hence it might be that even the sage would not be able
to avoid all error in such a state.
The question of the sage’s powers of judgement in abnormal states was, I think,
what triggered the dispute over the question whether virtue can or cannot be lost
(see the testimonia in SVF III, 237–244, 56ff). It is probably significant that
Chrysippus, unlike Cleanthes, is said to have held that virtue may be temporarily
lost in states of disease or madness.¹⁴ These counterexamples, then, can be avoided
by restricting the claim of distinguishability, as seems quite reasonable, to the case
of healthy and sober perceivers.
So far, I think, the Stoic defences rely on the weak interpretation. The cognitive
character of an impression is supposed to be due to its precision and accuracy, and
in situations where one knows that such accuracy is hard to come by, as with eggs
or coins, one will avoid error by refraining from judgement. The appeal to the
principle of peculiar features of individuals makes it clear that the distinctness of
an impression is assumed to be due to the distinctness of its object, not to any
special feature characterizing cognitive impressions as such. The examples of
madness and similar states can be set aside if one is willing to grant that these
are passing episodes, recognizable as abnormal once one has returned to the
normal state. The argument here would be, then, that impressions received in a
deranged state of mind are not accurate enough to be trusted. The Academic
objection merely shows that one will make mistakes while in such a state, not that
one cannot avoid error while sane and sober. It appears that the more general
question as to whether and how one can tell at any given time that one is not
dreaming or deluded did not play an important role in the dispute.¹⁵ This may
seem surprising to a modern reader, but it will appear less strange once one
realizes that the relevant point was covered by the third and, I think, most
damaging kind of examples.
(c) Tricks and deception. Two early instances of this type of argument show up
in the biographies of Zeno’s pupils Aristo and Sphaerus (DL 7.162 and 7.177
respectively), both presented as designed to refute the thesis that the wise man will
not have opinions. In the first story, Persaeus—himself a Stoic—‘refutes’ Aristo by
sending one of two twin brothers to deposit a sum of money with Aristo, and

¹⁴ DL 7.127: καὶ μὴν τὴν ἀρετὴν Χρύσιππος μὲν ἀποβλητὴν, Κλεάνθης δὲ ἀναπόβλητον· ὁ μὲν
ἀποβλητὴν διὰ μέθην καὶ μελαγχολίαν, ὁ δὲ ἀναπόβλητον διὰ βεβαίους καταλήψεις. Cp. Simplicius, In
Cat. 402,22–26: according to the Stoics, in states of derangement virtue is lost together with the entire
‘rational disposition’, which does not mean that vice takes its place, but that its ‘firmness’ is relaxed.
¹⁵ See Luc. 54, where Lucullus/Antiochus points out that if there were no difference between the
impressions of sane and insane people, one could not even be sure of one’s own sanity. He treats this as
an absurd consequence of the Academic position, not as a point raised by the Academics themselves.
172    

having the deposit collected by the other twin. It may be, as A.M. Ioppolo argues
(in Opinione e scienza, Naples 1986, 82f.), that the ‘refutation’, coming from a
Stoic, was aimed only at Aristo’s claim to be a sage himself; but it still raises the
question how a real sage would have avoided the mistake. The same goes for the
anecdote about Sphaerus. King Ptolemy Philopator is said to have trapped him
with a practical joke by ordering wax pomegranates to be served at the dinner
table. Sphaerus was apparently deceived into thinking they were real pomegran-
ates, and the king triumphantly pointed out that he had assented to a false
impression. Sphaerus then replied that he had assented only to the impression
that it was reasonable to think these were real pomegranates. Now, if this is
supposed to represent the reaction of a wise man, then it begins to look like a
retreat to the position advocated by Arcesilaus, namely that in the absence of (any)
cognitive impressions, the sceptic will be guided by ‘the reasonable’ (SE, M 7.158).
If it is not what the sage would say, the question arises once again how he would
have avoided error in such a situation.
What distinguishes these two examples from the kinds of cases mentioned
before is that—as far as one can tell—the observer is sane, sober, in full possession
of his perceptual faculties, and well placed to get a clear and accurate impression of
the object. He has no reason to think he might be hallucinating. But he also has no
reason to exercise special caution, as in the case of eggs or coins, because he does
not know, and has no reason to suspect, that he is dealing with objects of a sort
that might come in several very similar species or exemplars. In order to defend
the claim that the wise man will nevertheless not be taken in even in such
situations, the Stoics might say, as Sphaerus’s reply suggests, that the sage will
always exercise caution, never taking a sense impression to be true unless he has
thoroughly investigated the object, but being guided by reasonable assumptions
when a full investigation is not practicable. (After all, it would have been very
impolite of Sphaerus to examine the fruit at the king’s table to make sure it was
real.) Aristo should perhaps have accepted a deposit only from a person he knew
well enough to be certain that he had no identical twin. Alternatively, the Stoics
might have claimed that the sage has extraordinary powers of perception such that
he will immediately recognize a fake pomegranate or a previously unknown twin
brother because his impressions (and his memory) are so precise and accurate that
he will notice the slightest deviation. This is not a very plausible move, however,
given that the Stoics themselves seem to have pointed out that precise perceptual
discrimination between similar objects requires special training or experience, as
with the Delian farmer or the mother who can tell her twins apart (Luc. 57).
Caution would seem more appealing, and one might also argue that practical jokes
are not all that ubiquitous—in most cases of unimpeded perception, what it would
be reasonable to assume will really be so. This is not enough, however, to
guarantee infallibility. What is disconcerting about examples of this sort is that
even though cases of trickery or accidental resemblance may be rare, they cannot
    173

be predicted. Since Arcesilaus is already said to have argued for the general thesis
of ἀπαραλλαξία (SE, M 7.154), I would assume that he based his generalization on
this kind of case. Now, caution will definitely have to lead to complete suspension
of assent if there is reason to think that deceit might be ubiquitous—and that is
exactly what was suggested, presumably by Carneades, in the notorious sorites-
argument about the deceiver god (Luc. 47 and 49). The Stoics apparently
responded to the suggestion that such a god could produce a false impression
indistinguishable from a cognitive one just as Descartes did many centuries later,
by saying that a god would not do such a thing even if he could. They may also
have thought that not even a god could bring it about that there was no difference
at all between two distinct individuals, so that impressions produced by different
objects would necessarily have to be distinct (Luc. 50). But this is not enough to
solve the problem. The Academics might even grant the assumption that any two
distinct objects, however similar, must produce different impressions (Luc. 85);
for once it had been shown that any true impression could conceivably be very
similar to a false one, the Stoic sage would have to exercise caution in every single
case. That is to say, he would either have to refrain from assent in every case or
acquire an incredible amount of technical skill (Luc. 86).
It is at this point, I think, that the Stoics would have been tempted to resort to
the stronger interpretation of the clause ‘such as could not arise from what is not
that existing thing’, by postulating a peculiar feature that distinguishes cognitive
impressions as a class from non-cognitive ones. Antiochus, for example (Luc. 46,
51), maintains that all ‘empty’ impressions lack the perspicuousness that charac-
terizes cognitive impressions. This may be plausible for dreams and hallucin-
ations, but what, for example, would have prevented Aristo from getting a clear
and accurate impression of the twin brother who came to collect the deposit? At
Luc. 58, Antiochus rejects an argument ‘occasionally’ (interdum) used by the
Academics, according to which they did not wish to maintain that there was no
difference between the impressions themselves, but only that there was no differ-
ence between the ‘species et quasdam formas eorum’¹⁶. The expression is difficult
to understand, since it sounds like ‘appearance of an appearance’, but the com-
bination species et forma may indicate that Cicero is rendering the Greek εἶδος (for
a similar use of species, compare perhaps Luc. 99 specie probabile and 111 probandi
species). The Academic argument would then be that the impressions themselves
might be granted (for the sake of argument: 85) to be distinct, given that they arise
from different objects, but since there is no difference in kind between cognitive
and non-cognitive impressions, they may still not be distinguishable (cf. Luc. 40).
Lucullus/Antiochus responds by insisting, indignantly, that impressions have to

¹⁶ See also Luc. 40 and 84, where the same argument seems to be mentioned, but without the words
species et forma.
174    

be judged by their species, and that this implies the availability of a ‘mark of truth
and falsehood’.
But what exactly might be the feature or set of features that characterizes
cognitive impressions as a kind? Sextus, in one passage (M 7.252), says that
according to the Stoics, cognitive impressions differ from all others in the way
in which horned snakes differ from all other kinds of snakes. This is, I think, an
unfortunate move.¹⁷ Not only do we never find a more specific account of the
alleged special feature of cognitive impressions, but the suggestion that there is
such a feature also invites a regress argument: if we recognize cognitive impres-
sions by their special mark, then how do we recognize the mark? By another
cognitive impression? And how do we recognize that? (cf. M 7.427–429). On the
other hand, the snake story has an attractive aspect that makes it similar to the
light-analogy discussed above: horns are an intrinsic feature of horned snakes, and
one obviously does not need an inference to get from seeing a snake that has horns
to the belief that one is seeing a horned snake. But the truth of an impression must
be a relational property—as pointed out, once again, by Carneades, SE, M 7.168—
and hence any intrinsic feature could at best be necessarily linked to truth.
Since Sextus’s remark about the distinguishing feature has no parallels else-
where, it is more likely that most Stoics simply insisted that cognitive impressions
were recognizably more clear, accurate, and vivid than any others. This would
mean that the distinguishing characteristic of a cognitive impression was not some
additional feature, like a blue ribbon or perhaps a watermark, discernible only
with due training. Rather, the distinguishing feature or features would be ones that
are present in all presentations to a certain degree, such as clarity and precision.
There would be indubitable cases at both ends of the range—impressions that are
overwhelmingly clear or very murky—but there would also be an intermediary
range in which it was arbitrary or impossible to decide whether a given impression
was to count as a cognitive or not. The suggestion that the distinguishing feature
comes in degrees would seem to be confirmed by Chrysippus’s notorious advice
for responding to a sorites-argument (Luc. 93). Finally, Sextus discusses two
possible sets of ‘characteristics’ (ἰδίωμα) that are obviously a matter of degree:
clarity and force (ἐναργές, ἔντονον) or accuracy and precision (κατὰ χαρακτῆρα καὶ
κατὰ τύπον, M 7.408). Needless to say, the Academics reacted by arguing that true
and false impressions may have all of these characteristics to the same degree.
The suggestion of the special character of cognitive impressions as a kind,
however, invites the sort of account that we find in the Lucullus: it seems that
we can separate the content of an impression from its characteristic features as an
impression of a certain kind, and maintain that it is the special features that

¹⁷ The comparison with different kinds of snakes may have been inspired by the example, presented
as a difficulty for the Stoics, of snakes in a covered basket who look out one by one, so that one cannot
tell whether it is the same or a different snake raising its head each time: SE, M 7.410.
    175

guarantee its truth. Hence, the situation could be described by saying that we infer
the truth of an impression from its characteristic ‘mark of truth’. The result is,
however, an awkward combination of alleged self-evidence with knowledge
reached by inference.
I do not wish to suggest, of course, that the Stoics themselves ever explicitly
described their criterion as a sign of truth. But, given the strong interpretation of
the third clause in their definition of the cognitive impression, it seems to me that
the Academic interpretation was not unfair.

III

I assume that it was this version of their epistemology that both Philo and
Antiochus attributed to the Stoics. Antiochus apparently accepted it because he
had come to think that it was absurd to maintain that knowledge is impossible, but
continued to believe, with other Academics, that the Stoic definition was correct.
Philo, on the other hand, remained convinced by the Academic arguments
purporting to show that the kind of impressions postulated by the Stoics could
not be found. And once the cognitive impression had come to be considered as a
sign, that is, a piece of evidence from which to infer the truth of the corresponding
proposition, Carneades’ arguments would indeed seem to be compelling: to show
that an impression is reliable may justify its acceptance, but it cannot constitute a
demonstration of its truth. Or to put it in the terms of Sextus’s report: even the
most reliable witness may occasionally go wrong, and there is no way of guaran-
teeing, in any given case, that he has not made a mistake.
Now, Philo and other Academics had for quite a while been advocating the view
that knowledge cannot be attained, and that we must therefore be content with
fallible opinions. But the Academy after Carneades had also developed a sophis-
ticated account of more or less reasonable opinions. Philosophers were not
supposed to accept just any impression that might strike them as convincing;
they were expected to exercise caution, to ascertain that the situation was normal
in all relevant respects, and that there was no countervailing evidence from other
impressions or background beliefs to indicate the contradictory of the proposition
under scrutiny. The opinions one would arrive at in this way would be based on
impressions that could be considered reliable, even though there could be no
guarantee of their truth. Moreover, these opinions would coincide for the most
part with what a Stoic might say he had grasped by a cognitive impression.¹⁸
Hence one might come to think that the difference between the ‘probable’
opinions of the Academics and some rash and inconsiderate beliefs was far

¹⁸ See Luc. 105: sed ea quae vos percipi comprehendique eadem nos, si modo probabilia sint, videri
dicimus, and cf. Allen, ‘Academic Probabilism and Stoic Epistemology’, 1994, 104.
176    

more significant than the elusive line between the reasonably justified but fallible
opinions claimed by the followers of Carneades and the allegedly certain know-
ledge guaranteed by the Stoic criterion. Why not, then, follow ordinary life and
redraw the line between knowledge and mere opinion so as to separate well-
founded judgements from superficial beliefs?¹⁹ I would suggest that this was the
consideration that led Philo to his Roman innovations. He concluded, plausibly
enough, that the Stoics had set their standards too high. It was unnecessary
modesty on the part of the Academics to concede that all they could hope for
was opinion, not knowledge.
Our sources unfortunately do not permit us to say exactly how Philo would
have re-defined knowledge (κατάληψις), if indeed he explicitly offered a new
account. There seem to be two options:

(a) Philo might have said that a plausible and thoroughly ‘tested’ impression
was sufficient to warrant the inference to its truth. This would mean that he
continued to adhere to the Academic way of treating impressions as evidence or
‘messengers’ rather than as sources of non-inferential information. This line,
however, would be implausible, given that the Academic arguments for
indistinguishability—which Philo apparently continued to accept—were designed
precisely to show that such an inference could never be warranted.
(b) The second option would be to deny that any inference is needed to acquire
true and reliable information through the senses. One may use the procedures
outlined by Carneades to make sure that an impression is reliable, but it is a
mistake to think that one then goes on to infer the impression’s truth from its
reliability. The impression that there is a table in front of me, together with the
observation that this impression was received under standard conditions, may
justify my claim to know that there is a table. But if this claim is correct, it is
because, as the Stoics might have put it, my senses have revealed the presence of
the table and thus provided me with true and accurate information—or, in today’s
fashionable jargon, because I am in the right causal relation to the table. My
knowledge is not based on the spurious inference from ‘I have the impression that
p, and p is a reliable impression’ to ‘p is true’.
If this was the option Philo chose, he may have been the first to distinguish
clearly between justifying acceptance and offering a proof. He would also have
preserved the distinction between ‘evident’ truths grasped without inference and
other true propositions that may be derived from them by argument and

¹⁹ A consideration of this sort appears at Luc. 146, where Cicero points out that accomplished artists
like the painter Zeuxis or the sculptors Phidias and Polyclitus would be rather offended if told (by a
Stoic) that they had no knowledge. They would look more kindly upon the Academics, realizing that
they ‘reject only what can never be found, but leave in place all that is needed for their crafts’. Cicero
presumably means that they would be content with having their knowledge re-described as probable
opinion; but one might also think that it is absurd to deny that these people have a kind of knowledge.
    177

inference—the distinction that underlies the theory of inference from signs. In

other words, he would have offered something that could be called a criterion of
truth—a non-inferential means of grasping what is the case—though not an
infallible one. Even though we do not have any positive testimony, then, I am
inclined to believe that this would have been Philo’s view.
Once he had taken this step, he could also say that he was just reverting to a
conception of knowledge that had been held by the great founders of the
Academy, Plato and Aristotle.²⁰ For, even though they obviously thought, just as
did Philo himself, that the knowledge of an expert must consist of truths or even
necessary truths, they did not claim that an expert could never make a mistake.
Cicero tells us several times (Luc. 77, 113) that no one before Zeno had put
forward the view that the wise man will have no opinions, and I think this is
historically correct. If an expert did make a mistake, this would show simply
that his claim to knowledge was false in this instance, not that he had no
knowledge at all.²¹
This would still not show that the Academy had always agreed with Philo’s
Roman views, of course, for Arcesilaus and Carneades, as far as we can see, had
followed the Stoics in arguing that the wise man should not assent unless he was
certain he had grasped the truth. But Philo’s version of the ‘true Academic line’
seems to me to have no less plausibility as regards the early Academics than
Antiochus’s picture of a thoroughly Stoicized dogmatic orthodoxy.
In the famous summary of Photius (Bibl. cod. 212, 170a16), Aenesidemus is
said to have described the Academics of his day as ‘Stoics fighting Stoics’. The
dispute between Cicero’s teachers Philo and Antiochus was actually a case of one
Academic arguing against another Academic, but if my interpretation of Philo’s
Roman views is somewhere near the truth, then there would not be much to
choose between these two descriptions. Both parties to the dispute were dogma-
tists, differing only in their greater or lesser modesty; both held theories that can
be seen as versions or variants of Stoic epistemology; and neither of them could
any more be called a sceptic.²²

²⁰ For this point, see also Fin. V 76: nihil enim est aliud quam ob rem mihi percipi nihil posse
videatur, nisi quod percipiendi vis ita definitur a Stoicis, ut negent quicquam posse percipi nisi tale
verum, quale falsum esse non possit. itaque haec cum illis est dissensio, cum Peripateticis nulla sane.
²¹ For Aristotle see, for example, Physics II 8.199a33–35: ἁμαρτία δὲ γίγνεται καὶ ἐν τοῖς κατὰ τέχνην
(ἔγραψε γὰρ οὐκ ὀρθῶς ὁ γραμματικός, καὶ ἐπότισεν οὐκ ὀρθῶς ὁ ἰατρὸς τὸ φάρμακον).
²² I am grateful to James Allen for discussion and criticism of the first draft, and to David Sedley for
giving me a written version of his objections at Utrecht.
 13
Scepticism as a Kind of Philosophy

An examination of Sextus Empiricus’ claim that Pyrrhonist Scepticism is a kind of

philosophy different from both positive and negative dogmatism. What Sextus
describes in the first book of Outlines of Scepticism as the Pyrrhonist way of life is
not a different kind of philosophy, but a way of using reasoning and argument not
only to undermine philosophy itself, but even any kind of judgement based on
reasons. The label ‘investigators’ does seem to apply to the Academic Sceptics, but
theirs is not a different kind of philosophy; it is simply an undogmatic version that
abandons claims to certainty.

Scepticism has been one of the standard problems of epistemology in modern

times. It takes various forms—the most general one being the thesis that know-
ledge is impossible, but equally prominent are such versions as the notorious
doubt about the existence of an external world, inaugurated by Descartes’
Meditations, or doubts about the existence of objective values. Philosophers who
undertake to refute scepticism—still a very popular exercise—try to show that
knowledge is possible after all, or to prove the existence of an external world, and
so on. Sceptics, generally speaking, are seen as radical doubters—philosophers
who call into question assumptions that are usually taken for granted by ordinary
people as well as by other philosophers. Those doubts have to be refuted by
showing that they are in some way unjustified.
It was not always so. In ancient times, when the label Scepticism was intro-
duced together with a variety of others, Sceptics were seen primarily as philo-
sophers who suspend judgement, refrain from making any assertions, either
about philosophical problems or about anything whatsoever, including everyday
statements of fact. The word ‘sceptic’ literally means ‘investigator’ or ‘searcher’,
not ‘doubter’. Here, to illustrate the point, is the opening chapter of Sextus
Empiricus’ Outlines of Pyrrhonism, entitled ‘The most fundamental difference
among philosophies’:

When people are investigating any subject, the likely result is either a discovery,
or a denial of discovery and a confession of inapprehensibility, or else a continu-
ation of the investigation. This, no doubt, is why in the case of philosophical
investigations, too, some have said that they have discovered the truth, some

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0013
       179

have asserted that it cannot be apprehended, and others are still investigating.
Those who are called Dogmatists in the proper sense of the word think that they
have discovered the truth—for example, the schools of Aristotle and Epicurus
and the Stoics, and some others. The schools of Clitomachus and Carneades, and
other Academics, have asserted that things cannot be apprehended. And the
Sceptics are still investigating. Hence the most fundamental kinds of philosophy
are reasonably thought to be three: the Dogmatic, the Academic, and the
Sceptical. (PH 1.1–4)¹

Sextus presents scepticism as a kind of philosophy, distinguished from others not

by the content of its doctrines (there are none), but apparently by its attitude to
philosophical problems and theses. Modern so-called sceptics would fall into
Sextus’s second category—negative dogmatists, people who argue for the view
that knowledge is impossible on the basis of specific assumptions about what
knowledge would have to be. A Pyrrhonist Sceptic would suspend judgement
about the premises of such arguments as much as about their conclusions, and
hence not assert that knowledge is impossible, or that we do not know whether
there is an external world, and so on. In this chapter I propose to take a look at
Sextus’s claim that Scepticism is not a doctrine at all, but rather a different kind of
philosophy.
Sextus’s tripartition is probably an expansion of a simpler twofold division that
can be found in Diogenes Laertius (1.16): ‘Of the philosophers some were dog-
matists, others ephectics [suspenders of judgement].’ Diogenes’ classification
contrasts philosophers who have doctrines with others who suspend judgement,
but his subsequent explication of the two labels already begins to look like the
modern contrast between philosophers who believe they can find the truth and
sceptics who say one cannot: ‘dogmatists are those who make assertions about
things as being apprehended; ephectics are those who suspend judgement about
things as being inapprehensible’—that is, maintain that things cannot be known.
Sextus’s version is more sophisticated. He (rightly) treats the view that nothing
can be known as a philosophical doctrine, and hence puts the Academics on the
side of dogmatism as opposed to his own school, which professes to be simply
continuing the search. Now, if Sceptics are investigators who have no answers to
propose, in what sense do they count as philosophers?
Sextus’s answer should be contained in the general account of Pyrrhonism he
offers in the first book of his Outlines. It will emerge, I think, that ‘perpetual
investigator’ is actually not a very apt description of a Pyrrhonist—but, as I shall
argue later, an accurate description of the Pyrrhonists’ ancient rivals, the

¹ Tr. J. Annas and J. Barnes in Sextus Empiricus, Outlines of Scepticism, Cambridge 1994. I use this
translation for passages from PH unless otherwise indicated.
180    

Academics (precisely those philosophers whom Sextus stigmatizes as negative

dogmatists). But let us look first at Pyrrhonism.
Sextus’s books were written towards the end of a long tradition of non-
dogmatic philosophy, going back at least to the end of the fourth century ,
the lifetime of the founder and figurehead Pyrrho of Elis. The historical develop-
ment of what we now call Greek Scepticism is obscure, owing to the fact that
virtually everything that was written before Sextus has been lost, but one can at
least say that Sextus’s version of Pyrrhonism does not represent a uniform view
that was preserved through the centuries. Apart from Pyrrho himself, who may
have had many admirers during his lifetime, but apparently did not have many
disciples (apart from the notorious Timon²), we should include the ‘sceptical’
period of Plato’s Academy as well, at least from Arcesilaus in the third century
down to Carneades (second century ), because Aenesidemus, the reviver of the
Pyrrhonist sect in the first century , seems to have been an Academic who left
his school to found a rival enterprise, apparently because he found the Academics
of his own time insufficiently sceptical.³ Another influence, documented by the—
no doubt fictitious—‘genealogy’ of the Sceptical school from Timon to Sextus’s
pupil Saturninus in DL (9.115–116), which contains the names of several distin-
guished physicians, is that of the Empiricist school of medicine. The famous
Empiricist doctor Heraclides actually appears on DL’s list as the immediate
predecessor of Aenesidemus. The medical Empiricists from the third century 
on had an ongoing debate with the so-called Rationalists (λογικοί), doctors who
believed that one should try to understand and treat diseases by discovering their
(often unobservable) causes. Our main source for the disputes among the phys-
icians is Galen, who treats the Empiricists as (philosophical) Sceptics; as the
surname ‘Empiricus’ shows, Sextus himself must have been a member of this
school, and we know that he actually wrote a book on Empirical medicine (see
M 1.61), which has unfortunately been lost.
The form of Sextus’s exposition is obviously influenced by the style of the
philosophical historiographers and commentators of his own time: he follows
the pattern that would be used for the account of a school that had doctrines in the
three traditional ‘parts’ of philosophy—physics, logic, and ethics. So, we get first
some notes on the name of the school and its origins, including the question
whether Pyrrhonism can count as a school or sect (αἵρεσις) at all—to which the

² Timon of Phlius (approx. 325–230 ), best known for his satirical attacks on other philosophers,
is probably the ultimate source of all the surviving evidence about Pyrrho, since Pyrrho himself, as one
might expect, wrote nothing.
³ This has recently been doubted by Fernanda Decleva Caizzi (‘Aenesidemus and the Academy’,
Classical Quarterly 1992, 176–89), but see the reply by J. Mansfeld, ‘Aenesidemus and the Academics’,
in L. Ayres (ed.), The Passionate Intellect, Rutgers University Studies in Classical Humanities vii, New
Brunswick/London 1995, 235–48.
       181

answer is a guarded yes (PH 1.16–17). Then Sextus appears to summarize the
position of his school in the three traditional fields (chapters 9, 11, 12). He has
nothing to offer under physics. In the brief chapter entitled ‘Do Sceptics study
natural science?’ (PH 1.18), he candidly admits that Sceptics engage in the
discussion of questions from this field only for the purpose of exhibiting undecid-
able conflicts of opinion in order to reach or perhaps preserve their peace of mind.
He adds that they deal with logic and ethics in the same way. But as a matter of
fact, he does have things to say under the rubrics ‘The criterion of Scepticism’
(PH 1.9 21–24) and ‘What is the end of Scepticsm?’ (25–30), topics that would
normally be considered to belong to epistemology (‘logic’) and ethics, respectively.
Of course, the Sceptics do not have a philosophical view about either the criterion
of truth or the ultimate goal of life, but it turns out that they do use (if not argue
for) what they call a criterion for action, and they also have an ultimate aim,
namely tranquillity, which they claim to achieve by practising their philosophical
method. It looks as though the Pyrrhonists had at least a quasi-system.
But in lieu of the more detailed exposition of doctrines that would have
followed such a summary in the account of a dogmatic school, we get the extended
presentation of the famous Modes—a large collection of materials to be used in
setting up conflicts of opinions about all sorts of subjects, followed by an ingenious
set of epistemological meta-arguments, the Modes of Agrippa, designed to thwart
any attempt by dogmatically inclined philosophers to resolve the controversies
that surround philosophical as well as everyday questions. After that, we get a
highly scholastic series of short chapters about the slogans or utterances (φωναί)
through which the Sceptics expressed their attitude or, as they might prefer to say,
the affection (πάθος) that results from the exercise of their argumentative skills.
Here Sextus describes a variety of ways in which these utterances can be seen not
to be assertions of any sort—for that, of course, would count as dogmatism. This
section is a good illustration of the multiple strands of tradition behind Sextus’s
account of Scepticism. Some of the slogans, such as οὐ μᾶλλον (‘no more this than
that’), go back beyond Pyrrho to Democritus and Plato; others are attempts to
avoid even the appearance of assertion by using, for example, the form of a
question, but there are also phrases like ‘I have no apprehension’ (ἀκαταληπτῶ)
and even ‘Everything is inapprehensible’ that derive from the Stoic–Academic
debate about the possibility of knowledge. Finally, there are six chapters in which
Sextus endeavours to refute the claim, presumably made by other philosophers,
that Pyrrhonism is identical with some other school, notably the Academic and
medical Empiricism.
What are we to make of this portrait of a self-proclaimed group of
philosophers-without-answers? In chapter 8, Sextus had told us that while the
Sceptics do not have a school or sect in the standard sense of being adherents of
some system of philosophical doctrines that contains assertions about obscure
matters, they do belong to a school if by that word one understands ‘a way of life
182    

(αἵρεσις)⁴ which, to all appearances, coheres with some account (λόγος), the
account suggesting how it is possible to live correctly (where “correctly” is taken
not only with reference to virtue, but more loosely, and extends to the ability to
suspend judgment)’ . . . ‘For’ (he continues) ‘we coherently follow, to all appear-
ances, an account which shows us (or ‘suggests’, ὑποδείκνυσι) a life in conformity
with traditional customs and the law and ways of life and our own feelings’. I take
it that all of book I, and not just the brief chapter about the criterion for action, can
be seen as a version of this account—or story, as I might be inclined to translate
here, to distinguish it from either a doctrine or an argument. The story begins,
I think, in chapter 6, with the ‘principles’ of Scepticism. According to this chapter,
the initial motivation of the Pyrrhonist is the hope of reaching tranquillity, to free
himself from the disturbances that befall a person who observes ‘the general
disorder of things’. ‘Men of noble character’, faced with ubiquitous conflicts of
perceptual appearances as well as opinions or doctrines, feel impelled to investi-
gate what is true and what is false. What leads the Sceptic into philosophical
investigations is disturbance and confusion, but he engages in the search for truth,
not just in order to find answers to puzzling questions, but in order to attain peace
of mind. Contrast this picture with Plato’s and Aristotle’s description of the
philosopher as someone who starts out from bewilderment, indeed, but then
desires to find the truth for its own sake. It is true, of course, that the Hellenistic
schools advertised their doctrines, and hence the study of philosophy, as a way of
reaching happiness as well, and Epicurus is even reported to have said that the
study of nature would be useless if it did not lead to tranquillity. But this was
precisely because philosophy was taken to be a way of finding the truth. The
conception of philosophy as a search for truth was also used in the first chapter of
the Outlines quoted above, but here we are told that a Pyrrhonist philosopher is
interested in finding the truth only as a way of reaching peace of mind. This is
why, when he finds himself unable to discover the truth, but nevertheless relieved
of his worries once he has given up on the project, the Sceptic also loses interest in
the investigation of philosophical problems. Scepticism itself, indeed, was defined
accordingly in chapter 4, as ‘an ability to set out oppositions among things which
appear and are thought of in any way at all, an ability by which, because of the
equipollence in the opposed objects and accounts, we come first to suspension of
judgment and afterwards to tranquillity’ (PH 1.8). The Sceptic’s subsequent
investigations are undertaken in order to exhibit the equal strength of arguments

⁴ Annas and Barnes (1994) translate this word as ‘persuasion’, which I find misleading, especially for
readers without Greek, who would naturally take it as translating a word like πίστις. I use ‘way of life’
for lack of a better term. As Roberta Ioli has shown (in Ἀγωγή and related terms in Sextus Empiricus,
MPhil thesis, Cambridge 1999, the use of the word in a philosophical context and in contrast to αἵρεσις
suggests following a lead—either a person or a set of instructions. So in its second occurrence here (l.23)
it seems to correspond to the phrase διδασκαλία τεχνὦν (teaching of kinds of expertise) in the
description of the ‘everyday observances’ followed by the Sceptics (PH 1.9.24). For ‘way of life’, see
the explication of the word by Sextus at PH 1.145.
       183

on both sides of each conflict, which will lead the investigator to suspend judge-
ment; and suspension of judgement is supposed to lead to tranquillity. Sextus
explains how this allegedly happens in his chapter about the goal of Scepticism,
PH 1.9.25–30, by telling an anecdote:

What is said to have happened to the painter Apelles befalls the Sceptic too. They
say that Apelles was painting a horse and wanted to represent the foam at the
horse’s muzzle. He was so unsuccessful that he gave up and hurled at the
picture the sponge he used to wipe the paints off his brush. The sponge touched
the picture and produced a representation of the foam.

So, for the Sceptic, tranquillity follows unexpectedly, not upon the discovery of
truth, but upon giving up the search. The Sceptic finds himself unable to decide
between conflicting appearances or opinions and suspends judgement—and then
notices, to his surprise, that his previous worries have left him together with the
attempt to arrive at a decision. This is clearly not a philosophical theory about the
goal of life, but it is a story about the aim of the Sceptical philosopher as a person,
and an explanation of his way of life. If we take this as a part of the story
mentioned in chapters 4 and 8, it tells us that a Pyrrhonist will use his skill to
construct a situation of equipollence whenever he encounters a conflict of doc-
trines or appearances, which will then lead him to abandon the project of finding
answers, and consequently be no longer troubled by the problem.
But we were also told that the Sceptical story suggests a way of living correctly—
and that is the point taken up in the chapter about the Sceptical criterion. It
contains the Pyrrhonists’ answer to the charge, brought against those who claim to
suspend judgement on all matters at least since the time of Pyrrho, that a person
who refrains from making any judgements at all will be unable to act, since action
implies a decision or belief about whether things are one way rather than another.⁵
Obviously, the Sceptic cannot base any judgements or decisions on a principled
method of establishing the truth or making moral decisions. Having given up the
attempt to find guidance in reason and argument, he might be at a loss as to how
to conduct his life—or indeed, as the dogmatist opponents claimed, unable to act
at all. But this, as Sextus points out, is not the case, for instead of reason-based
judgements the Sceptic will make use of appearances—he will respond to the way
things appear to him, act as his instincts such as hunger or thirst impel him to act,
without thinking that he is right about any view with which another person might
disagree.

⁵ I have discussed some sceptical replies to this argument in ‘Sceptical Strategies’ (in M. Schofield
et al. (eds.), Doubt and Dogmatism, Oxford 1980, 54–83), repr. in Striker, Essays on Hellenistic
Epistemology and Ethics.
184    

I will not go into the vexed question whether the Pyrrhonists’ ‘appearances’ are
to count as some kind of beliefs or not. It seems to me that the debate about this
question⁶ has suffered to some extent from a failure to distinguish between
different senses of ‘belief ’. It may be taken in the strong sense of ‘judgement’,
meaning assent to a proposition justified by appropriate reasons one is prepared
to produce in order to defend the truth of one’s assertion, but it can also be used in
a much weaker, dispositional sense, according to which it is sufficient for the
ascription of a belief to an agent that she acts or behaves in a certain way. If I avoid
an approaching car, for example, I thereby show that I believe that there is a
dangerous heavy object coming towards me that might kill me if I don’t get out of
the way. I could no doubt offer reasons, both for the belief and for the action, but
I probably didn’t think of them. In fact, my dog might have reacted in exactly the
same way, though it cannot offer reasons to justify its ‘belief ’. I would say that the
Pyrrhonist conception of ‘following appearances’ is on the model of this kind of
behaviour, and it is a matter of terminological choice whether we want to speak of
belief here or not.
Appearances and natural urges arise passively and involuntarily (PH 1.22) and
hence are not based on judgements as to what is or is not the case, ought or ought
not to be done, but they are sufficient to keep the Sceptic going, as it were.
A simple version of this reply to the dogmatic objection may go back to Pyrrho
himself and was later elaborated in the sceptical Academy. But while earlier
answers addressed only the question of how a Sceptic can orient himself in a
particular situation, Sextus’s version purports to show that we can make do with
appearances in every aspect of ordinary life. Following appearances or, as he calls
it, ‘everyday observances’ (βιωτικὴ τήρησις PH 1.23) has four parts: first, ‘guid-
ance by nature’, which provides the Sceptic with sense-impressions and thoughts;⁷
second, ‘necessitation of feelings’, such as hunger and thirst, which will lead the
Sceptic to go for food and drink; third, the ‘handing down of laws and customs’,
which leads the Sceptics to accept, for example, piety as good ‘in an everyday
fashion’, and impiety as bad; and fourth, ‘teaching of kinds of expertise’, which will
permit the Sceptic to practice the crafts in which he has been trained. So, Sextus
can claim that the Sceptic will lead a perfectly ordinary life, including conformity
to the moral views of his community, and that he will even be able to make a living
as a practitioner of some craft by following the instructions of his teachers. None

⁶ For the debate see, for example, the essays by M.F. Burnyeat, M. Frede, and J. Barnes reprinted in
M.F. Burnyeat and M. Frede, The Original Sceptics, Indianapolis 1997.
⁷ The Greek text is: ‘ὑφηγήσει μὲν φυσικῇ καθ’ἣν φυσικῶς αἰσθητικοὶ καὶ νοητικοί ἐσμεν. M. Frede
(‘The Empiricist Attitude towards Reason and Theory’, in Apeiron xxi, 1988, 95) finds in this passage
the recognition of reason as a natural human capacity, but I think one should be careful not to conclude
that this means reasoning. The word νοητικοί should probably be taken only in the minimal sense of
being able to think and understand language, as, for example, in PH 2.10, where Sextus seems to refer
back to this passage. There can be no question, of course, that the Pyrrhonists made use of reasoning in
their arguments against other philosophers, but I think they kept it out of their sceptical way of living.
       185

of this, or so we are invited to believe, requires any reasoned decision about the
truth or falsity of potentially controversial views, and hence the Sceptic never
needs to be disturbed. But in case he should find himself tempted to enter a
dispute, he will presumably use his Sceptical technique of argument to rid himself
of the desire to find out how things really are.
Such, in outline, is the account that guides the Sceptic’s way of life—one that
coheres remarkably well with the anti-rationalism of the Empirical doctors,
though avoiding, at least in Sextus’s book, their dogmatic assertion that nothing
can be known about things non-evident (PH 1.236). There is, however, a compli-
cation in Sextus’s case, arising from his frequent use of the Modes of Agrippa
(PH 1.164–177). This set of argument-forms is no doubt an effective device to stop
people from trying to solve puzzles by justifying some particular view about how
things really are, but it is not a method for reaching equipollence as advertised in
the definition of Scepticism. It is, in fact, a piece of negative dogmatism designed
to convince dogmatists that no judgement can ever be sufficiently justified to
count as an instance of knowledge.
And, pace Jonathan Barnes,⁸ it seems to me to be just as questionable as the
older Academic argument for the impossibility of knowledge against the Stoics.
Briefly, the dubious assumptions here are (1) that no judgement can be accepted
as a piece of knowledge without the backing of reasons; and (2) that every
judgement can be disputed. The argument overlooks the vast number of facts
we claim to know unhesitatingly without having the faintest idea of how we would
defend or justify them if challenged, and without consciously relying on any
reasons. Examples would be not only everyday perceptual beliefs, but also such
things as knowing who our parents are, in what city we are living, etc., and
generalities such as that cats are animals, books are made for reading, and humans
are language-users. The first sort can occasionally be called into question when
there is reason to suspect delusion or trickery, but the default-option, as it were, is
truth. To say that we are never justified in making even such modest claims is to
introduce the method of Unreasonable (‘hyperbolic’) Doubt made famous by
Descartes, and I see no reason to follow Descartes on this point. Aristotle was
right, I think, to say that if there is to be knowledge, some truths must be known
without proof. Since he was interested in scientific knowledge, not justification of
knowledge-claims in general, he then went on to consider the first principles of
demonstration, rather than humdrum bits of everyday knowledge (the knowledge
we need to have in order to follow any kind of systematic instruction: An.Po.
A 1.71a1–2). But his move, though never intended as a Refutation of Scepticism,
seems to me to be the right response to the later sceptics as well. A Refutation of
Scepticism is no doubt a hopeless enterprise once we accept the sceptic’s premises,

⁸ J. Barnes, The Toils of Scepticism, Cambridge 1990.

186    

but why should we not consider his argument as a reductio ad absurdum of its
premises?
Sextus’s fondness for this technique is understandable if one considers that the
old set of ten Modes contains quite a few weak items and so is not likely to
convince a dogmatist that no question is ever decidable. Sextus may be using
Agrippa’s arguments only ad hominem, assuming that the dogmatist will agree
with their premises, but to the extent that he seems to think they will have no
escape, his practice threatens to undermine the claim that the Pyrrhonist will lead
a life without dogmatic assent.
Now, since Sextus’s definition of Scepticism appeals to equipollence, and
Agrippa’s Modes must have been introduced after Aenesidemus,⁹ let me set this
question aside for the moment and consider only the official account of the
Sceptical way of life. Is it also a kind of philosophy? Yes and no, I think. If by
philosophy we mean a search for truth, successful or otherwise, the Sceptical way
of life can hardly qualify. Contrary to Sextus’s initial claim that the Sceptic goes on
investigating, philosophical investigations seem to be precisely what the Sceptic’s
way of life is designed to avoid. The impressive apparatus of the Sceptical modes is
supposed to be used for one purpose only—namely to rid us of the foolish
attachment to settling questions by reason and argument. Or, as one might put
it, to bring out the paradoxical nature of the Pyrrhonist enterprise, reasoning and
argument are used only to undermine themselves. But this vehement anti-
rationalism seems to be based upon a rationale itself—the Sceptics’ claim that
no controversies will ever be settled; that abandoning the search for truth will give
you peace of mind; and that there is no need to fall back on reason and argument
in everyday life because all of it can be conducted by following the appearances
that come to us involuntarily, untainted by the influence of reason. One might be
tempted to call this set of claims a philosophical theory, however rudimentary—
but in fairness to the Sceptics, one has to recognize that Sextus takes great care to
make sure that it is not mistaken for one. No part of the set is presented as the
conclusion of an argument. Sextus carefully qualifies the claim that the attempt at
settling a dispute always ends in a stalemate by saying that it refers only to the
cases investigated by the Sceptic so far. The full formulation of the Sceptical slogan
‘there is an equally strong argument opposed to every argument’¹⁰ would have to
run as follows: ‘To every argument I have scrutinized which purports to establish
something in dogmatic fashion, there appears to me to be opposed another

⁹ Nothing is known about Agrippa, and the only source that ascribes the Five Modes to him is DL
(9.88). F. Caujolle-Zaslawsky (Dictionnaire des philosophes antiques, ed. R. Goulet, Paris 1989, vol. I, 71)
suggests that Agrippa might in fact have been simply a character in a book by the Sceptic Apellas (see DL
9.106), a man about whom we also do not know anything else. Still, both Sextus and Diogenes say that
the Five Modes were introduced later than Aenesidemus’ Ten Modes.
¹⁰ Here I prefer the translation ‘argument’ for λόγος to Annas and Barnes’ ‘account’, because the
slogan was, after all, not invented by the Sceptics—it goes back, minus the significant word ἴσος (equal),
to the sophist Protagoras.
       187

argument, purporting to establish something in dogmatic fashion, equal to it in

convincingness or lack of convincingness’ (PH 1. 203). And to make sure that even
this very cautious formulation not be taken as an assertion, Sextus will insist that
any utterance by a Sceptic should only be seen as an expression of the way he is
affected. The claim that tranquillity follows upon suspension of judgement is
presented in the form of a biographical anecdote—no promise is made that the
experience can easily be repeated, or that tranquillity necessarily must follow
suspension (although I must say that the phrase ‘as a shadow follows a body’,
used by both Sextus (PH 1.29) and Diogenes Laertius (9.107), sounds a little
suspicious). The ‘everyday observances’ that Sextus offers as guidance for the
Sceptic’s way of life are a set of replies to dogmatic claims that this or that activity
might require reasoned judgement and deliberation. Sextus points out, in each
case, that no reasoning is needed to perform in (allegedly) normal ways; he does
not try to defend the view that it is right or reasonable to follow appearances.
Hence a Pyrrhonist would probably not be impressed if one objected to him
that his overall story is neither very plausible nor very appealing. For example, a
dogmatist would be likely to object that several of the Ten Modes seem to present
only apparent conflicts. To take a simple example: if A likes apples and B does not,
this need not (and usually does not) lead to a dispute as to whether apples are
good-tasting or not; having a good taste is not a property of apples taken by
themselves, but only of apples as tasted by people. Again, if eating apples is healthy
for humans but unhealthy for cats, this is no reason for dispute, because these facts
are relational—but they seem to be pretty well-established facts. So, the Sceptic’s
claim that things in general are full of conflicting appearances seems to be vastly
exaggerated, making his remedy for getting rid of these conflicts seem less
urgently needed. But even if we consider only genuine controversies such as
those among dogmatic philosophers, it seems to be false to say that no dispute
can ever be resolved, since some views at least are generally agreed to be mistaken.
Obviously, dogmatists would (and did) also object to Sextus’s claim that dogmatic
belief about real goods or evils will inevitably lead to insufferable troubles. And
finally, why should one believe that giving up the search for answers will lead to
tranquillity rather than to profound gloom?
But Sextus, as I said, does not purport to rely on argument to convert others to
Pyrrhonism. He is not in the business of arguing that everybody should, let alone
ought to, become a Pyrrhonist. His comments on the Sceptical slogans describe
them as being like a purgative that removes itself from the body along with the bad
humours (PH 1.206).¹¹ This is, I think, an apt description of Sextus’s arguments,

¹¹ For this description as a parody of the dogmatists’ model of philosophy as a therapy of the soul,
see A.J. Voelke, ‘Soigner par le logos: la thérapeutique de Sextus Empiricus’, in ‘Le Scepticisme Antique’,
Cahiers de la revue de théologie et de philosophie 15, 1990, especially 192–4. He quotes the following
188    

for, taken as a philosophical theory, Pyrrhonism is self-defeating—an argument to

show that arguments are useless.
The Pyrrhonist story thus has a tenuous status, being presented as nothing
more than an expression of the Sceptic’s—possibly temporary—state of mind. But
then one should admit that, in Sextus’s version at any rate, the story is presented
with great vigour and vivacity. Going through an interminable array of arguments
that appear to leave one stymied again and again may well have the intended or
promised effect of making the reader think that the dogmatists’ enterprise is
doomed to failure. And if you happened to be an anti-rationalist (who recognizes,
as Sextus explicitly does, that Empiricism is just another theory, PH 1.236)—what
else could you do to get your message across?
So much for Sextus’s version of the Pyrrhonist account. If this is what was
meant by Scepticism as a kind of philosophy, then it is a philosophy in a peculiar
sense—a discipline that uses argument and reasoning, not as a method to discover
the truth, but only as a therapeutic device to deter people from trying to be guided
by reason. But why would anyone wish to describe this as a continuation of the
ordinary philosopher’s search for truth? I do not know, of course, whether the
anti-rationalism of the Outlines was characteristic of the entire Pyrrhonist move-
ment from Aenesidemus on—though it fits in rather well with Photius’ summary
of Aenesidemus’ books, and with the set of Modes against causal explanations
ascribed to him at PH 1.180–185. But it certainly was not a part of the Pyrrhonists’
Academic inheritance—and I suspect, indeed, that the label ζητητικός (inquirer)
was taken over from the Academy. It is well known that Sextus’s description of the
Academics as negative dogmatists is not entirely fair, at least as far as the sceptical
Academy from Arcesilaus to Carneades was concerned.¹² Cicero repeatedly tells
us in his Academica that neither Arcesilaus nor Carneades wished to claim that
they knew that nothing could be known, nor even that they themselves knew
nothing, as Socrates allegedly did (Ac. I 45). But the same Cicero also informs us
that in the generation after Carneades there were philosophers in the Academy
who did accept the conclusion of the celebrated anti-Stoic argument to the effect
that nothing can be known (ἀκαταληψία) and proposed to make do with plausible
opinion rather than knowledge. These were presumably Aenesidemus’s contem-
poraries, and if Aenesidemus himself was an anti-rationalist, it is easy to see why

comment by Plutarch: οὐ γὰρ ὡς ἐλλέβορον, οἶμαι, δεῖ θεραπεύσαντα συνεκφέρειν τῷ νοσήματι τὸν λόγον,
ἀλλ’ ἐμμένντα τῇ ψυχῇ συνέχειν τὰς κρίσεις καὶ φυλάσσειν· φαρμάκοις γὰρ οὐκ ἔοικεν ἀλλὰ σιτίοις ὑγ
εινοῖς ἡ δύναμις αὐτοῦ . . . de cohib. ira 453 D–E: ‘For I do not think that reason should be used in one’s
cure as we use hellebore, and be washed out of the body together with the disease, but it must remain in
the soul and keep watch and ward over the judgements. For the power of reason is not like drugs, but
like some wholesome food . . . ’ tr. W.C. Helmbold).
¹² For this point see my earlier paper ‘Über den Unterschied zwischen den Pyrrhoneern und den
Akademikern’, Phronesis 26, 1981, 153–71 (English tr. in G. Striker, Essays on Hellenistic Epistemology
and Ethics, Cambridge 1996). At that time, however, I did not sufficiently appreciate the importance of
Pyrrhonist anti-rationalism.
       189

he would have been dissatisfied with their position. For the appeal to the greater or
lesser plausibility of impressions, accompanied by a set of rules for the confirm-
ation or testing of candidate impressions would, of course, reinstate the enterprise
of trying to get somewhere near the truth by finding good reasons for one’s beliefs.
From the point of view of an enemy of reason, it would hardly make a difference
whether your claims were modest or arrogant, claiming just a reasonable degree of
plausibility or claiming certainty. Sextus makes this point in his chapter about the
difference between his own school and the Academy. But there he also recognizes
that not all Academics took this line. He exempts Arcesilaus—who, as one realizes
at this point, had not been mentioned in the introductory chapter—from the
charge of dogmatism and goes so far as to say that

his way of life (ἀγωγή) is virtually the same as ours. For he is not found making
assertions about the reality or unreality of anything, nor does he prefer one thing
to another in point of convincingness or lack of convincingness, but he suspends
judgment about everything. And he says that the aim is suspension of judgment,
which, we said, is accompanied by tranquillity. He also says that particular
suspensions of judgment are good and particular assents bad. (PH 1.232–233)

The first sentence distinguishes Arcesilaus from Carneades and later Academics
who treated the convincingness or otherwise of impressions as a criterion for
action. This theory was invented by Carneades, who offered it as a reply to the
argument that it is impossible to act without belief, though not as his own serious
proposal for the conduct of life or for philosophical investigations. Arcesilaus, as
we hear from Plutarch (adv. Col. 1122A–F), had given a different reply: he
suggested that assent, and hence judgement, was not necessary for action, since
people could act by simply responding to sense-impressions and corresponding
impulses in the way that, according to the Stoics themselves, other animals do.
This is, indeed, very similar at least to the first part of the ‘Sceptical criterion’. But
Sextus does not here mention what he reports elsewhere (M 7.158), namely that in
another context Arcesilaus also offered a different ‘criterion’: that the wise man, in
the absence of cognitive impressions, that is, knowledge, may regulate his actions
by appeal to ‘the reasonable’ (τὸ εὔλογον). This omission is no doubt significant,
for it shows that Arcesilaus could, on other occasions, admit the appeal to reason
that the Pyrrhonists so strenuously avoided. Sextus, it seems, is trying to paint
Arcesilaus as a predecessor of his own anti-rationalist stance.¹³
The second sentence is a little puzzling, since it appears to ascribe to Arcesilaus
a view about the aim of life (τέλος) that he is not likely to have held. But the

¹³ Mansfeld (‘Aenesidemus . . . ’, n. 2 above) makes the intriguing suggestion that this attempt to turn
Arcesilaus into a Pyrrhonist might actually go back to Aenesidemus, who might have tried to win over
some fellow Academics to his new school.
190    

remark that suspension of judgement is accompanied by tranquillity is, of course,

Sextus’s own addition, and it seems more likely that Arcesilaus might have said
that suspension of judgement was the aim of philosophical discussion, which would
also explain why he approved of particular acts of suspension, and disapproved of
particular acts of assent. This would not mean that he saw suspension of judge-
ment as a means to attain happiness or tranquillity, but that he considered it as the
appropriate attitude for a philosopher. Cicero tells us that both Arcesilaus and
Carneades invested a lot of effort into keeping their interlocutors and students
from rash assent. He even describes Carneades as performing a ‘Herculean labour’
(Luc. 108), given the human propensity to go down the slippery slope of assenting
to one impression after another. We are also told that the Academics used to hold
back with their own views in a debate in order to avoid the undue influence of a
teacher’s authority on their students (Luc. 60). What seems to lie behind these
efforts is not the hope for tranquillity, but rather an appeal to intellectual honesty
or the norms of rational inquiry—not to make a firm assertion unless you are sure
that it is sufficiently justified, and to avoid error as far as possible. This appeal to
intellectual honesty and avoidance of rashness was a part of the Academics’
Socratic and Platonic heritage, and it seems clear from Cicero’s report that they
considered it as a norm of philosophical investigation. They advocated suspension
of judgement as a matter of rational caution, not because they regarded it as the
inevitable outcome of any controversy, philosophical or otherwise. If they tried to
reach a situation of equipollence, this would be an effective way of keeping people
from assenting to a philosophical view just because they found it plausible or
appealing, but even in the absence of strong counterarguments, the Academics
would no doubt have recommended suspension of judgement. I take it that
Carneades’ ingenious critical arguments in many fields of philosophy were
designed to show, against the best philosophers of his day, that the questions
remained open, however convincing a dogmatic theory like the Stoic might have
looked. As far as we can see, the Academics also did not resort to epistemological
meta-arguments of the sort of Agrippa’s Modes in order to establish the undecid-
ability of controversies quite generally and regardless of subject matter. Of course,
Agrippa’s Modes were introduced a long time after either Arcesilaus or Carneades,
but the argument-forms of these Modes could be found already in Aristotle’s
Posterior Analytics (A 3), to which Arcesilaus at least would no doubt have had
access. However, Agrippa’s Modes are ways of blocking the attempt to resolve a
dispute by further investigation, purporting to show that no resolution can ever be
reached. The Academics, by contrast, seem to have been interested precisely in
keeping investigation going, and the arguments ascribed to Arcesilaus, and par-
ticularly to Carneades, always address specific issues, often advancing the discus-
sion by suggesting improvements that their opponents would, at times, be happy
to take over. In short, Academic suspension of judgement seems to have been
motivated by philosophical considerations rather than the sheer inability to decide
       191

between equally strong arguments for conflicting views. Hence, it is not surprising
to see that Carneades’ reply to the inactivity-argument looks so very different from
the Pyrrhonist one. He had no specific objection to the use of reason to arrive at a
belief, and so he offered a modest alternative to the Stoic demand for cognitive
certainty. Provisional, fallible assent will be enough for everyday life and compat-
ible with suspension of judgement if, by judgement, one means an assertion that
purports to be based on certainty, but we do not need to be content with the
involuntary appearances that we find ourselves having apart from any activity of
reason.
So, the Academics present a rationalist version of Scepticism, and one that can
properly be described as a matter of leaving all philosophical questions open and
continuing the search for truth. Their advocacy of refraining from judgement
would itself be based on argument—and one such argument, ironically, has found
its way into Sextus’s text:

. . . when someone propounds to us an argument we cannot refute, we say to him:

‘Before the founder of the school to which you adhere was born, the argument of
the school, which is no doubt sound, was not yet apparent, although it was really
there in nature. In the same way, it is possible that the argument opposing the
one you have just propounded is really there in nature but is not yet apparent to
us; so we should not yet assent to what is now thought to be a powerful
argument.’ (PH 1.33–34)

Sextus presents this as an example of opposing ‘present to past or future things’,

but clearly the ‘future thing’ is not yet available. Hence, we do not have a situation
of equipollence—as Sextus admits, since he begins by saying that the Sceptic
cannot refute an argument presented to him. Why then does the Sceptic not
acquiesce for the time being, as Sextus himself suggests in a similar passage
(M 8.473–475)? If he refuses to give his assent, this is not because he is unable
to decide which side is the more convincing. For Sextus, this argument presum-
ably serves the same purpose as Agrippa’s set—it blocks either assent or further
investigation and thus preserves the Sceptic’s peace of mind. But like Agrippa’s
Modes, it belies the claim that suspension of judgement comes to him ‘passively’
and is a mere affection. Sextus might no doubt retort that he is using the argument
only ad hominem, to show his dogmatic opponent that he, the dogmatist, should
not take it that he has established anything with certainty. But there is no
suggestion in Sextus’s text that the Sceptic is inviting his opponent to join him
in further investigation, or to go looking for flaws in the proposed argument.
Rather, he appears to rely on an inductive argument to show that even the most
compelling philosophical theory might one day be overthrown, and to suspend
judgement because he wants to avoid rash assent. Hence, he seems for once to go
against his passive affections and to follow reason.
192    

Whether or not the argument Sextus uses here ultimately goes back to the
Academy we cannot tell. But it has always seemed very persuasive to me, and it
spells out a plausible general reason for the cautious attitude of the Academics.
This is an argument that applies specifically only to judgements based on elaborate
theoretical justifications, since it would be preposterous to claim that most of our
particular factual beliefs turn out by hindsight to have been mistaken. But as
regards philosophical theories, it seems to me that the inductive basis for this
argument has grown continually over subsequent centuries of philosophy. Still,
I would not wish to say that the attitude of open-minded, perpetual investigation
distinguishes the Academic philosophers from their more dogmatically inclined
rivals so drastically as to deserve to be called a different kind of philosophy. The
division into dogmatists and suspenders of judgement is in the end probably just a
polemical move, introduced to highlight a difference in attitude but definitely not
in method or subject matter. Few philosophers these days call themselves sceptics,
because Scepticism has come to be identified with negative dogmatism. But the
attitude adopted by the ancient Academics may actually be more widespread these
days than it ever was in antiquity.
In the philosophy department at Harvard University hangs a portrait of the
psychologist and philosopher William James holding a book with the faint
inscription EVER NOT QUITE. Now, William James was not a Sceptic, either
in the ancient or in the modern sense—indeed, one of his most famous essays
bears the title The Will to Believe (1896). He writes there about religious belief in
particular, but among other arguments he makes a point that would be a signifi-
cant objection to the attitude of suspending judgement advocated by the
Academics. James insists that the injunctions to avoid error and to go on looking
for the truth should be kept distinct, and that the search for truth may sometimes
require taking the risk of believing something of which one is not, nor cannot
be, certain. So, William James was not a Sceptic—but he was, of course, an
academic.¹⁴

¹⁴ Thanks are due to James Allen for reading and commenting on the first draft of this chapter.
Earlier versions were read at Cornell University and the University of Sheffield. I am grateful for
discussion and advice on both occasions, and in particular for Gail Fine’s set of written comments.
 14
Epicurean Epistemology

This chapter presents Epicurus’s theory of knowledge as a response to the

epistemological pessimism of Democritus. The first section discusses the three
‘criteria of truth’—sense-impressions, preconceptions, and the feelings of pleasure
and pain; and also Epicurus’s account of perceptual error, by which he tried to show
that the senses do after all offer true and consistent, if sometimes fragmentary,
information. The second section deals with Epicurus’s scientific method, that is, the
confirmation and refutation of beliefs or theoretical hypotheses. In a final section,
I briefly discuss the role of the feelings as criteria in ethics, and the kind of reasoning
Epicurus used to call ἐπιλογισμός.

Epicurus was perhaps the first Greek philosopher who found it useful or
necessary to begin the exposition of his natural philosophy with a chapter on
epistemology. He had set aside the part labelled logic or dialectic, the theory of
reasoning and argument, along with its technical terminology, as a superfluous
distraction (DL 10.31) and replaced it by a treatise on how to find and establish
the truth.
By the end of the fourth century , the question whether philosophers could
establish the truth of their theories had become more pressing than ever. Earlier
cosmologists had often rejected the senses as a source of reliable information
because they seemed to offer conflicting evidence. But the theories of the philo-
sophers, allegedly based on reason, had also been challenged by Pyrrho, an older
contemporary of Epicurus, because the philosophers contradicted one another
just as much as did the senses. Even Epicurus’s predecessor in atomism,
Democritus, had expressed deep pessimism about the possibility of attaining
knowledge. He had accepted the impossibility of deciding which among the
conflicting sense-impressions might be true, but at the same time acknowledged
the senses as our basic source of information about the world, expressing the
dilemma in the vivid image of a dialogue between the mind and the senses, in
which the senses address the mind by saying ‘Wretched mind! Do you take your
evidence from us and then overthrow us? Our overthrow is your downfall’
(Democritus fr. B 125; tr. Barnes).
Epicurus had to show a way out of this dilemma, and he did so in a treatise
(now lost) entitled Canon or About the Criterion that served as introduction to his

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0014
194    

natural philosophy. His work was obviously influential, and for several centuries
after Epicurus, epistemology was discussed by the Hellenistic philosophical
schools alongside logic—for those who did not reject it—in terms of the question
of the criterion or criteria of truth.

The Criteria

The word ‘criterion’—literally, instrument of judgement—had been used occa-

sionally by Plato and Aristotle to refer to a cognitive faculty, either reason or the
senses. This usage continued throughout the Hellenistic period, including in
Epicurus’s own writings, and did not belong to any particular school.¹ Thus
Epicurus speaks of criteria in this sense when he urges repeatedly that the student
of nature must consider the available evidence from all the criteria, both the mind
and the senses (Letter to Herodotus 38; 82). Epicurus’s innovation consisted in
speaking of criteria of truth as means of determining or deciding about the truth of
beliefs or scientific theses. He had probably taken over the title ‘Canon’ from
Democritus, who wrote a work entitled Canons, but Epicurus may have been the
first to take literally the metaphor implied by the word. A canon was a mason’s
rule or straightedge, an instrument used to determine the straightness or other-
wise of walls or beams etc. This instrument had to be straight itself to serve as a
standard of straightness for other things. Following this model, Epicurus desig-
nated as criteria of truth kinds of basic or irrefutable truths that could serve to
assess the truth of beliefs.
Diogenes Laertius introduces the three letters of Epicurus included in Book 10
of his Lives of Eminent Philosophers with a brief summary of the Canon. Not
surprisingly, it suggests that the treatise was mainly concerned to argue for the
truth of Epicurus’s criteria:

In the Kanon, Epicurus says that the criteria of truth are sensations, preconcep-
tions, and feelings (the Epicureans also include the applications of the mind to an
impression) . . . ²

¹ For the notion of a criterion in Hellenistic philosophy see G. Striker κριτήριον τῆς ἀληθείας,
Göttingen 1974, and ‘The problem of the criterion’(1990), both repr. in Striker, Essays on Hellenistic
Epistemology and Ethics, Cambridge University Press 1996.
² A note on translation: I follow most recent Epicurus scholars in translating the word αἴσθησις by
‘sensation’ where it is not used to refer to one of the sense faculties. I use ‘impression’ or ‘sense-
impression’ to render φαντασία and adjectives derived from it, as in the phrase φανταστικὴ επιβολὴ τῆς
διανοίας (‘application of the mind to an impression’). Note that the interpretation of this phrase is
controversial; for different interpretations, see for example A.A.Long and D.N. Sedley. The Hellenistic
Philosophers, Cambridge University Press 1987, vol. 1, 90) and E. Asmis, ‘Epicurean Empiricism’ in
J. Warren (ed.) The Cambridge Companion to Epicureanism, Cambridge 2009, 94.
   195

For, he says, all sense perception is irrational and does not accommodate
memory. For neither is it moved by itself, nor when moved by something else
is it able to add or subtract anything.
Nor is there anything that can refute the senses: neither can like sense refute like,
because of their equal validity, nor unlike unlike, since they are not judges of the
same things; nor can reason, since all reason depends on the senses; nor can an
individual sensation refute another, since we pay attention to all of them.
(DL 10.31–32; tr. Long and Sedley with modifications)

The first argument emphasizes the passivity of the senses: they can only receive or
register what moves them, but neither add nor omit anything. This in itself would
not be enough to show that the information they receive must be correct, but the
mention of memory indicates what Epicurus had in mind: memory is a function of
thought, not of perception, and it may be used to modify the incoming informa-
tion. In this way Epicurus could explain what others would describe as perceptual
errors as the result of interference by the mind. Instead of rejecting a sense-
impression as false, one should distinguish between what is already given and
the modifications one has inadvertently made, leading to a false belief. As
Epicurus puts it succinctly in Key Doctrine 24:

If you reject any single sensation absolutely and do not distinguish in an opinion
between what is still awaited and what is already present according to sensation
and feelings and any application of the mind to an impression, you will throw
into confusion even your other sensations by your foolish opinion, so that you
will be rejecting the criterion altogether.
(DL 10.147, tr. Long and Sedley with modifications)

The distinction between the actual content of a sensation and a perceptual

judgement based on it allowed Epicurus to deny the claim that the senses offer
conflicting evidence. According to him, individual sensations provide only partial
information about external objects; this information is correct, but it may not be
sufficient to justify a judgement about the object itself as opposed to its appear-
ance. So, for example, a tower looks small and round when seen from afar, though
one may find out that it is actually square and large when one comes closer. An oar
seen through water does look crooked even though one knows that it is straight.
The sound of a trumpet is not as loud coming from far away as it would be if one
were standing next to the trumpeter, and so on. These are familiar facts about
sense perception—and indeed the argument from conflicting appearances would
not have been so successful if people had not been familiar with them. But people
quickly learn to adjust their judgements to the circumstances, taking into account
their distance from the tower, or the fact that the oar is seen through water, and
196    

suspending judgement when they realize that they do not have enough informa-
tion. The alleged contradictions arise only on the level of perceptual judgement,
because we are prone to forming such beliefs based on very little information.
Again, this is not surprising, since things observed nearby under normal condi-
tions tend to be just as they appear or are perceived to be. One might, therefore, be
inclined to go in the direction the Stoics took by trying to distinguish true from
false sense-impressions. But as Democritus and others had already argued, this
will not work, since there is no standard of judgement above the senses.
The next argument in Diogenes’ summary of the Canon repeats this point and
also emphasizes the limitations of the senses.
The claim that the senses cannot refute one another because they do not judge
the same things can only hold with respect to the proper sensibles—colour, shape,
and motion for sight, sound for hearing, hardness and softness for touch and so on
(see Lucretius. DRN 4.486–499). If something looks like honey, but then turns out
to taste and smell like soap, one might argue that the senses of taste and smell have
refuted sight; but it is still true that the thing has the colour of honey, and that is all
that the eyes could tell one.
Here Epicurus seems to endorse the most restrictive conception of sense
perception advocated by Plato in the Theaetetus (184B–186E). A passage in
Lucretius’s long discussion of optical illusions confirms this. Lucretius has just
explained that the shadow may appear to be an object that follows the body
around, but that it seems to move only because the moving body is blocking light
in different places. Yet this does not show that the eyes deceive us—it is up to the
mind, not the eyes, to determine whether the shadow is a single moving thing or
just the result of the body blocking the light (DRN 4.379–387). But unlike Plato in
the Theaetetus, who goes on to argue that there is no truth to be found in the
senses, Epicurus insists that reason cannot refute the senses, since it is entirely
dependent on them. For, as Diogenes explains a few lines further down, ‘all our
notions also derive from the senses, by encounter or analogy or similarity or
composition, with some contribution from the mind as well’ (10.32).
A true perceptual judgement such as ‘This is honey’ will then be the result of an
‘application of the mind’—an act of attention to the content of many sensations
that leads to a complex impression.
Epicurus’s second criterion, the preconceptions (προλήψεις), is described by
Diogenes in a rambling paragraph that uses terminology from various schools:

Preconception, they say, is as it were a cognition (κατάληψις), or a correct

opinion, or conception, or universal notion stored inside, that is, a memory of
what has frequently appeared from outside, e.g., ‘Such and such a thing is a man’.
For as soon as the word ‘man’ is uttered, immediately its outline also comes to
mind by means of preconception, the senses leading the way. Thus what pri-
marily underlies each name is something evident.
   197

And we would not inquire about the things we investigate if we had not had prior
knowledge of them. For example, ‘Is what’s standing over there a horse or a cow?’
For one must at some time have come to know the form of a horse and of a cow
by means of preconception. Nor would we have named something if we had not
previously learnt its outline by means of preconception.
(DL 10.33, tr. Long and Sedley)

Diogenes’ attempt to capture the sense of the word ‘preconception’ with many
different terms no doubt reflects the fact that Epicurus did not like to give
definitions; but it also incidentally shows that by the time of Diogenes Laertius,
a rich vocabulary was available to describe what he had in mind. General concepts
played a role in the theories of all the Hellenistic schools, though their origin and
epistemological status were seen in different ways. According to Cicero (ND 1.44),
the term prolepsis was introduced by Epicurus himself. The prefix pro- indicates
that these concepts must be grasped prior to something else, and this in two
different ways.
First, general concepts are associated with words and must be known before
one can understand or use the corresponding word. In this role, the preconcep-
tions do not function as criteria of truth, since general terms are used in true and
false statements alike. In fact, these concepts presumably furnish the mind with
the memories it sometimes uses to modify the content of a sensation to arrive at an
erroneous perceptual judgement.
The role of preconceptions in investigations, illustrated by Diogenes with the
rather trivial example of a horse or cow, is explained by Epicurus himself at the
beginning of the Letter to Herodotus:

First, then, Herodotus, we must grasp the things which underlie words, so that we
may have a reference point against which to judge matters of opinion, inquiry
and puzzlement, and not have everything undiscriminated for ourselves as we
attempt infinite chains of proofs, or have words which are empty. For the
primary concept corresponding to each word must be seen and need no add-
itional proof, if we are going to have a reference point for matters of inquiry,
puzzlement and opinion. (Ep.Hdt. 37–38, tr. Long and Sedley)

The preconceptions represent Epicurus’s solution to the notorious puzzle raised

in Plato’s Meno: how can you investigate something if you do not know at all what
it is? (Meno 80D). The answer consists in a distinction: we must indeed know what
it is that we are investigating in the sense of understanding the corresponding
word, but this is not the same as having an expert’s knowledge of its nature. This
understanding must be given at the start of an investigation and not be in need of
any proof, since such a requirement would involve us in an infinite regress, or
else we would be left with meaningless words. The suggestion that one
198    

must understand the terms one uses probably also comes from the Meno (see
75B–76A). But Epicurus adds that this pre-existing knowledge must also serve as a
reference point for the results of an investigation. Here he is introducing a label for
a practice that was (and is) common among philosophers: our conception of a
thing sets adequacy conditions for possible accounts of its nature. A theory of the
soul, for example, must explain how it enables animals to move and to perceive,
and will be rejected if it implies consequences incompatible with our initial
assumptions. In this sense, then, the preconceptions function as criteria.
Aristotle counts such presuppositions among what he calls phainomena or
endoxa—generally accepted beliefs, including facts of observation. Epicurus
reserved the term phenomena (φαινόμενα) for perceptual observations. This is a
useful clarification by contrast to Aristotle, whose wide class of phenomena
included common opinions, empirical observations, and even the views of philo-
sophers. But the comparison with Aristotle may also cast some doubt on
Epicurus’s claim that the preconceptions must not only be accepted without
further proof, but also themselves be evident truths. The argument here seems
to be the same as in the case of sense-impressions: we have no higher standard to
which we could appeal to establish the truth of our preconceptions, and therefore
we must accept them all as true. Now, it is surely correct to insist that some of the
assumptions that underlie our use of general terms must be kept constant—it will
not do, for instance, to claim that cows are made by sculptors, or that numbers can
walk. However, this does not rule out the possibility that a scientific investigation
might end up modifying our initial conception. A trivial example of this would be
that whales, according to the zoologists, are mammals, not fish. Here one might
say that the common conception of fish is wider and less precise than the scientific
one. But it is probably no surprise that Epicurus’s most famous theological thesis,
according to which anger and concern about human affairs are incompatible with
the preconception of the gods as blessed and immortal beings, did not find many
adherents outside the Epicurean school, even though offering physical explan-
ations of phenomena like thunder and lightning might have had some effect
against superstitious beliefs in supernatural powers.
In the case of the sensations, Epicurus could eventually try to vindicate his
claim that they are all true by offering explanations of optical illusions or different
sensations of taste in terms of atomism. In cases of concepts like those of the gods
or of justice—concepts that obviously involve some contributions of the mind, not
just perceptual observation—this kind of explanation was not available. The
evidence provided by preconceptions, even if they are formed on the basis of
true perceptions, is more limited than Epicurus seems to have assumed.
The third kind of criterion, feelings of pleasure or pain, are described by
Diogenes as criteria of choice and avoidance rather than of truth. This is indeed
their role in Epicurean ethics, but several passages in the Letter to Herodotus
show that they also functioned as criteria of truth. They are usually mentioned
   199

alongside the sensations and presumably have the same role, except that they are
impressions of bodily states rather than of external objects. ‘Pleasure’ and ‘pain’
are used as generic terms, not only for physical affections, but also for positive or
negative emotions such as joy, fear, or anger. Diogenes’ error is understandable,
since pleasure and pain are primarily mentioned as showing what is good or bad
for a person, though Epicurus emphasizes that this does not mean that everything
pleasant is to be pursued or everything painful to be avoided. Like the sensations,
they do not reveal the nature of the states that cause them. What brings about the
pleasant life is not the enjoyment of luxuries, but ‘sober reasoning that seeks out
the causes of all choice and avoidance’ (Letter to Menoeceus (Ep.Men.) 132). The
greatest pleasure, according to Epicurus, consists in the absence of all pain and
distress, but it takes philosophy to find this out. In the context of Epicurus’s
philosophy of nature, the feelings can be subsumed under perception, as indeed
they seem to be in some later sources.³
Epicurus’s account of the criteria and their truth offers a sophisticated theory of
the relations between the two cognitive faculties, reason or thought (διάνοια) and
the senses. Sense perception is the foundation of all knowledge because it is the
only way we can come in contact with the world around us. But the information
furnished by the senses may seem confusing and even self-contradictory. This led
philosophers to the argument from conflicting appearances, and to the contempt,
expressed by many of Epicurus’s predecessors, for perceptual beliefs. Democritus
had insisted on the fundamental role of the senses, but he had also apparently
accepted the conclusion that they offer conflicting information, and hence cannot
reveal the truth. Epicurus’s response was different: though he agreed that we have
no higher faculty that could distinguish between true and false sense-impressions,
he distinguished—like Plato—between sensations and perceptual judgements and
argued that we should recognize the limitations of the senses. We need to use
reason to organize the information we receive through the senses, and also to
understand the way the sense organs are functioning. Memory is needed for the
formation of general concepts and the development of language to communicate
what we experience. Concepts are formed by the mind from a multitude of sense-
impressions, usually from several senses. Even a simple concept like that of a cow
or horse includes not only the visual appearance of those animals, but also the
facts that they move, eat grass, make characteristic noises, and so on. These
concepts are not separately existing abstract objects like Platonic Forms, accessible
only to thought. Reason, therefore, remains dependent on the senses both for the
formation of its concepts and for observations of the natural world. But, as Lucretius
puts it, ‘the eyes cannot take cognizance of the real nature of things’ (DRN 4.385).

³ See, for example, Cicero, Fin. 1.30–31.

200    

The discovery of the causes and explanations of natural phenomena—in particular,

of course, the concepts of atoms and void, the elements of atomism that cannot be
perceived by the senses—is the domain of the mind.
This theory itself is independent of atomism—as should be expected if it was to
serve as the epistemological foundation of Epicurus’s philosophy of nature.
However, at this point we must take a look at a rival interpretation of Epicurus’s
thesis about the truth of sense-impressions that came up already in antiquity. In
Sextus Empiricus’s report on Epicurus’s epistemology (Adversus Mathematicos
(M) 7.203–216) we find the following:

But some are deceived by the difference between the impressions which seem to
be derived from the same object of sense—for instance a visible object—because
of which the object appears of another colour or of another shape, or altered in
some other way. For they have supposed that of the impressions thus differing
and conflicting, one of them must be true and the opposing one false. This is silly,
characteristic of people who do not understand the nature of things. For it is not
the whole solid body that is seen—to base our argument on objects of sight—but
the colour of the solid body. And of the colour, some is on the solid body itself, as
when one sees things from close up or from a moderate distance, and some is
outside the solid body and exists in the adjacent spaces, as in the case of things
seen from a great distance. And this is altered in the intervening space and takes
on its own shape, which produces an impression corresponding to what it really
is like.
And just as neither the sound in the brass instrument that is struck, nor the
sound in the mouth of the man who shouts, is what we hear, but the sound that
reaches our sense; and just as no one says that the person who hears a faint sound
from a distance is mishearing because, on coming close, the same sound is
perceived as louder—so I would not say that vision is deceived because it sees
the tower as small and round from a great distance, but from close at hand as
larger and square. Rather, I would say that it is telling the truth because when the
object of sense appears to it small and of that shape, it really is small and of that
shape, since the edges of the images are rubbed away as they travel through the
air; and again when it appears large and of a different shape, it is correspondingly
large and of a different shape, since it is no longer the same object that is both at
once. For it is left to the distorted opinion to believe that the object of vision seen
from close at hand is the same as that seen from a distance.
(M 7.206–209; tr. Bury with modifications)

This argument is obviously based on Epicurus’ theory of perception. He held that

the perceptions of the distance senses—vision, hearing, and smell—are caused by
atomic films or emanations constantly streaming from the surfaces of things. In
   201

the case of vision, these ‘images’ (εἴδωλα) may be damaged or distorted when
coming from a longer distance, and hence produce impressions different from
those received close at hand. In the last lines of Sextus’s report, it turns out that the
objects of vision are taken to be those images themselves rather than the objects
from which they come, and the truth of sense-impressions is thus guaranteed by
their correspondence to the images. This might look like a shortcut to truth, but,
of course, it means that what we perceive is not the external object we naturally
take it to be.
Contrast this with Lucretius’s explicit statement at DRN 4,256–258: ‘In this
connection, you should not consider it strange that, although the images that
impinge on our eyes are not visible, the objects themselves are seen.’ In other
words, the images are the means of perception, not its objects. Lucretius, the
faithful Epicurean, is more likely than Sextus to report Epicurus’s doctrine accur-
ately, and a passage in Plutarch’s treatise Against Colotes (adv.Col.) 1121B–D) may
actually show how the rival interpretation arose.
Plutarch claims that Colotes was unwilling to accept the consequences of the
view that all sense-impressions are true—namely the doctrine of the Cyrenaics,
according to which we can only perceive our own affections (πάθη): ‘For those
who say that when we encounter a round image, or another one that is bent, the
sense receives a true impression, but who will not let it declare as well that the
tower is round or the oar is bent, insist on their own affections and appearances,
but will not agree that the external objects are like that . . . ’ Imagining a dialogue,
Plutarch then has an Epicurean say ‘By Zeus, but when I come closer to the tower
or touch the oar, I will declare that the oar is straight, the tower angular, whereas
this man [sc. the Cyrenaic] will only agree that it seems and appears so, even when
he gets close.’ Plutarch replies: ‘By Zeus indeed—for he is better at seeing and
preserving the consequences of the claim that every sense-impression alike is
trustworthy in itself, and none more than any other, but to an equal degree.’
Plutarch obviously thinks that the Epicurean is taking the impression received
near the tower to be more reliable than the one received from a distance, and
hence goes against his own principle. But this is to misunderstand the Epicurean
view about the limitations of sense perception. The impression one receives from
afar is vague—as Lucretius puts it, ‘they [the towers] do not look like objects close
at hand that are really round, but resemble them in a shadowy fashion’ (DRN
4.362–363), and so an Epicurean would refrain from making a judgement about
the exact shape of the tower while he can only see it from a distance. That is the
point of the distinction between what is already present and what is still awaited.
One might perhaps say that one sees a tower in the distance, but that one cannot
yet discern its shape. Lucretius admits that it may be difficult to observe the
distinction, and so people will sometimes make mistakes, but the error lies in the
judgement of the mind, not the senses: ‘nothing is more difficult than to separate
202    

patent facts from the dubious opinions that our mind at once adds of its own
accord’ (4.466–467).
It is understandable that superficial or hostile readers of Epicurus would take
his theory of perception as evidence in support of the truth of sense-impressions,
but this reverses the order of the argument. Epicurus argues for the existence of
the images that reach the eye in order to explain the functioning of the sense
organs (see e.g. Ep.Hdt. 46–48), and he assumes that those images may be
damaged or distorted en route in order to explain why things look different
when observed from different distances. These differences between impressions
are among the phenomena a scientific theory must seek to explain—as science
does to this day, though not in Epicurean terms.
It is better, then, to side with the faithful Epicurean Lucretius. Epicurus’s theory
can be best understood as a response to the problems raised by his predecessors
who seemed to despair of the senses as a source of knowledge. And Epicurus, in
fact, agrees with one of the less pessimistic statements ascribed to Democritus.
Here is Sextus quoting from Democritus’s Canons:

These are his words:

Of knowledge (γνώμη) there are two forms, the one legitimate, the other bastard;
and to the bastard belong all these: sight, hearing, smell, taste, touch. And the
other is legitimate, and separated from that.
Then preferring the legitimate to the bastard, he continues:
When the bastard can no longer see anything smaller, or hear, or smell, or taste, or
perceive by touch . . . (*but more fine*)
Thus according to him too, reason, which he calls legitimate knowledge, is a
criterion.
(M 7.139, Democr. fr. B 11; tr. Barnes)

The text is corrupt at the end of the second quotation, but it seems clear that
Democritus went on to say that reason, the legitimate form of knowledge, must
take over, for example, to discover that the world is made up of atoms and void.
We do not know whether Democritus had found a way to answer his own
reproach of the senses I quoted at the beginning of this chapter, but Epicurus, as
I have tried to show, did so.
Epicurus’s Canon was intended as an introduction to the science of nature, and
he was surely right to insist that perceptual observation is indispensable. Even
though it might be difficult to avoid perceptual errors in individual cases, the
‘manifest facts’ (φαινόμενα) to which he appealed in his physics were not individ-
ual sense-impressions. One does not need to look for a special type of sense-
impression that is infallibly correct, as the Stoics did; it is enough, as Epicurus says
many times, to pay attention to all the available evidence.
   203

Scientific Method

At the beginning of the Letter to Herodotus, Epicurus reminds Herodotus of the

rules to keep in mind for a scientific inquiry: first, it must be clear ‘what underlies
words’, that is, one must have grasped the preconceptions that belong to the
subject under inquiry; this was discussed above. The next rule indicates how the
evidence furnished by the cognitive faculties must be used to construct opinions or
theories about things not accessible to direct observation:

Second, we should observe everything in accordance with the sensations and in

general the present applications of the mind or of any of the discriminatory
faculties, and likewise also in accordance with the feelings which exist in us, in
order to have a basis for sign-inferences about what is still awaited and about the
non-evident.
(Ep. Hdt. 38, following the passage about preconceptions quoted above;
emphasis added)

It is characteristic of passages like this one that Epicurus speaks of the evidence
coming from all the available sources of evident truths. Here we also find the
‘applications of the mind’ that the later Epicureans counted among the criteria
(DL 10.31). We do not have a definition of the phrase, but the contexts in which it
appears suggest that what is meant is a conscious attempt of the mind to arrive at
accurate information—in other words, careful observation as opposed to casual
looking or listening that might lead to errors. So Epicurus explains at Ep. Hdt. 51
that ‘error would not exist if we did not have in ourselves another movement
connected with the application to an impression, but distinct from it’. Since the
evidence obtained by an application of the mind still comes through the senses,
Epicurus himself probably did not treat it as an additional criterion.
The ways of testing the truth or falsity of opinions are briefly described in the
next sentence: ‘by this movement, falsehood arises if it is not attested or contested,
but when it is attested or not contested, the truth.’ A more detailed exposition
appears in Sextus Empiricus:

Of opinions, then, according to Epicurus, some are true, some false. True are
those attested and those uncontested by what is evident. Attestation is perception
through an evident impression of the fact that the object of opinion is such as it
was believed to be. For example, if Plato is approaching from far off, I form the
conjectural opinion, owing to the distance, that he is Plato. But when he has come
close, there is further testimony that he is Plato, now that the gap is reduced, and
it is attested by the evident itself.
Non-contestation is for the non-evident thing posited and believed to follow from
the evident appearance. For example, Epicurus, in saying that there is void, which
204    

is non-evident, confirms this through the evident fact of motion. For if void does
not exist, there ought not to be motion either, since the moving body would lack
a place to pass into as a result of everything’s being full and solid. Therefore the
non-evident thing believed is uncontested by the evident appearance, since there
is motion.
Contestation, on the other hand, is something which conflicts with non-
contestation. For it is the elimination of the evident appearance by the positing
of the non-evident thing. For example, the Stoics say that void does not exist,
judging something non-evident; but once this is posited about it, the evident
appearance, namely motion, ought to be co-eliminated with it. For if void does
not exist, necessarily motion does not occur either, according to the method
already demonstrated.
Likewise, too, non-attestation is opposed to attestation, being confrontation
through what is evident of the fact that the object of opinion is not such as it
was believed to be. For example, if someone is approaching from far off, we
conjecture, owing to the distance, that he is Plato. But when the gap is reduced,
we recognize through that which is evident that it is not Plato. That is what non-
attestation is like: the thing believed was not attested by the evident appearance.
Hence attestation and non-contestation are the criterion of something’s being
true, while non-attestation and contestation are the criterion of it’s being false.
And that which is evident is the foundation and basis of everything.
(M 7.211–216; tr. Long and Sedley with modifications, emphasis added)

Sextus is clearly not quoting from Epicurus directly, and his report is influenced
by later discussions,⁴ but it offers at least some examples to illustrate the method.
For purposes of discussion, it is easier to organize the text by types of opinion
rather than by truth or falsehood.
Judgements or opinions about observable objects can be accepted as true if they
are ‘attested’ by perception, false if they are ‘not attested’. This sounds odd, for
why should a belief for which one does not have confirmatory evidence count as
false? After all, there will inevitably be many things or events that cannot be
revisited to obtain additional evidence—for instance, the end of a horse race, or a
bird flying past. But Sextus’s example shows what appears to be intended: opin-
ions about observable objects must be rejected as false if subsequent perception
shows that things are not as they were believed to be. Epicurus’s method serves to
verify or falsify perceptual judgements, and so beliefs about things or events that
cannot be so verified fall outside the scope of the method. This is understandable,
since the inquirer into nature is not primarily interested in particular facts or

⁴ For a more detailed discussion of the anachronisms in this passage see D.N.Sedley, ‘On Signs’ in J.
Barnes et al. (ed.), Science and Speculation, Cambridge/Paris 1982, 239–272.
   205

events, but in types of things or events that occur regularly and can be perceived
by many observers, such as the waxing and waning of the moon, or indeed the
appearance of oars in and out of water. These are what Epicurus calls the
phenomena or ‘common perceptions’ (Ep.Hdt. 82)—observable facts that are
confirmed by everybody’s repeated experience. It is these that form the evidential
basis for his theory: for example, the indisputable fact that there are bodies,
‘attested by the senses everywhere’ (Ep.Hdt. 39), and the equally public fact that
there is movement, are the phenomena to which Epicurus appeals for the funda-
mentals of atomism, atoms, and void.
For hypotheses about unobservable objects—things that are either too small to
be perceived, like atoms, or too distant to be carefully observed, such as the stars or
the phenomena of meteorology—direct confirmation or disconfirmation is
impossible, and so they must be tested through their relation to observable facts,
that is, contestation or non-contestation by the phenomena. An opinion about the
unobservable is false if it has an observable consequence that conflicts with an
observable phenomenon. The standard example here, also used by Sextus, is
Epicurus’s argument for the existence of void: ‘if there were not what we call
void or place or the untouchable, the bodies would not have anything through
which to move, as they are plainly seen to move’ (Ep.Hdt. 40). The movement of
bodies, therefore, is a sign of the existence of void.
A hypothesis can be accepted as true if it is not contested by the phenomena. As
in the case of false beliefs about perceptible objects, this condition might seem too
generous: lots of hypotheses, for example, about events in the distant past are
consistent with the phenomena, but that is no reason to accept them all as true.
However, once again, the scope of the method is narrower than the general term
‘opinion’ suggests. The method is intended to test hypotheses about the unob-
servable causes of phenomena, and as Epicurus’s practice in the letters shows,
those will count as conflicting with the phenomena if they either fail to explain
what they are supposed to explain or, in the case of celestial events and objects, if
they are inconsistent with the ways in which similar events or things that can be
closely observed are produced. The evidence gained through the senses thus shows
us what kinds of hypotheses are possible, and in this sense the phenomena form
the basis of sign-inferences for what we cannot directly observe. In some cases,
such as the existence of void, Epicurus maintains that there is only one possible
explanation, and hence only one way of agreeing with the phenomena. In the case
of celestial motions or meteorological phenomena such as thunder or lightning,
however, the phenomena around us indicate that there are several ways in which
events like lunar eclipses, the waxing and waning of the moon, or the noise in the
clouds that we hear as thunder, can come about, since similar effects can be
produced in more than one way. Each of the several hypotheses consistent with
the phenomena will then represent at least a possible explanation. Since we cannot
verify or falsify the hypotheses by direct observation, scientific method, according
206    

to Epicurus, requires that we list them all and accept them as possible—or perhaps
even as true, if not in our world, then in others.
Lucretius illustrates this with an example:

There are some phenomena for which it is not sufficient to state one cause: you
must mention several causes, though only one of these will be the true cause. Let
me illustrate this point. If you saw a lifeless human body lying at some distance,
you would naturally enumerate all the possible causes of death, to ensure that you
mentioned the one true cause. For you could not be certain that the victim had
perished by the sword, or by cold, or by disease, or maybe by poison. But we do
know that it is something of this kind that has occasioned the death. And the
same applies to numerous phenomena. (DRN 6.703–711)

By listing all the possible causes of death, we can be sure to have captured the
actual cause, though we cannot determine which one it was. In the case of
meteorological or celestial phenomena, Epicurus often admits that the list of
possible causes might be longer, but he insists that it is unscientific to name
only one cause, because the observable phenomena call for several possible ones.
Astronomers who try to settle on a particular cause are wasting their time—and
they also open the door to the possibility of divine intervention, which makes
them lapse into mythology (Ep.Pyth. 113). The gods, of course, are excluded by
Epicurus’s theological doctrine—but also, I think, because Epicurus assumed that
there must be an observable model or analogue for any acceptable explanation,
and divine intervention has never been observed. The claim that hypotheses must
be based on observable models is not trivial, since it sets further limits on possible
theories, as can be seen in Epicurus’s polemic against other theories that do not
involve supernatural agents. For example, here is Epicurus’s argument against the
claim that the soul is incorporeal:

There is the further point to be considered, that according to the most common
usage we apply the term ‘incorporeal’ to that which can be conceived as existing
by itself. But it is impossible to conceive anything that is incorporeal except the
void; and the void cannot act or be acted upon, but merely allows bodies to move
through it. Hence those who call the soul incorporeal are silly, for if it were so, it
could neither act nor be acted upon. But as it is, both these properties evidently
belong to the soul. (Ep.Hdt. 67)

Aristotle, for one, certainly thought that his theory of the soul agreed with the
observable phenomena—this was after all a requirement of his own philosophical
method. But for Epicurus, the only conceivable incorporeal entity was the void.
Anything that could affect or move a body had to be a body itself, and so he
concluded that the soul must be a body consisting of extremely fine particles that
   207

made it too fine to be directly perceived. Here it is the metaphysics of atomism,

not the phenomena, that rules out an alternative theory. Epicurus was confident
that the concepts we acquire through perceptual experience and observation of the
world around us must, in principle, be sufficient for understanding everything that
goes on in the universe: what is inconceivable cannot exist.

Epistemology and Ethics

What we have considered so far is a general outline of epistemology and a method

for natural science that is narrower in scope. Those of Epicurus’s arguments for
atomism that involve only a single way of agreement with the phenomena could
well be regarded as proofs, although Epicurus seems to avoid the term for his own
arguments. But they were treated as proofs in later discussions of inferences from
signs between Epicureans and Stoics.⁵ They are not demonstrations in the
Aristotelian sense, where the premises of a proof not only imply, but also explain
the conclusion. Epicurus’s arguments are what would now be called inferences to
the best explanation—a label that may be misleading because those inferences are
not formally valid. This does not mean that the arguments are bad or unaccept-
able. The claim that there must be empty space if there is to be motion may well be
true or at least well supported by the evidence, but the existence of empty space
cannot be logically deduced from the possibility of motion.
But sign-inferences are not the only kind of argument used in philosophy or
elsewhere, and Epicurus’s general account of the relations between reason and
perception need not be taken to be merely a preface to the methodology of atomist
physics. Inferences from signs occur in arguments from perceptual evidence and
so belong, for example, to the natural sciences or to trial cases. But what, for
instance, about arguments in ethics? Here we have very little information in the
remaining fragments of Epicurus’s work, and it seems likely that there was no
explicit discussion of the subject in the Canon. The only apparently technical term
that seems to refer to some kind of philosophical reasoning is epilogismos
(ἐπιλογισμός), a word that tends to be translated differently in different contexts.
In a study of its use, Schofield⁶ has suggested ‘appraisal’ or ‘assessment’ as a
general translation, and this seems plausible, since it does not imply any particular
method. Epicurus speaks of epilogismos once in a side remark about the study of
time in the letter to Herodotus (72–73). The argument seems to be that we do not

⁵ For those later discussions about sign-inferences, some of them preserved in a papyrus fragment of
Philodemus’s treatise On signs, see D.N. Sedley (op. cit.1982), J. Barnes, ‘Epicurean Signs’, Oxford
Studies in Ancient Philosophy suppl. vol., 1988, 91–134, and J. Allen, Inference from Signs, Oxford
University Press, 2001, study IV.
⁶ Schofield, M.‘Epilogismos: An Appraisal’, in M. Frede and G. Striker (eds.) Rationality in Greek
Thought. Oxford University Press 1996, 221–37.
208    

have a preconception of time as an independently existing object, and that we

should, therefore, consider the observable events in relation to which we speak of
much or little time. We would then realize that we think of time as a property of
days and nights and their parts, periods of motion and rest, and so on. This
insight, he says, does not require proof, but epilogismos—by which he means
perhaps a reflection on what we think of when we use phrases like ‘a long time’
and the like. One might call the reasoning involved consideration, reflection, or
assessment, but one would hardly call it a method.
A more revealing passage can be found, I think, in the letter to Menoeceus
(133). This passage is close to the end of the letter, after an exposition of the best
life according to Epicurean hedonism that concludes with a praise of practical
wisdom (φρόνησις). Then comes the triumphant rhetorical finale:

For who, do you think, is superior to the man who holds only pious beliefs about
the gods, who is forever free from the fear of death, and who has found out by
reasoning (ἐπιλελογισμένου) the end established by nature? He discerns that the
highest limit of goods is easy to fulfil and easy to acquire, that bad things have
their limit in being either short in duration or easy to bear . . . (Ep.Men. 133)

It seems clear that this person has not just thought about the goal of life, but
arrived at the correct (Epicurean) conception, and he has reached this insight by
epilogismos. I would suggest that the reasoning is sketched out in the preceding
pages—a summary of what was no doubt set out in more detail in Epicurus’s lost
treatise about the goal of life (Περὶ τέλους, mentioned at DL 10.28). It leads from
the basic premise that goodness and badness consist in pleasure and pain via a
consideration of the differences between necessary and unnecessary, natural and
unnatural desires to the conclusion that once all the natural and necessary desires
are fulfilled, all pain and distress are gone and one has reached a state where no
good is lacking—that is, happiness. This philosophical argument is one that owes
something to Plato (regarding the desires) and to Aristotle (regarding happiness
as the state where no good is lacking). One could also describe it as an assessment
of the facts of psychology and biology, though I suspect that epilogismos was
simply the term Epicurus used for any kind of philosophical reasoning that is not a
form of deduction or inference from signs. But one should note that the argu-
ment’s first and most important claim—that the good is pleasure—comes from
Epicurean epistemology: pleasure is the first thing we naturally recognize as good
by perception, and our notions of good and bad are derived from the feelings of
pleasure and pain. Pleasure and pain must, therefore, serve as our guides in choice
and avoidance. The criterion of the feelings thus guarantees that hedonism in
some version must be true; but we still need reason to understand the nature of
pleasure and desire, and what this means for the happy life.
   209

As in natural philosophy, so in ethics: the basic facts and concepts derive from
the senses, but philosophical argument and reflection must be used to arrive at the
true conception of happiness.
Epicurus has often been described as an empiricist—no doubt by contrast to
Plato, for whom the Forms were the true objects of scientific understanding. This
is plausible to the extent that Epicurus, like Democritus, clearly saw in sense
perception the necessary foundation for scientific knowledge of the world. But the
labels ‘empiricism’ and ‘rationalism’ did not yet exist in Epicurus’s time⁷—in fact,
they first appeared in the methodological debates of the Hellenistic doctors—and
neither Democritus nor Epicurus seems to have thought that all knowledge must
in the end go back to sense perception. Humans have two faculties for attaining
knowledge, and Epicurus, reasonably enough, followed Democritus in arguing
that they needed to use both.

⁷ See J. Allen, ‘Experience as a source and ground of theory in Epicureanism’, Apeiron 37, 2004,
89–106, and Asmis , op. cit. (2009).
 15
Mental Health and Moral Health
Moral Progress in Seneca’s Letters

This chapter examines the Stoic conception of moral guidance or education

as a kind of psychotherapy, exemplified by the moral progress of Seneca’s
friend Lucilius as described in the Epistulae Morales. The Stoic project of
moral reform goes beyond the more modest aims of modern psychotherapy,
but it can be quite plausibly described as requiring the assistance of others—
friends or teachers—much as the recovery from illness requires the assistance of
a doctor.

The Stoic conception of philosophical teaching as a kind of therapy of the soul has
received a good deal of attention in recent years, due not only to Martha
Nussbaum’s The Therapy of Desire,¹ but also to a revived interest in the role of
the emotions and their relation to moral character.² Because the Stoics regarded all
emotions as illnesses and made much of their methods designed to get rid of them
in order to attain the ideal state of freedom from passion (apatheia), most of the
discussion has focused on the treatment of the passions. I do not intend to
rehearse that subject here, but instead to draw attention to the later stages of
Stoic moral education, which begins when the worst excesses of the passions—
uncontrollable propensities to emotional outbursts such as fits of rage or incon-
solable grief—have been left behind. Philosophical therapy will no doubt have to
begin with the passions because they are the cause of the most violent and
dangerous diseases of the soul, according to the Stoics; but once a person has
come to realize this and begun to work towards her psychic health, there is still a
long and difficult way to go. One might say that such a person will have reached a
minimal level of sanity, but she is still far from the ideal, and not just because she is
in danger of relapsing. This second part of Stoic moral education, or spiritual

¹ M. Nussbaum, The Therapy of Desire, Princeton, NJ 1993.

² See M. Graver, Stoicism and Emotion, Chicago 2007, with extensive bibliography.

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
?DOI: 10.1093/oso/9780198868385.003.0015
      211

guidance as one might also call it, is the main subject of Seneca’s Letters to Lucilius
Epistulae Morales (ep. mor.), which present the reader with a case history, as it were,
of advanced Stoic therapy. Seneca makes it clear at the beginning of the series that
both Lucilius and he himself are past the worst dangers and temptations, partly no
doubt because they are both of an age—beyond fifty—where some of the passions,
at least, no longer constitute a perpetual threat. It seems to me that a look at this part
of the therapeutic programme might shed some light on an intriguing question
raised by the analogy between medical treatment and philosophical training: namely
the relation between mental and moral health (if there is such a thing).
Modern psychotherapists do not usually promise to improve the moral char-
acter of their patients. Though the notion of mental health is notoriously murky, it
seems safe to say that the aim of contemporary psychotherapy is the modest one of
making people able to function ‘normally’ in everyday life, where this includes,
say, the ability to work, have reasonably good relations with others, and enjoy
one’s life to the extent that it might be enjoyable. Still, even if mental health is
clearly not the same as moral virtue, there seem to be some connections. Think of
the case of certain criminals who are considered to be psychologically disturbed,
such as child abusers or wife beaters. It would seem that normal social functioning
includes a minimal sense of morality, and psychotherapists who deal with such
patients are at least also trying to improve their patients’ moral behaviour. Perhaps
one could suggest that mental health is a prerequisite for moral improvement. But
there seems to be at least some overlap between psychotherapy and moral
education, and so one might ask whether the Stoic—or more generally, the
Hellenistic—conception of ‘caring for the soul’ might still have a point.
I should begin by acknowledging that the modern distinction between mental
health and moral virtue is, of course, no innovation. The Stoic slogan ‘All fools are
madmen’ was a paradox, intended to shock rather than to be taken literally. So
Cicero remarks at Tusculan Disputations (Tusc.) IV 30 ‘there is a kind of mental
health that may belong also to fools, as when a mental disturbance is removed by
medical treatment and purgation’. The point of the Stoic slogan was that moral
depravity is just as bad as lunacy. But since the ancient notion of mental disturb-
ance was no doubt much narrower than ours, modern psychotherapists might still
find themselves in the same field as their ancient counterparts—witness, for
example, the contemporary notion of ‘emotional disturbance’. And while the
aims of modern psychotherapists are more modest than those of the ancients,
they are still similar in that ancient moralists as well as modern psychotherapists
are in the business of helping their patients to lead better or happier lives. The
main difference seems to lie in the fact that happiness is no longer supposed to
have much to do with morality. But with this, of course, most ancient philo-
sophers strongly disagreed. Perhaps a look at the Stoic ideal can also help us to
understand why they thought moral virtue would guarantee psychological well-
being as well.
212    

II

To demarcate the earlier stages of Stoic therapy, mainly concerned with the
extirpation of the passions, from the more advanced stages, it is useful to begin
with a distinction that Cicero introduces in Tusc. IV 29–32, and which shows that
the Stoics extended the analogy between the treatment of the body and the
treatment of the soul to cover not only the cure of diseases, but also the part
that Plato (in the Gorgias 464B–465D) called ‘gymnastics’. Doctors are needed to
treat disorders of the vital functions of the body; trainers will start with basically
healthy bodies and aim at strength and beauty. The Stoic therapist has both
functions: he will have to deal with outright ‘diseases’, mainly the passions, in
the first place, but once the patient has recognized these as the illnesses they
allegedly are and has gained some degree of control over them, his further
education will aim at strength and beauty as well:

As in the body a certain fitting shape of the limbs together with a certain lovely
coloring is called beauty, so in the mind we speak of beauty with respect to an
evenness of beliefs and judgments combined with constancy, firmness and
stability, such as follows virtue or comprises virtue itself. Likewise, strength of
soul resembling the strength and sinews and effectiveness of the body is also
described by similar terms. (Tusc. IV 31)

Virtue, then, is not just mental health, but inner beauty and strength as well. This
should come as no surprise, since the Greek word used for moral goodness in
Hellenistic times is precisely to kalon—beauty. But how exactly are we to under-
stand this notion of virtue as beauty of the soul?
Seneca offers some clues in ep. mor. 120, which deals with the question of how
we acquire the (correct Stoic) notion of the good. This notion, he tells us there,
comes from observation and analogy—namely, the analogy between body and soul:

I will tell you what this analogy is. We had come to know the health of the body:
from this we came to think that there is also a health of the mind. We had come
to know the strength of the body: from this we inferred that there is a vigor of the
mind as well. Certain actions that were kind, humane, courageous had amazed
us: we began to admire them as though they were perfect.³

³ Quae sit haec analogia dicam. Noveramus corporis sanitatem: ex hac cogitavimus esse aliquam et
animi. Noveramus vires corporis: ex his collegimus esse et animi robur. Aliqua benigna facta, aliqua
humana, aliqua fortia nos obstupefecerant: haec coepimus tamquam perfecta mirari (4–5).
Seneca probably took over the analogy between outer and inner beauty from the Stoic Panaetius;
cp. Cicero, Off. I 14: Itaque eorum ipsorum quae aspectu sentiuntur nullum aliud animal pulchritudi-
nem, venustatem, convenientiam partium sentit; quam similitudinem natura ratioque ab oculis ad
animum transferens multo etiam magis pulchritudinem constantiam ordinem in consiliis factisque
conservandam putat.
      213

Beauty appears in the picture after a longish portrait of the virtuous man. Here
Seneca says:

From what then did we come to understand virtue? It revealed itself to us by its
order, its grace and constancy, the harmony of all actions among themselves, and
a greatness that rises above everything else.⁴

This is a description of beauty, given that the notion of beauty, familiar at least
from Plato on, was indeed one of orderliness and symmetry.
Seneca has switched from the good to virtue because he can already assume, at
this stage in the correspondence, that these two are the same. Once we have
grasped the analogy between body and soul, it seems, we are prepared to observe
the inner beauty of the virtuous person. He then goes on to explain that this
impression is produced by a threefold consistency. There is, first, the fact that the
virtuous man never curses his fate, always accepts whatever happens without
complaint—so there is no conflict between his thoughts and desires and the course
of events ordained by Nature (12–13). Second, the virtuous man is unwavering in
his judgements—a consistency that is due to his unfailing adherence to the truth
(Vero tenor permanet, falsa non durant, 19). Third and last, the wise man is at one
with himself, he is the only one who is truly one person or who performs only a
single role: Magnam rem puta unum hominem agere. Praeter sapientem autem
nemo unum agit. (cf. Epictetus, Dissertationes IV 2.10). The wise man is exactly as
he wants to be (22).
In other words, there is consistency between inner and outer events, consist-
ency among what we would call beliefs due to their truth, and complete congru-
ence between the person’s moral standards and her actual performance.
The first kind of consistency is postulated, of course, on the basis of Stoic
metaphysics—their belief in the rationality and goodness of the world order. I take
it that this part is difficult to make sense of from a contemporary point of view,
given that hardly anyone, with the possible exception of some very religious
people, would be inclined to share the Stoics’ metaphysical or cosmological
convictions. The second and third kinds are more interesting. With regard to
the second, we have to keep in mind that the judgements or beliefs in question will
include value judgements and moral principles, and that the Stoics assumed that
desires are to be identified with value judgements. The inner harmony and peace
that is supposed to go with the sage’s infallibly correct judgement is due to the

(No other animal, therefore, perceives the beauty, the loveliness, and the harmony of parts, of the
things that sight perceives. Our rational nature transfers this by analogy from the eyes to the mind,
thinking that beauty, constancy and order should be preserved, and much more so, in one’s decisions
and in one’s deeds. Tr. E.M. Atkins, with minor modifications).
⁴ Ex quo ergo virtutem intelleximus? ostendit illam nobis ordo eius et decor et constantia et omnium
inter se actionum concordantia et magnitudo super omnia efferens sese (11).
214    

absence of conflicting aims as well as factual beliefs. Since a Stoic would be firmly
convinced that only moral goodness has any real value at all, he would never be
inclined to go against his moral principles, and the Stoics evidently thought that
this would rule out all inner conflicts. We might find this over-optimistic—there
may be situations in which even moral obligations do clash and there is no clear
way of making the right decision. Nonetheless, it is plausible to think that a high
degree of consistency in one’s value judgements and moral beliefs would be a
desirable thing indeed, saving one from agonizing doubts or guilty feelings. It is
also plausible to think that a person who is not plagued by doubts and indecision
might be liked and admired by many others—assuming, of course, that her beliefs
and principles are reasonable and well founded.
The third kind of consistency may be the hardest to achieve, but is probably
crucial for the Stoic claim that virtue will make one happy. Perhaps the best
contemporary term for this inner state might be ‘integrity’—if by this one means a
character that will not admit the all-too-common compromises between one’s
considered judgements and one’s actual behaviour. Notoriously, people are apt to
judge others more harshly than themselves, and when challenged on this point,
can be very ingenious in finding excuses to defend themselves against the charge
of having broken the rules they expect others to follow. A person of integrity
might be one who will not do this—either because she does in fact always behave
in the way she thinks is right, or because she is willing to admit her mistakes. The
unity ascribed to the Stoic sage is obviously of the first sort—he is supposed to be
above all temptations, unhesitatingly and confidently following his rational judge-
ment. I assume that the unity Seneca has in mind here is distinct from the previous
sort of consistency because it is possible that a person might be entirely consistent
in judgement as well as in action, yet find it difficult to abide faithfully by her own
standards (cp. Chrysippus, Stoicorum Veterum Fragmenta (SVF) III 510). The
ideal Stoic, by contrast, will never feel tempted to stray from the path of virtue; he
is sure of himself, calm and serene, and, according to Seneca at least, even joyful in
the thought that no evil can befall him.
This, then, is the state of character and mind that Seneca and his friend aspire
to. Its beauty lies in its harmony and regularity, its strength in its constancy; and
both of these can only be achieved when all illnesses and disturbances of the
passions have been removed.

III

Let me now turn to ep. mor. 75, the letter in which Seneca sets out a Stoic ranking
of the last stages of moral progress before full virtue. Seneca’s description
(75.9–14) starts from the top—the class of people who are closest to wisdom but
do not yet fully possess it. They have already left behind all vices and passions, and
      215

they have learned what is really to be embraced; they are also beyond the danger of
backsliding. All they lack is self-confidence—as Seneca puts it, ‘they do not know
that they know’ (scire se nesciunt). This is because they have not yet tested their
own strength sufficiently. Only experience, it seems, can show them that they are
actually beyond all danger. Before we return to this point, let me look at the other
stages.
On the second level are those who no longer suffer from any ‘diseases’
(ingrained emotional dispositions that have become part of one’s character),
though there are still some episodes of affection. A person may, for example,
occasionally get angry but will no longer be irascible, or she may feel fear of death
or of pain, but she will no longer be obsessed with it, and so on. Since there are
these episodes, the danger of relapsing is still present, and the person is still on the
slippery slope (in lubrico) where one episode may lead on to the next and thus
reintroduce the emotional dispositions the Stoics called diseases. Clearly, these
people are below the level of the first group.
The third group consists of those who have shed some but not all of their
emotional diseases, and who are obviously also still liable to occurrent emotions,
presumably of all kinds. This third rank, Seneca says, is probably the best he and
Lucilius can hope for, and it already requires considerable effort. There may be
some hope for them to advance even further, but Seneca fears that they will be
held back by their vices, which make them treat their efforts at moral improve-
ment as a kind of leisure time occupation (quantum vacat, 16). Yet the rewards, if
they should manage to overcome their laziness, so to speak, are truly
magnificent—and Seneca concludes with one of his many vignettes of the blessed
state of the Stoic sage.
The three ranks I have just described are I think the ones Seneca wants to
distinguish, but his presentation is somewhat confused because it emerges in §10
that some Stoics did not recognize Seneca’s first stage—sages who do not yet know
they are sages—and hence described the level closest to perfection as the one
Seneca wants to rank second.⁵
It is easy to see why some Stoics would have refused to recognize Seneca’s first
rank as a stage of progress, for people at this stage actually are already sages,
whether they know it or not, so why count them still among those who are making
progress? ‘Nobody is beyond the danger of badness unless he has shed it entirely;
but nobody has shed badness unless he has replaced it by wisdom’, as Seneca

⁵ The side-remark about the alternative view of the first stage leads him to offer a brief explanation
of the difference between what the Stoics called ‘diseases’ (morbi)—ingrained emotional dispositions—
and ‘affects’ (adfectus)—episodes of emotion that he claims to have often mentioned before. In fact, the
relevant passage only comes later (ep. 85.10ff.), and I suppose this paragraph was inserted here before
Seneca’s own succinct description of the second rank in s.13. What has happened, I think, is that he has
combined a three-stage schema with one that has only two, lacking his own first rank; but then the first
rank of the other schema should simply count as his own second.
216    

reports their view (10). So these people would put Seneca’s first rank on the side of
wisdom, but they would probably not deny that there is a period during which a
person who has reached wisdom would not yet be aware of it, since the claim that
the transition from folly to wisdom will not be noticed by the sage-to-be was a
well-known—and sometimes ridiculed—bit of Stoic doctrine (see SVF III
539–542). The doctrine itself is curious and may well have been introduced in
response to some epistemological puzzle raised by Diodorus Cronus, as David
Sedley has suggested.⁶ However, what interests me at this point is not why the
Stoics held this view, but what Seneca makes of it. It seems fairly clear from the few
brief passages that report the Stoic view that the Stoics did not think of the
transitional period as being very long, so as to count as a distinct stage on the
way to the ultimate end. Their point was probably just that even the sage would
not be able to discern exactly the moment in which he makes the transition to
complete wisdom—after all, it would be the step from making no mistakes to
being infallible. More importantly, one would expect the aspiring sage to be more
concerned with practising virtue than with observing his own progress at every
moment. I suspect that the reason why Seneca wants to see the transitional period
as a stage before the final end is that a person who had not yet realized that he had
become a sage would also not yet enjoy what Seneca elsewhere (de vita beata
3.3–4; 4.4–5; 15.2) describes as the ‘supervenient’ consequences of wisdom—
tranquillity, security, and a sense of freedom. At the beginning of the correspond-
ence with Lucilius (ep. mor. 9.21), Seneca quotes with approval, as expressing a
kind of natural common notion, the line of an unknown comic poet:
He is not happy who does not believe he is (non est beatus esse se qui non putat).
The person who thinks he lacks something cannot count as being content with his
life, even if he is mistaken in his belief. A sage who does not yet realize that he is
one could not be certain that he has achieved wisdom,⁷ and this would prevent
him from enjoying the tranquillity and the sense of liberation that come with the
thought that one has left all dangers behind. Knowing that you have reached the
goal is crucial for happiness. And so while Seneca sometimes emphasizes the point
that joy and tranquillity are consequences, not constituents of the end, which is
simply living in agreement with nature, he also often identifies the happy life itself
with the inner state that results from perfect virtue. This is not a deviation from
orthodox Stoicism, I think. The Stoics would insist on the distinction between
virtue and its psychological consequences in order to make the point that virtue
must be desired for its own sake, not for the pleasures it brings—this is the

⁶ See his ‘Diodorus Cronus and Hellenistic Philosophy’, Proceedings of the Cambridge Philological
Society, N.S. 23, 1977, 94–5.
⁷ See the descriptions in Plutarch, mor. progr. 75D (SVF III 539): ἀγνοεῖ καὶ ἀμφιδοξεῖ and Stoic.
repugn. 1043A: τὸν ἐκ φαύλου γενόμενον σπουδαῖον . . . τῆς ἀρετῆς μὴ αἰσθάνεσθαι παρούσης ἀλλ’ οἴεσθαι
τὴν κακίαν αὑτῷ παρεῖναι.
      217

argument Seneca makes in the De vita beata. But once virtue has been achieved, it
cannot fail to bring the joy and security that make happiness complete.
The picture that Seneca offers here gives a fairly clear account of the path of
moral reform that leads from the initial recognition of one’s wretched and corrupt
state to the supreme self-assurance and serenity of the Stoic sage. It is strikingly
different from the better-known Aristotelian account of moral education which
begins in childhood and leads, if all goes well, to a stage of maturity at which the
individual has acquired a solid moral character and feels secure and competent in
the exercise of his rational capacities in virtuous action. The Stoic story requires a
break in a person’s development that can only come at the adult stage. It seems
inevitable, according to the Stoics, that a child should initially develop into a
creature of more or less uncontrolled emotion, due mainly to the corrupting
influence of parents and family (the very people whom Aristotle would expect
to watch over the natural and orderly development to virtue). Hence the Stoics
thought that a conversion would be needed to set one on the path to virtue. This is,
of course, why Stoic spiritual guidance is presented as analogous to medical
treatment. Aristotle’s more optimistic picture, one might say, covers only the
part that is parallel to gymnastic training. But while Aristotle seems to think that if
the natural talents are given, and the external circumstances are favourable,
education to virtue will work fairly smoothly, the Stoics hold that even after the
initial conversion, striving for virtue is a perpetual struggle against the ever-
present dangers of the passions. It is no wonder that the Stoic theory should
have been so much more influential than the Aristotelian one among Christians
who believed in original sin. Nevertheless, the Stoic story also includes the steps
that lead to moral and intellectual perfection, and so can be compared not just to a
cure, but also to the training that produces strength and beauty in the body. By
treating this process as a case of recovery, however, the Stoics may be offering a
more insightful account of what is meant by such old adages as that virtue is hard
to attain. They emphasize the fact that one’s performance often falls short of one’s
own standards and expectations. Where Aristotle suggests that practising the right
way of acting, under the guidance of parents and teachers, will make morally good
conduct almost a matter of instinct, the Stoics ask us first to recognize our faults,
and then to be always on the alert against possible temptations. They harp upon
the distance between most people’s professed moral ideals and their everyday
ways of living—one of the reasons, no doubt, why the almost superhuman figure
of the Stoic sage plays so important a role in their teaching. Where Aristotle seems
to think that intelligent upper-class males should have no great difficulties in
becoming practically wise adults (phronimoi), the Stoics, being less elitist and
aristocratic, insist that it takes reflection and conscious effort as well as practice to
become morally good.
It may be that the Stoic Panaetius, an aristocrat and well-known admirer of
Aristotle, tried to strike a compromise here by focusing on the rules of conduct for
218    

good men in an ordinary and less exalted sense (Cicero, de Officiis (Off.) I 46; III
13–16) as opposed to the ideal of perfection represented by the sage. But he did
not, therefore, deviate from orthodox Stoic doctrine—the ideal, and hence also the
conception of the human good, remains perfect moral virtue alone. However,
those who might be regarded as having achieved that ideal now become heroes or
saints—rare exceptions from the ordinary run of people rather than the respect-
able good citizens Aristotle seems to envisage for his ‘polity’ (and Cicero for the
Roman senate). There certainly is a point to this move. Embracing the accepted
standards of the upper classes can easily lead to complacency, and this is one
danger of which later Stoics at any rate seem to be acutely aware. Epictetus cites as
a sign of moral progress that a man will never claim to be worth much or to
possess any knowledge; and he advises those who are on the path to virtue to keep
a watchful eye on themselves as though they were their own enemy lying in wait
(Encheiridion 48.2–3). Seneca’s letters are full of self-deprecating remarks, con-
stantly reminding himself and his friend of the long distance that lies between
them and the ideal.
As a result of the Stoics’ insistence on ordinary people’s deficiencies, their
conception of the human good turned into something that appears to be so far
beyond most people’s reach that one is tempted to write it off as a conception of
human happiness. The Stoics claimed that nothing short of perfection can make
us happy, but also, of course, that moral perfection would be sufficient to
guarantee happiness. If one looks at Seneca’s enthusiastic descriptions of the
inner state of the sage, one might be inclined to agree with him on this last
point. What Seneca imagines and describes with all his literary and poetic skills is
a state in which a person is completely at one with herself and willing to accept
whatever happens, including the worst misfortunes, in the belief that all that she is
responsible for, and all that really matters, is her own moral integrity, and that
nothing inflicted by others or simply happening by accident can deprive her of
this. It seems plausible to think that such a person would be free from ordinary
people’s daily anxieties and worries—she is not deeply attached to anything that
lies beyond her own control, but she is disposed to appreciate whatever small,
good things come along. (The dark side of this portrait, as I have argued else-
where,⁸ is the implied lack of attachment to other people, friends as well as
humankind in general). The best examples I can think of that might show how
this could be seen as a form of happiness are the Stoics’ own favourite, Socrates, on
the one hand, Christian saints and martyrs on the other. Moral heroes and saints,
indeed—but how plausible is it to claim that only such people can be happy?
The Stoics were given to arguing that wisdom and happiness can be achieved
only if a person firmly believes that virtue is the only good, so that no trade-off

⁸ In Following Nature (1991); repr. in Striker, Essays on Hellenistic Epistemology and Ethics,
Cambridge 1996, 278–80.
      219

between moral and other alleged goods is possible. I do not think one has to agree
with the Stoics on this point. As far as I can see, it should be sufficient if a person is
convinced that serious moral evil outweighs all other evils, so that she would hate
herself and be miserable if she committed a grave moral wrong. Moral virtue may
require difficult sacrifices, however—and that, more prosaically speaking, is pre-
sumably the reason why saints and heroes are rare exceptions. One may appre-
ciate the role that moral ideals can play in keeping one from becoming
complacent, but there is also something to be said for putting up with some
(not all) of one’s own weaknesses—for example, the ones one is prepared to
forgive or overlook in one’s own friends and everyday companions. This should
not be confused with complacency: to say that one has made a forgivable mistake
is not the same as saying that one should go on cheerfully doing more of the same,
rather than trying to avoid similar mistakes in the future. Taking this into account,
one might arrive at a more modest conception of a satisfactory human life while
still acknowledging that most of us could and should try to do better than we
actually do. I suppose this is why Seneca thinks that he and Lucilius might find it
quite an achievement already to have reached even the third rank on his scale of
progress⁹—but this would obviously be no reason for self-congratulation.

IV

Whether we accept the Stoic ideal of perfection and happiness or not, I think we
might agree that the Stoic account of moral reform adds some important features
to an account of moral education, features that are missing from Aristotle’s
optimistic portrait of education as well as from the pessimistic view that both
Plato and Aristotle take of the chances of progress for anyone below the level of a
chosen few. It also seems plausible to compare moral reform to something like
recovery from illness or, later on, gymnastic training, if one starts from the
assumption that the person who is trying to improve her moral character begins
with a kind of conversion that makes her see herself as weak and perhaps even
despicable.
Which brings me back to the analogy between mental and moral health from
which I began. We have seen how Stoic moral therapy goes beyond the mere
restoration of ‘normal’ mental and social functioning by aiming at a state of inner
harmony and contentment that may appear desirable indeed, but that also
probably lies beyond what most of us could ever hope to achieve. Stoic moral
therapy also has a much larger scope than any version of psychotherapy in that it

⁹ Compare Seneca’s defence of his modest achievement at vita beata 17.3–4: Exige itaque a me, non
ut optimis par sim, sed ut malis melior: hoc mihi satis est, cotidie aliquid ex vitiis meis demere et errores
meos obiurgare.
220    

includes the attempt to make people realize that they need treatment that we find,
for example, in many of Epictetus’s little sermons. So the Stoics are obviously
much more ambitious than modern psychotherapists, who hope at best to help
their patients to fit in with the rest of us, being no worse, but also no better off than
their fellow humans.
On the other hand, Stoic moral therapy in its first stage after the necessary
conversion appears to be curiously one-sided and limited in treating the emotions
as the single cause of all vices and disturbances. Modern psychotherapy recognizes
a host of other factors that may lead to psychic disorders, including some that have
grave moral consequences. Some patients such as so-called sociopaths seem, in
fact, to suffer from a lack of ordinary emotional capacities rather than excess. The
same is likely to be true of cold-blooded criminals like concentration-camp
doctors: while it might be possible to concoct a story according to which their
crimes are not due to a lack of human feeling, but only to some excessive
attachment to a misguided idea of science, or an ideology of racism or whatever,
this would seem to be an attempt to make the facts fit the theory. Other psycho-
logical disorders will nowadays be blamed on traumatic experiences such as the
death of a close relative, or abuse by parents—and so on. This makes it difficult
even to compare the Stoic account with the much more complicated one we would
be given now. If there is overlap, it must be limited to a few specific cases. But
perhaps it might make sense to compare the Stoic convert who is trying to reform
herself to a person who is depressed through lack of self-esteem. Modern psycho-
therapists would probably see their task as one of reconciling a patient with her
own shortcomings, making her feel better about herself by giving up exaggerated
expectations. This applies, for example, to the case of anorexic young women who
think they are fat and ugly, but also (more common among graduate students) to
people who fall into utter despair because they are not geniuses. On the Stoic side,
I suppose, this would be described, not as adjusting one’s standard to one’s
feeble performance, but as learning to accept things as they are when one cannot
change them, and realizing, perhaps, that the situation is not as bad as one might
have believed. When it comes to moral standards, though, it is at the very least
questionable whether a psychotherapist should try to make her patient feel better
about herself by being content with what may be a rather nasty character (not to
mention outright criminal impulses). The first adjustment could be described as
taking a more realistic view—it isn’t, after all, a disaster if you are not very slim, let
alone if you are not a genius; the second, however, might end up fostering
unjustified complacency and self-deception. But the idea of living up to certain
standards, both of society at large and of one’s own, seems to play a significant role
in both cases. So, I would suggest that even a modern psychotherapist may have to
observe a distinction between accepting one’s own limitations and accepting one’s
moral faults by making them look like excusable weaknesses. However, this is a
point where a modern psychotherapist might well decide that she is no longer
      221

responsible and be content with the avoidance of actual criminal wrongdoing. Yet
her patient might still feel the need for further help, being really contrite and
unhappy with the kind of person she is. It seems to me that something like Stoic
moral therapy might well be what such a patient would be looking for—and the
analogy with training would not be inappropriate in that it emphasizes the need
for guidance and support. Here is Seneca once again (ep. mor. 52.2–3):

No man by himself has sufficient strength to rise above [his folly]; he needs a
helping hand, and someone to extricate him. Epicurus remarks that certain men
have worked their way to truth without anyone’s assistance, carving out their
own passage . . . Again, there are others who need outside help, who will not
proceed unless someone leads the way, but who will follow faithfully . . . We
ourselves are not of that first class, either; we shall be well treated if we are
admitted into the second. Nor need you despise a man who can gain salvation
only with the assistance of another; the will to be saved means a great deal, too.
(tr. R.M. Gummere)¹⁰

The point where psychotherapy and moral education overlap, then, is perhaps
also the point where they were later separated, one becoming a branch of
medicine, the other of education, both having gone together for a long time in
the old role of the confessor or spiritual adviser. If this is correct, then the kind of
moral guidance the Stoics offered might legitimately be seen as a close ally or
complement of ordinary psychotherapy. It would be part of a more comprehen-
sive perspective on what is needed for a satisfactory human life. Mental health is
no doubt a prerequisite, but another important factor may well be living up to
one’s own standards, and the process that makes one able to do this may, though it
need not, take a form that is still quite illuminatingly compared to a kind of
therapy.¹¹

¹⁰ Nemo per se satis valet ut emergat; oportet manum aliquis porrigat, aliquis educat.
Quosdam ait Epicurus ad veritatem sine ullius adiutorio exisse, fecisse ipsos viam . . . quosdam indigere
ope aliena, non ituros si nemo praecesserit, sed bene secuturos . . . Nos ex illa prima nota non sumus; bene
nobiscum agitur, si in secundum recipimur. Ne hunc quidem contempseris hominem qui alieno beneficio
esse salvus potest: et hoc multum est, velle servari.
¹¹ Earlier versions of this chapter were read at the University of Siena, Cornell University, and as a
Keeling lecture in London, 2004. Since I have been working on other subjects in the following years,
I have not attempted to bring this essay up to date with respect to the many publications on Stoic ethics
and Stoic psychology of the intervening years. As far as I can see, it still raises a point that might be of
interest.
 Bibliography

Ackrill, J.L. Aristotle’s Categories and De Interpretatione. Oxford: Clarendon Press 1963
Alesse, F. Panezio di Rodi. Testimonianze, Elenchos XXVII. Naples 1997
Alexander of Aphrodisias. In Aristotelis Analyticorum Priorum Librum I Commentarium,
ed. M. Wallies, (CAG II, 1). Berlin 1883
Algra, K. ‘The Mechanism of Social Appropriation and its Role in Hellenistic Ethics’,
Oxford Studies in Ancient Philosophy xxv, 2003, 265–96
Allen, J. ‘Academic Probabilism and Stoic Epistemology’, Classical Quarterly XLIV, 1994,
85–113
Allen, J. Inference from Signs. Oxford University Press, 2001
Allen, J. ‘Experience as a Source and Ground of Theory in Epicureanism’, Apeiron 37, 2004,
89–106
Allen, J. ‘Aristotle on Disciplines of Argument: Rhetorik, Dialectic, Analytics’, Rhetorica 25,
2007, 87–108
Annas, J. ‘Cicero on Stoic Moral Philosophy and Private Property’, in M. Griffin and
J. Barnes (eds.) Philosophia Togata. Oxford University Press 1989
Annas, J. The Morality of Happiness. Oxford University Press 1993
Annas, J. ‘Ethics in Stoic Philosophy’, Phronesis 52, 2007, 58–87
Annas, J. and J. Barnes. Sextus Empiricus, Outlines of Scepticism. Cambridge University
Press 1994
Asmis, E. Epicurus’ Scientific Method. Ithaca 1984
Asmis, E. ‘Epicurean Empiricism’, in J. Warren (ed.) The Cambridge Companion to
Epicureanism. Cambridge University Press 2009
Badawi, A. La transmission de la philosophie grecque au monde arabe. Paris 1968
Balme, D.M. Aristotle’s De Partibus Animalium I and De Generatione Animalium I,
translated with notes by D.M.B. Oxford 1972
Barnes, J. ‘Aristotle’s Theory of Demonstration’, Phronesis 14, 1969, 123–50; revised edition
in J. Barnes, M. Schofield, and R. Sorabji (eds.) Articles on Aristotle I 1975, 65–87
Barnes, J. Aristotle’s Posterior Analytics, translated with notes by J.B. Oxford 1975
Barnes, J. The Presocratic Philosophers. London 1979
Barnes, J. ‘Proof and the Syllogism’, in E. Berti (ed.) Aristotle on Science: The ‘Posterior
Analytics’. Padua 1981, 17–59
Barnes, J. Sheep Have Four Legs, Proceedings of the World Congress on Aristotle vol. III.
Athens 1982, 113–19
Barnes, J. ‘Epicurean Signs’, Oxford Studies in Ancient Philosophy suppl. vol., 1988, 91–134
Barnes, J. ‘Antiochus of Ascalon’, in J. Barnes and M. Griffin (eds.) Philosophia Togata.
Oxford: Oxford University Press 1989, 51–96
Barnes, J. The Toils of Scepticism. Cambridge: Cambridge University Press 1990
Barnes, J. Aristotle’s Posterior Analytics, translated with notes. Oxford: Clarendon Press
1975; 2nd edition 1994
Barnes, J. ‘Logical Form and Logical Matter’, 1990; repr. in J. Barnes, Logical Matters.
Oxford: Oxford University Press 2012, 43–143
Barnes, J., M. Schofield, and R. Sorabji (eds.). Articles on Aristotle I. 1975, 65–87
246 

Barnes, J. et al. (ed.s), Science and Speculation, Cambridge University Press 1982
Becker, A. Die Aristotelische Theorie der Möglicheitsschlüsse. Junker, Berlin 1933
Bochenski, I.M. La Logique de Théophraste. Fribourg (Suisse) 1947
Bonitz, H. Index Aristotelicus. 2nd ed. Berlin 1870, repr. Graz, Akademische Druck-u.
Verlagsanstalt 1955
Brinton, A. ‘Pathos and the “Appeal to Emotion”: An Aristotelian Analysis’, History of
Philosophy Quarterly 5, 1988, 207–19
Broadie, S. Ethics with Aristotle. New York: Oxford University Press, 1991
Brunschwig, J. Aristote, Topiques, vol. I. Paris 1967
Brunschwig, J. ‘La forme prédicat de la matiere’, in P. Aubenque (ed.) Etudes sur la
Métaphysique d’Aristote, Paris 1979
Burnyeat, M. ‘The origins of non-deductive argument’, in J. Barnes et al. (eds.) Science and
Speculation. Cambridge/Paris: Cambridge University Press 1982, 193–238
Burnyeat, M. ‘Enthymeme: Aristotle on the Logic of Persuasion’, in D.J. Furley and
A. Nehamas (eds.) Aristotle’s Rhetoric. Princeton University Press 1994, 3–55
Burnyeat, M. ‘Enthymeme: Aristotle on the Rationality of Rhetoric’, in A.O. Rorty (ed.)
Essays on Aristotle’s Rhetoric. Oakland, CA: University of California Press 1996, 88–115
Burnyeat, M. and M. Frede. The Original Sceptics. Indianapolis, IN: Hackett, 1997.
Byrne, P.H. Analysis and Science in Aristotle. SUNY Press, New York 1997
Cooper, J.M. Reason and Human Good in Aristotle, Indianapolis: Hackett 1986)
Cooper, J.M. ‘Aristotle on Natural Teleology’, in M. Schofield and M.C. Nussbaum (eds.)
Language and Logos. Cambridge: Cambridge University Press 1982, 197–222
Cooper, J.M. ‘Hypothetical Necessity’, in A. Gotthelf (ed.) Aristotle on Nature and Living
Things. Pittsburgh, PA: Mathesis Publications, 1986, 151–67
Cooper, J.M. ‘Rhetoric, Dialectic, and the Passions’, Oxford Studies in Ancient Philosophy
11, 1993, 178–84
Corcilius, K. ‘Two Jobs for Aristotle’s Practical Syllogisms?’, Logical Analysis and History of
Philosophy 11, 2008, 163–84
Corcoran, J. ‘Completeness of an Ancient Logic’, Journal of Symbolic Logic 37, 1972, 696–702
Corcoran, J. (ed.). Ancient Logic and its Modern Interpretations. Dordrecht 1974
Corcoran, J. ‘Aristotle’s Natural Deduction System’, in J. Corcoran (ed.) Ancient Logic and
its Modern Interpretations. Dordrecht, D. Riedel 1974, 85–131
Crubellier, M. ‘The Programme of the Aristotelian Analytics’, in Dialogues, Logics and
Other Strange Things. London: College Publications, 2008, 121–47
Decleva Caizzi, F. ‘Aenesidemus and the Academy’, Classical Quarterly 1992, 176–89
Dyck, A.R. A commentary on Cicero’s De Officiis. Ann Arbor, MI: University of Michigan
Press, 1996
Ebbinghaus, K. Ein formales Modell der Syllogistik des Aristoteles, Hypomnemata 9.
Göttingen 1964
Fernandez, P. ‘Reasoning and the Unity of Aristotle’s Account of Animal Motion’, Oxford
Studies in Ancient Philosophy 47, 2014, 151–204
Fine, G. On Ideas. Oxford: Oxford University Press 1993
Fortenbaugh, W.W. ‘Aristotle’s Rhetoric on Emotions’, Archiv für Geschichte der Philosophie
52, 1970, 40–70
Fortenbaugh, W.W. Aristotle on Emotion. New York 1975
Frede, D. ‘Mixed Feelings in Aristotle’s Rhetoric’, in A.O. Rorty (ed.) Essays on Aristotle’s
Rhetoric. London: University of California Press, 1996, 258–85
Frede, M. ‘An Aristotelian Dilemma’, Ajatus 22, 1959, 87–92
Frede, M. ‘Stoic vs. Aristotelian Syllogistic’, Archiv f. Gesch. Phil. 56, 1974, 16ff. (repr. in
Frede 1987)
  247

Frede, M. ‘The Title, Unity, and Authenticity of the Aristotelian Categories’, (1983), repr. in
M. Frede, Essays in Ancient Philosophy. Minneapolis, MN: University of Minnesota
Press, 1987, 11–28
Frede, M. ‘Stoics and Skeptics on Clear and Distinct Impressions’, (1983), repr. in Frede
1987, 151–76
Frede, M. ‘The Empiricist Attitude towards Reason and Theory’, in Apeiron xxi, 1988.
Frede, M. and Striker G. (eds.) Rationality in Greek Thought. Oxford University Press1996
Gauthier, R.E. and J.Y. Jolif. L’Éthique a Nicomaque. 2nd ed. Louvain/Paris 1970
Geach, P.T. ‘History of the Corruptions of Logic’, in P.T. Geach, Logic Matters. Oxford:
Blackwell 1972, 44–61
Glucker, J. ‘Antiochus and The Late Academy’, Hypomnemata 56. Göttingen 1978
Gomperz, H. ‘Die deutsche Literatur über die . . . aristotelische Philosophie 1899 und 1900’,
Arch. Gesch. Phil 16, 1903, 274f
Gosling, J.C.B. and C.C.W. Taylor. The Greeks on Pleasure. Oxford University Press 1982
Goulet, R. (ed.). Dictionnaire des philosophes antiques. Paris 1989
Graver, M. Stoicism and Emotion. Chicago, IL: University of Chicago Press 2007
Griffin, M.T. and E.M. Atkins (eds.). Cicero: On Duties. Cambridge: Cambridge University
Press 1991
Hankinson, R.J. The Sceptics. London/New York 1995
Hintikka, J. ‘Aristotle’s Different Possibilities’, in J. Hintikka (ed.) Time and Necessity.
Oxford, Clarendon Press, 1973, 27–40
Hintikka, J., ‘An Aristotelian Dilemma’, Ajatus 22(1959), 87–92
Hirzel, R. Untersuchungen zu Cicero’s philosophischen Schriften, III. Leipzig 1883
Hume, D. Enquiry concerning the Principles of Morals, ed. Selby-Bigge. 2nd ed. Oxford
University Press 1902
Ioli, R., Ἀγωγή and related terms in Sextus Empiricus, MPhil thesis, Cambridge 1999
Ioppolo, A.M. Aristone di Chio e lo Stoicismo antico. Naples, Bibliopolis 1980
Ioppolo, A.M. Opinione e scienza. Naples, Bibliopolis 1986
Irwin, T. ‘Aristotle’s Conception of Morality’, Proc. of the Boston Coll. in Ancient Phil. I,
1986, 115–43
Irwin, T. ‘Stoic Naturalism and its Critics’, in B. Inwood (ed.) The Cambridge Companion to
the Stoics. Cambridge: Cambridge University Press 2003, 345–64
Joachim, H.H. Aristotle on Coming-to-be and Passing-away, text, introduction and com-
mentary by H.H.J. Oxford University Press 1922
Kassel, R. Aristotelis ars rhetorica. Berlin, de Gruyter 1976
Kenny, A. ‘Practical truth in Aristotle’, in B. Morison and K. Ierodiakonou (eds.) Episteme,
etc. Oxford University Press 2011, 277–84
Kneale, W. ‘Modality De Dicto and De Re’, in E. Nagel, P. Suppes, and A. Tarski (eds.)
Logic, Methodology and Philosophy of Science. Stanford University Press 1962, 622–33
Kneale, W. and M. The Development of Logic. Oxford University Press 1962
Kraut, R. ‘In Defense of the Grand End’, Ethics 103, 1993, 361–74
Lear, J. Aristotle and Logical Theory. Cambridge: Cambridge University Press 1980
Lee, T.S. Die griechische Tradition der aristotelischen Syllogistik in der Spätantike,
Hypomnemata 79. Göttingen 1984
Leighton, S. ‘Aristotle and the Emotions’, Phronesis 27, 1982, 144–74
Lévy, C. Cicero Academicus, Collection de l’Académie Française de Rome 162, 1992
Lloyd, A.C. Necessity and Essence in the Posterior Analytics, in E. Berti (ed.) Aristotle on
Science: The ‘Posterior Analytics’. Padua 1981, 157–71
Locke, J. An Essay Concerning Human Understanding, ed. P.H. Nidditch. Oxford:
Clarendon Press 1975
248 

Long, A.A. and D.N. Sedley. The Hellenistic Philosophers. 2 vols. Cambridge University
Press 1987
Long, A. A., Stoic Studies, Cambridge University Presss 1996
Maier, H. Die Syllogistik des Aristoteles, Bd. II 1. Tübingen 1900
Malink, M. ‘Aristotle on Principles as Elements’, Oxford Studies in Ancient Philosophy LIII,
2017, 163–213
Mansfeld, J. ‘Aenesidemus and the Academics’, in L. Ayres (ed.) The Passionate Intellect, Rutgers
University Studies in Classical Humanities vii. New Brunswick/London 1995, 235–48
McCall, S. Aristotle’s Modal Syllogisms. Amsterdam 1963
McDowell, J. ‘Deliberation and Moral Development in Aristotle’s Ethics’, in S. Engstrom
and J. Whiting (eds.) Aristotle, Kant, and the Stoics. Oxford University Press 1996, 19–35
Mendell, H. Making Sense of Aristetelian Demonstration,
Mignucci, M. Aristotele, Gli Analitici Primi a cura di M.M. Naples 1969
Mignucci, M. ‘ὡς ἐπὶ τὸ πολύ et nécessaire dans la conception Aristotélicienne de la science’,
in E. Berti (ed.) Aristotle on Science: The ‘Posterior Analytics’. Padua 1981, 173–203
Moraux, P. Der Aristotelismus bei den Griechen. Berlin 1973
Mueller, I. ‘Greek Mathematics and Greek Logic’, in J. Corcoran (ed.) Ancient Logic and its
Modern Interpretations. Dordrecht 1974
Nielsen, K.M. ‘Deliberation as Inquiry’, Philosophical Review 120, 2011, 383–421
Nortmann, U. ‘Über die Stärke der aristotelischen Modallogik’, Erkenntnis 32, 1990, 61–82.
Nussbaum, M. The Therapy of Desire. Princeton, NJ: Princeton University Press 1993
Pacius, I. In Porphyrii Isagogen et Aristotelis Organum Commentarius Analyticus. Frankfurt
1597 (repr. Hildesheim 1966)
Patterson, R. ‘The Case of the Two Barbaras’, Oxford Studies in Ancient Philosophy 7, 1989,
1–40
Patzig, G. Die aristotelische Syllogistik. 3rd ed. Göttingen: Vandenhoeck & Ruprecht 1969
Perin, C. ‘Substantial Universals in Aristotle’s Categories’, Oxford Studies in Ancient
Philosophy 33, 2007, 125–44
Philoponus, Ioannnes. In Aristotelis Analytica Posteriora commentaria, ed. M. Wallies,
(CAG XIII.2). Berlin 1905
Pohlenz, M. Τὸ πρέπον (1933), repr. in his Kleine Schriften 1, ed. H. Dörrie. Hildesheim
1965, 100–39
Rapp, C. Aristoteles: Rhetorik, Erster Halbband. Berlin: de Gruyter 2002
Robinson, R. Plato’s Earlier Dialectic. 2nd ed. Oxford University Press 1953
Ross, W.D. ‘Aristotle’s Prior and Posterior Analytics’, editior maior w. commentary. Oxford,
Clarendon Press 1949
Schäublin, C., A. Graeser, and A. Bächli. Marcus Tullius Cicero, Akademische
Abhandlungen: Lucullus. Hamburg: Meiner 1995
Schofield, M. The Stoic Idea of the City. Cambridge University Press 1991
Schofield, M. ‘Epilogismos: An Appraisal’, in M. Frede and G. Striker (eds.) Rationality in
Greek Thought. Oxford University Press 1996, 221–37
Sedley, D. ‘Diodorus Cronus and Hellenistic Philosophy’, Proceedings of the Cambridge
Philological Society, N.S. 23, 1977
Sedley, D. ‘The End of the Academy’, Phronesis XXVI, 1981, 67–75
Sedley, D. ‘On Signs’, in J. Barnes et al. (eds.) Science and Speculation. Cambridge/Paris,
Cambridge University Press 1982, 239–72
Sedley, D. ‘Sextus Empiricus and the Atomist Criteria of Truth’, in Elenchos 13, 1992, 19–56
Sherman, N. ‘The Role of Emotions in Aristotelian Virtue’, Proceedings of the Boston Area
Colloquium in Ancient Philosophy 9, 1993, 1–33
  249

Shorey, P. ‘ΣΥΛΛΟΓΙΣΜΟΙ ΕΞ ΥΠΟΘΕΣΕΩΣ in Aristotle’, American Journal of

Philology X, 1889
Sigwart, Ch. Beiträge zur Lehre vom hypothetischen Urteil. Tübingen 1871
Smiley, T. ‘What is a Syllogism?’, Journal of Philosophical Logic 2, 1973, 136–54
Smiley, T. ‘Aristotle’s Completeness Proof ’, Ancient Philosophy XIV, 1994, 25–38
Smith, M.F. Lucretius, On the nature of things. 2nd ed. Indianapolis, IN: Hackett 2001
Smith, R. ‘The Mathematical Origins of Aristotle’s Syllogistic’, Archive for the History of
Exact Sciences 19, 1978, 201–10
Smith, R. ‘Some Studies of Logical Transformations in the Prior Analytics’, History and
Philosophy of Logic 2, 1981, 1–9
Smith, R. ‘The Relationship of Aristotle’s two Analytics’, Classical Quarterly 32, 1982,
327–35
Smith, R. ‘Completeness of an Ecthetic Syllogistic’, Notre Dame Journal of Formal Logic 24,
1983, 224–32
Smith, R. ‘Aristotle as Proof Theorist’, Philosophia Naturalis 21, 1984, 590–7
Smith, R. Aristotle, Prior Analytics.Hackett, Indianapolis 1989
Smith, R. Aristotle, Topics I, VIII and Selections. Oxford: Clarendon Press 1997, 193–238
Solmsen, F. Die Entwicklung der aristotelischen Logik und Rhetorik. Berlin 1929
Sorabji, R. Necessity, Cause and Blame. Duckworth, London 1980
Striker, G. Review of J. Barnes, Aristotle’s Posterior Analytics. Oxford 1975; Zeitschrift für
philosophische Forschung 31, 1977, 317–8
Striker, G. review of H. Tarrant, Scepticism or Platonism?, in Ancient Philosophy 11, 1991,
202–6
Striker, G. κριτήριον τῆς ἀληθείας, (Orig. in German, 1974; English transl. in Striker, Essays
on Hellenistic Epistemology and Ethics. Cambridge University Press 1996, 22–76)
Striker, G. ‘The Problem of the Criterion’, in S. Everson (ed.) Epistemology (Companions to
Ancient Thought 1). Cambridge 1990, 143–60; also repr. in Striker, Essays 1996
Striker, G. Essays on Hellenistic Epistemology and Ethics. Cambridge: Cambridge University
Press 1996
Striker, G. Aristotle: Prior Analytics, Book I, transl. w. notes. Oxford: Clarendon Press 2009
Tarrant, H. Scepticism or Platonism? Cambridge: Cambridge University Press 1985
Thiel, N.M. Die Bedeutung des Wortes Hypothesis bei Aristoteles (Diss.). Freiburg 1919
Thom, P. ‘Termini Obliqui and the Logic of Relations’, Archiv f. Geschichte der Philosophie
59, 1977, 143–55
Thom, P. The Syllogism. Munich: Philosophia 1981
Thom, P. The Logic of Essentialism. Dordrecht: Reidel, 1996
von Arnim, H. Stoicorum Veterum Fragmenta III, 1903; reprint, Stuttgart 1964
van Rijen, J. Aspects of Aristotle’s Logic of Modalities, Synthese Historical Library 39.
Dordrecht: Reidel 1989
Van Straaten, M. Panétius, sa vie, ses écrits . . . Amsterdam 1946
Voelke, A.J. ‘Soigner par le logos: la thérapeutique de Sextus Empiricus’, in ‘Le Scepticisme
Antique’, Cahiers de la revue de théologie et de philosophie 15, 1990
Wedin, M. Aristotle’s Theory of Substance. Oxford: Oxford University Press 2000
Wesely, M. ΑΝΑΛΥΣΙΣ ΠΕΡΙ ΤΑ ΣΧΗΜΑΤΑ, Peitho 3, 2012, 83–113
Wieland, W. ‘Review of Patzig, Die aristotelische Syllogistik’, Philos. Rundschau 14, 1966,
24–5
Wolf, U. Möglichkeit und Notwendigkeit bei Aristoteles und heute. Munich 1979
 16
Panaetius Peri tou kathēkontos in
Cicero’s De Officiis

This chapter presents a study of Panaetius’s best-known innovation in Stoic

ethics—the notion of decorum in both a wider and a narrower sense. In its wider
sense, decorum appears on a list of four natural human starting points for virtue.
Based on the uniquely human sense of beauty and the analogy between visible
beauty and beauty of mind, it is the outer manifestation of virtue that leads
humans to recognize virtue as a good and to strive to become virtuous. In the
narrower sense, it includes moderation, the traditional fourth cardinal virtue, but
also, motivated by the human desire to win the approval of others, socially
pleasant behaviour. Decorum in this sense goes beyond morally correct conduct; it
may be a social virtue, but not a moral one.

Stoic Ethics without Metaphysics?

Cicero’s De Officiis (Off.) has not attracted much attention from philosophers in
the last few decades. The rediscovery of Hellenistic philosophy has been focused
mainly on the theories of the earlier period of each school, with the exception
perhaps of the Sceptics. For Stoic ethics, Cicero’s other book, De Finibus (Fin.),
contains much more information than the De Officiis, which is based on the work
of the second-century Stoic Panaetius On appropriate action (Περὶ τοῦ
καθήκοντος) and does not include an exposition of the general theory. And
while Panaetius used to be described in the 1920s and 1930s as an innovator or
perhaps even a heretic in his school, most scholars now agree that there is no good
reason to think that he diverged from the central tenets of Stoic ethics, or
presented something like a second-class moral doctrine for people who might
find the ideal of the Stoic Sage too demanding. Panaetius could hardly have been
the head of the Stoic school if he had been seen as a heretic. But in the Περὶ τοῦ
καθήκοντος he was, after all, writing about appropriate action—correct but not
necessarily virtuous, and such action is, according to the Stoics themselves,
common to both sages and fools, virtuous and vicious persons. In a work that
offers advice for everyday life, the theoretical foundations could be left largely in

From Aristotle to Cicero: Essays on Ancient Philosophy. Gisela Striker, Oxford University Press. © Gisela Striker 2022.
DOI: 10.1093/oso/9780198868385.003.0016
       ’    223

the background, as Cicero himself suggests (Off. I.7). Compared to books like the
Academica, De Finibus or even the Tusculans, Off. contains little philosophical
argument. Furthermore, the advice Panaetius offered was clearly intended for
upper-class gentlemen, sounding sometimes rather offensive to modern, more
egalitarian readers, at other times quaint or even hilarious, as when Cicero
mentions singing in the forum as an obvious impropriety (1.145). One would
not recommend this work to an aspiring politician today—although it was
apparently used in just this way in Europe until the eighteenth century.
On the other hand, given that Cicero explicitly asserts that he was mainly
following a single author and work in the first two books of Off., we probably have
more precise information about this work of Panaetius than about the books of
any other Stoic author before the first century —that is, before Seneca and
Epictetus. Panaetius was clearly much admired during his life and in the following
century. Though he may not have departed from Stoic orthodoxy in his ethics, he
did apparently introduce some innovations, and he is known to have shown some
independence in other fields, for example, rejecting the belief in divination, and
the notorious doctrine of cyclical conflagration and eternal return—perhaps
under the influence of Plato and Aristotle. So I thought it might be worthwhile
to take a closer look at Off. and comparing it, where possible, with the third book
of the Fin. and other sources on earlier Stoic ethics. I do not intend to engage in
Quellenforschung here, trying to distinguish fragments of Panaetius’s work from
additions by Cicero—this was done, with more or less success, by the philologists
of the earlier twentieth century and the editors of collections of testimonia about
Panaetius, and continued in the recent commentary by A. Dyck.¹ But I will assume
that the general structure of Cicero’s first two books, and their main theoretical or
programmatic passages, go back to Panaetius. It seems to me that a study of
Panaetius’s way of presenting an account of appropriate action reveals an inter-
esting and indeed probably innovative way of dealing with Stoic ethics that may
also explain some of the influence of Cicero’s book in early modern times.
One reason why Cicero chose Panaetius’s famous work as a model was no
doubt the fact that Panaetius wrote in elegant literary Greek, avoiding the awk-
wardness of the technical terminology that Cicero is at pains to explain in Fin.
III. We do not hear, for example, of the ‘selection’ of ‘preferred’ or ‘dispreferred’
‘indifferents’ as distinct from the choice of real goods; the Stoic sage is barely
mentioned, and then mainly as an unattainable ideal. Appropriate actions or
duties are classified under the rubrics of what is honourable or virtuous and
what is useful or expedient, using a common distinction that had probably not
been obliterated in spite of the efforts of philosophers from Socrates on. To the

¹ See M. Van Straaten, Panétius, sa vie, ses écrits, Amsterdam 1946; F. Alesse, Panezio di Rodi:
Testimonianze, Elenchos XXVII, Naples 1997; A.R. Dyck, A commentary on Cicero’s De Officiis, Ann
Arbor, MI 1996; J. Annas, ‘Ethics in Stoic Philosophy’, Phronesis LII, 2007, 58–87.
224    

dismay of some other Stoics, Panaetius had even promised to discuss cases where
the honourable and the useful would appear to conflict, though in fact he never
wrote this last book of his treatise. Orthodox Stoics protested that such conflicts
could not possibly arise, since the honourable and the useful must always coincide,
and Cicero spent some pages of his own last book confirming that Panaetius had
indeed announced his plan to write about those conflicts, and emphasizing that
the conflicts would be only apparent, due to the difficulties that ordinary non-wise
people might sometimes find in trying to decide about the right way to act.
It is a pity that we do not have Panaetius’s own introduction to his treatise.
What we find in Cicero is first, of course, a dedicatory epistle to his son, then a
brief statement emphasizing the importance of the subject to be discussed. Cicero
also mentions that he will be mainly following Stoic doctrine, though not without
using his own judgement, as usual. But then follows a section of rather pedantic
criticisms of Panaetius’s treatise—though neither the author nor the title of his
work had been mentioned before. Cicero complains that Panaetius did not begin
with a definition of appropriate action ‘as one should, in order for the reader to
understand what the subject is’ (I 7). He briefly states the Stoic definitions of both
right action and appropriate action (κατόρθωμα and καθῆκον) as ‘perfect’ and
‘middle’ duty (I 8), without trying to explain the reason for the distinction; then he
continues by criticizing Panaetius’s organization of the treatise for omitting some
sub-sections (I 10). I doubt that Cicero would have left this somewhat ill-tempered
set of notes as it is, had he found the time to revise his work and polish up its style.
At the very least he would have mentioned the author and the title of the work he
is following, and probably praised Panaetius for his elegant style rather than
complaining about his way of organizing the book.² All one can learn from this
section is that Panaetius saw his treatise as an aid to deliberation (I 9) and
organized it around the three main considerations that arise in making a decision:
one would consider first whether a proposed action is honourable or not, then
whether it is useful, and sometimes one would face difficulties when those criteria
appear to be in conflict.³ Panaetius’s (and Cicero’s) work was organized around
these three problems. A technical definition of appropriate as opposed to right
action might not have been much help in the absence of the general theory that
explains and motivates the distinction, but Panaetius did, of course, offer explan-
ations of the main criteria for decision—the honourable (honestum⁴) and the
useful, each apparently at the beginning of the relevant discussion.

² See the compliments to Panaetius at the end of Fin. IV 79.

³ Note the order in which these considerations are mentioned, which might imply a ranking: if a proposed
action is not honourable (i.e. dishonourable), it might be ruled out, so that conflicts would not arise.
⁴ I use the translation ‘honourable’ rather than ‘morally good’, since the honestum includes more—
for instance, theoretical wisdom and ‘great deeds’. Unless otherwise indicated, in this as in most
quotations in the text I follow the translation by E.M. Atkins in M.I. Griffin and E.M. Atkins, Cicero,
On Duties, Cambridge 1991, with occasional modifications and two terminological changes: officium is
mostly rendered as ‘appropriate action’ rather than ‘duty’, decorum as ‘fitting’ (a term still used in
aesthetic theory) rather than ‘seemly’.
       ’    225

Cicero turns to Panaetius’s exposition after his critical remarks, with the
description of the honourable (I 11–17). It seems to begin just like the general
exposition of Stoic ethics in Fin. III, with the subject of οἰκείωσις—the natural
instinct for self-preservation that enables living things to seek out what they need
to sustain and perfect themselves, and the equally natural instinct for social living,
exemplified by the desire for a partner to produce offspring and the standard
example of parental care for children. These are impulses that all animals are said
to have in common. It may be worth noting, however, that Cicero’s exposition in
Fin. III 16–19 describes only the instinct for self-preservation, introducing the
social instinct only much later (III 62–68).
Here, the two natural impulses are simply mentioned, and then we turn to the
special case of the natural concerns of humans as rational animals. In book I,
Cicero describes four natural tendencies or desires, each based on specifically
human capacities. The development of these capacities and the achievement of the
aims of the associated desires is said to lead to what appear to be the four cardinal
virtues, or versions of them, which together constitute the honestum—‘what is
honorable even if it is not admired, and what we may truly call worthy of praise,
even if it is not praised by anyone’ (1.14). Panaetius’s procedure is well illustrated
by a simile explicitly ascribed to him by Stobaeus:

Panaetius said that what happens in the case of the virtues is similar to a
situation in which many archers have a single target, and this target contains
lines of different colors. Now each of the archers has the aim of hitting
the target, but one by hitting the white line, another the black one, and
yet another aiming at a line of some other color. For just as these archers
make it their ultimate end to hit the target, but different archers choose
different ways of achieving this, so all the virtues make happiness their end,
which lies in living in agreement with nature, but each of them arrives at it in a
different way.⁵

The virtues have a common goal, namely happiness or the best human life, but
they reach it by aiming at different parts of it. So the four starting points or
resources (ἀφορμαί)⁶ that Panaetius ascribed to rational animals would lead them
to pursue different aims, the attainment of which would make them virtuous. This

⁵ Ὅμοιον γὰρ ἔλεγεν εἶναι ὁ Παναίτιος τὸ συμβαῖνον ἐπὶ τῶν ἀρετῶν, ὠς εἰ πολλοῖς τοξόταις εἷς σκο
ὸς εἴη κείμενος, ἔχοι δ’ οὖτος ἐν αὑτῷ γραμμὰς διαφόρους τοῖς χρώμασιν· εἶθ’ ἕκαστος μὲν στοχάζοιτο
τοῦ τυχεῖν τοῦ σκοποῦ, ἥδη δ’ ὁ μὲν διὰ τοῦ πατάξαι εἰς τὴν λευκὴν εἰ τύχοι γραμμήν, ὁ δὲ διὰ τοῦ εἰς
τὴν μελαίναν, ἄλλος δὲ διὰ τοῦ εἱς ἄλλο τι χρῶμα γραμμῆς. Καθάπερ γὰρ τούτους ὡς μὲν ἀνωτάτω τέλος π
ιεῖσθαι τὸ τυχεῖν τοῦ σκοποῦ, ἤδη δ’ ἄλλον κατ’ ἄλλον τρόπον προτίθεσθαι τὴν τεῦξιν, τὸν αὐτὸν τρόπον
καὶ τὰς ἀρετὰς πάσας ποιεῖσθαι μὲν τὲλος τὸ εὐδαιμονεῖν, ὅ ἐστι κείμενον ἐν τῷ ζῆν ὁμολογουμένως τ
ῇ φύσει, τούτου δ’ ἄλλην κατ’ ἄλλον τυγχάνειν. (Stob. ecl. II p. 63, 10–64, 12 Wachsmuth/Hense; test. 54
Alesse; SVF III 280). Panaetius may have been inspired by Aristotle; see EN 1.1094a23–24.
⁶ The Greek word that Panaetius used, ἀφορμαί, can be translated in both ways. F. Alesse prefers
‘resource’; presumably the Greek term covers both.
226    

does seem to be an innovation, but not a heresy—the human virtues, not surpris-
ingly, are based on human rationality. It is, first of all, the source of the desire for
knowledge, which leads to wisdom, but also—given human language—the desire
to communicate and live together with others—the tendency that is perfected by
justice. (Here Panaetius appears to be inspired by Aristotle’s Politics I 2.1253a9ff.)
The desire to find the truth is then also said to be linked to ‘a certain desire for
leadership’, as a result of which ‘a mind well equipped by nature is not willing to
obey anyone except a person who offers advice or teaching, or whose commands
are given justly and legitimately for the sake of some benefit’. Evidently, animals
do not deliberate about right and wrong, and so the desire to make one’s own
decisions as far as possible is also based on rationality. This desire is said to be the
source of magnanimity and contempt for merely human things. Here Panaetius
has replaced the virtue of courage by magnanimity—a change that does not seem
to have been unusual, though normally greatness of soul is listed by the Stoics as
subordinate to courage.
Next comes what was probably Panaetius’s most important innovation. We are
told that

The power of nature is not insignificant in this too, that this one animal alone
perceives what order there is, what is fitting, what limit there is in words and
deeds. No other animal, therefore, perceives the beauty, the loveliness, and the
congruence of parts of the things that sight perceives. Man’s rational nature then
leads him to transfer those qualities by analogy from the eyes to the mind,
[a transfer that would seem easier in Greek than in Latin, since the word for
aesthetic beauty and excellence of character and action was the same, καλόν] and
he thinks that beauty, constancy and order should be preserved, and much more
so, in one’s decisions and in one’s actions. He is careful also to do nothing in an
indecent or effeminate way, in all his opinions and actions not to think or do
anything driven by lust. (I 14)⁷

This passage takes the place of the fourth cardinal virtue, temperance or moder-
ation, and indeed in the following paragraph Cicero says that modesty and
temperance are contained in this one of the four ‘parts’ of the honestum from
which everything honourable arises. But as the long chapter devoted to this virtue
will show, it is not identical with moderation (σωφροσύνη).

⁷ Nec vero illa parva vis naturae est rationisque, quod unum hoc animal sentit quid sit ordo, quid sit
quod deceat, in factis dictisque qui modus. Itaque eorum ipsorum quae aspectu sentiuntur nullum aliud
animal pulchritudinem, venustatem, convenientiam partium sentit; quam similitudinem natura ratio-
que ab oculis ad animum transferens multo etiam magis pulchritudinem constantiam ordinem in
consiliis factisque conservandam putat, cavetque ne quid indecore effeminateque faciat, tum in
omnibus et opinionibus et factis ne quid libidinose aut faciat aut cogitet.
       ’    227

Cicero does not give a name to the virtue that arises from the desire for inner
beauty. At this point, we might note that the analogy between visible beauty and
the beauty of mind does not seem to apply to a particular part of virtue, but to
human virtue as a whole. Yet the desire for inner beauty is distinct from the three
starting points mentioned before, and so it can indeed be seen as another resource
that might lead humans to virtue. The relation to moderation or temperance in
particular remains unclear.
The substitution of magnanimity for courage, and decorum—as I will call it for
lack of a proper term—for temperance, could be explained by the fact that it might
be difficult to find specific desires that lead to courage (to face great risks?) or
temperance (to avoid excess?), and indeed courage appears in the long chapter on
magnanimity (I 61ff.), along with a desire for freedom (I 68) in the sense of self-
determination, which explains the ‘desire for leadership’ mentioned before, and
temperance, of course, reappears in the chapter on the decorum.
The four specifically human desires or ‘resources’ (ἀφορμαί) are the starting
points of appropriate action in the sense that one can figure out how one should
act by considering what should be done in order to achieve the aims of these desires.
The impulse to self-preservation, common to all animals, is treated in book II,
indicating that self-interest should be considered only after virtue. But in this
chapter I am more interested in the cognitive side of the human sense of
beauty—the ability to recognize beauty in both the aesthetic and the moral sense,
which turns out not only to lead to a desire for order and moderation in one’s own
actions, but also to enable humans to recognize and appreciate virtue in all its forms.
Cicero explains the wider function of the sense of beauty at the beginning of the
chapter in which he deals with the appropriate actions derived from the fourth
virtue, I 93:

Next we must discuss the one remaining part of honorableness. Under this
appear a sense of shame and what one might call the ordered beauty of life,
restraint and modesty, a calming of all the agitations of the mind and the due
measure in all things.
Under this heading is included what in Latin may be called decorum; the Greek
for it is πρέπον. The essence of this is that it cannot be separated from what is
honorable, for what is fitting is honorable, and what is honorable is fitting. It is
easier to understand than to explain what the difference is between the honorable
and the fitting.⁸

⁸ Off. 1 93: sequitur ut de una reliqua parte honestatis dicendum sit, in qua verecundia at quasi
quidam ornatus vitae, temperantia et modestia omnisque sedatio perturbationum animi et rerum
modus cernitur. hoc loco id quod dici latine decorum potest, graece enim πρέπον dicitur decorum.
Huius vis ea est ut ab honesto non queat separari; nam et quod decet, honestum est et quod honestum
est, decet; qualis autem differentia sit honesti et decori, facilius intellegi quam explanari potest.
228    

Τὸ πρέπον was Panaetius’s term, a word that had long been used in Greek aesthetic
theory for the beauty of works of art, in literature as well as in music and the visual
arts.⁹ Cicero probably mentions it only in his longer treatment, since initially he
had just wanted to give a list of the four cardinal virtues. One difficulty that
resulted from the inclusion of decorum and the language of aesthetics was
terminological: πρέπον was used as an explication or synonym for the Greek
word καλόν, which describes not only beauty in its aesthetic form, but anything
that one might find admirable, and was also the standard term for moral goodness
or nobility. This may explain why Cicero finds it ‘easier to understand than to
explain’ the difference between decorum and honestum.
We can find a much more detailed account of the analogy invoked by Cicero in
his first description of the decorum in Seneca’s Letter to Lucilius 120, where Seneca
explains how we arrive at our first notion of the good. Seneca describes how we
recognize virtue in others in just the terms by which Cicero characterizes beauty:
virtue is ‘revealed to us by its order and decorousness (decor), its constancy and
the harmony of all actions among themselves . . . ’ (Epistulae Morales 120.11).
Here is Cicero’s attempt at explanation:

For whatever it is that is fitting, it becomes apparent when the honorable has
preceded it. Therefore what is fitting manifests itself not only in the part of the
honorable that we are now to discuss, but in the three earlier ones as well . . .
Therefore, the fitting of which I am speaking relates to everything that is
honorable, and in such a way that it does not take subtle reasoning to discern
it, but that it is in plain view. For there is such a thing as what is fitting, and it is
understood in every virtue and can be separated from virtue in thought rather
than in fact. Just as the loveliness and beauty of the body cannot be separated from
health, so the fitting of which we are speaking is entirely blended with virtue, but
can be distinguished by thought and reflection. (I 94–95, emphasis added)

The analogy between the—alleged—relation between health and beauty and that
between virtue and what is fitting shows what is meant by the phrase ‘when the
honorable has preceded it’: as good looks follow upon health of the body, so what
is fitting will be apparent when virtue is present in the soul. In other words,
fittingness is the outer appearance of inner beauty.
Clearly, this is not a description of a fourth virtue, but of a quality that goes with
virtue and virtuous action in general. So, after this passage Cicero acknowledges
that there are two senses of the word ‘decorum’:

⁹ The best survey of the history of this term before and up to Panaetius is still the article by
M. Pohlenz, τὸ πρέπον (1933); repr. in his Kleine Schriften 1 (ed. H. Dörrie), Hildesheim 1965, 100–39.
       ’    229

Now this has two senses: first, we understand a fittingness of a general kind,
involved with honorable behavior as a whole; secondly, something subordinate to
this, which relates to individual parts of what is honorable.¹⁰ (I 96)

Cicero goes on to illustrate what is meant by decorum by comparing it with the

decorum sought after by poets and playwrights who have to match the words they
give to the personages in their plays to the character of the people represented, and
in the later discussion of the individual decorum he keeps coming back to
examples from literature to illustrate proper conduct in social life. In I 97–98 he
explains that nature herself has assigned their roles to humans, where the first
corresponds to the definition of the general decorum, the second to that of the
subordinate, which is then further described and forms the subject of the whole
chapter.
What follows is the famous doctrine of the four personae or roles that every
human individual has to play in life.
However, let me for the moment stay with the epistemological role of the
general notion. Behind the description of decorum/πρέπον as the outward mani-
festation of inner beauty lies its brief appearance in Plato’s Hippias Maior as a
candidate for the definition of beauty (293Dff.). It comes up as a friendly sugges-
tion from Socrates’ double, and after several failed attempts of his own, Hippias
accepts it enthusiastically. But Hippias as portrayed by Plato is exceedingly obtuse,
and so it is no surprise that he is unable to defend this proposal against the
objections raised by Socrates—just as Nicias in the Laches (194Dff.) is unable to
defend the Socratic view that courage is a kind of knowledge. When Socrates asks
whether τὸ πρέπον is what makes things appear beautiful or what makes them be
so, Hippias opts for appearing, which leads to his defeat. Socrates proceeds to
interpret appearance as mere and possibly deceptive appearance and argues that
this cannot be what they were looking for, namely what accounts for things being
really beautiful and not just appearing to be beautiful. Panaetius, though, was not
Hippias, and he knew to avail himself of a distinction that seems to be missing in
Plato’s dialogue—the distinction between appearance as outer manifestation and
appearance as mere deceptive semblance. To use the analogy cited by Cicero, if
physical beauty is the outward appearance of physical health, the fact that it may
be imitated by cosmetics does not show that it must always be a mere semblance.
Similarly, if the constancy and order of conduct is the manifestation of inner
beauty, one cannot exclude the possibility that it is mere appearance, but it will
still be what makes inner beauty recognizable by others. And that, I think, is the
function of the general decorum: it allows people to appreciate virtue and to try to

¹⁰ Est autem eius descriptio duplex; nam et generale quoddam decorum intellegimus, quod in omni
honestate versatur, et aliud huic subiectum, quod pertinet ad singulas partes honestatis.
230    

attain it without having a deeper understanding of the inner state that is the cause
of the observable conduct.
Let me now take a step back and compare Panaetius’s treatment of the
honestum with the more conservative exposition of Stoic ethics in Fin. III 21ff.
Both begin, as I said, with οἰκείωσις, but Cato, Cicero’s speaker in Fin. III,
mentions only the two natural instincts that humans have in common with
other animals—self-preservation and concern for others. Cato outlines the psy-
chological development of a human being that begins by following the impulse for
self-preservation, trying to obtain the things that will keep it alive and healthy and
avoiding what might harm it. At some stage of this way of living—probably with
the advent of reason, that is, at the age of fourteen or so—the human being is
supposed to form a conception of the good by realizing that it was orderliness and
harmony that made her own (and perhaps also other people’s) conduct good. This
realization then leads her to see order and harmony as the only real good and to
relegate the objects of her previous choices to the status of mere ‘indifferents’,
things that are pursued only because nature wants us to pursue them, not because
they have any intrinsic value by themselves. Cicero’s Cato does not explain the
reasoning that is supposed to lie behind this radical change—and in book IV
(26–28) of the Fin., Cicero as speaker criticizes it as an unintelligible break in what
ought to be the pursuit of human perfection, including not just self-preservation,
but the virtues as goods of the soul as well.
Panaetius’s account of what one might call specifically human οἰκείωσις pro-
vides a way of avoiding the dramatic change postulated in Cato’s story. But it did
not follow the alternative urged upon the Stoics in Fin. IV, namely including the
desire for virtue in the general aim of self-perfection. Instead, Panaetius intro-
duced four specifically human desires, the fulfilment of which would constitute
virtue itself. Virtue is recognized as a good due to the human sense of beauty, and
it can be attained by acting in accordance with natural human tendencies. The
objects of the natural instinct for self-preservation are not among those specifically
human aims—in Panaetius’s scheme, they appear under the rubric of the useful,
the pursuit of which is indeed appropriate, but not necessarily admirable. This is
reflected in the difference between the definitions of the highest good ascribed to
Panaetius’s immediate predecessors, Diogenes of Babylon and Antipater: while
they defined the end in terms of the reasonable selection of natural things—the
objects of the original impulses given to animals by nature—Panaetius is said to
have defined it as ‘living according to the resources (ἀφορμαί) given to us by
nature’.¹¹ In this way he avoided another objection to the Stoic theory, probably
going back to the Academic sceptic Carneades, namely that the Stoics were
postulating two goals of life rather than one. According to the theory as set out

¹¹ Clement Alex. Stromata II 21: Παναίτιος τὸ ζῆν κατὰ τὰς δεδομένας ἡμῖν ἐκ φύσεως ἀφορμὰς τέλος
ἀπεφήνατο.
       ’    231

in Fin. III, humans strive for agreement with the rationality and order of the
universe by trying to obtain the ‘natural things’, while at the same time treating
them as indifferent with regard to happiness. In Panaetius’s theory, the four
special aims of rational animals are also admirable, so that achieving those aims
will be tantamount to virtue. It remains the case that the appropriate actions
derived from those aims can be described as reasonable selections of things that
are not moral goods—for example, giving money to the poor rather than spending
it all on one’s own amusements.
However, an ‘orthodox’ Stoic like Cato would no doubt object that Panaetius
had given up the crucial distinction between two sorts of values that motivated the
terminological distinction between preferred and dispreferred indifferents on the
one hand, real goods and evils on the other. In Panaetius’s book, the so-called
natural advantages or preferred indifferents are treated under the rubric of
the useful or beneficial, separately from the honourable, that is, virtue, but there
are appropriate actions derived from the useful in this sense as well as from the
honourable. Now, according to orthodox Stoic doctrine, what is useful cannot be
separated from what is good, and only what is honourable is really useful—and
that should mean that the ‘natural advantages’ should not be described as useful,
but precisely as indifferent.
Yet Cicero assures us (Off. III 11–12) that Panaetius did not abandon the
central Stoic thesis that virtue is the only good,¹² and that he also held that
nothing can be useful that is not honourable. Panaetius was evidently following
ordinary usage in treating moral values separately from prudential values, but one
might think that this can no longer be seen merely as a matter of ‘speaking with
the vulgar’.
Or can it?
Let me first point out how close the theory, as presented by Cicero, comes to a
number of Stoic theses connected with the distinction between goods and indif-
ferents. The thought that material goods do not count towards the real good that
lies in virtue comes up in many places in Off. I, beginning with the introductory
outline of the honourable. It is implicit in the description of the virtue of
magnanimity that takes the place usually given to courage: magnanimity or
greatness of spirit is said to lead to ‘contempt for merely human things’ (I 14;
cf. 66–67). This seems to me to capture what the Stoics expressed by calling such
things ‘preferred indifferents’: they are worth taking when easily available without
injustice to others, but they will be disregarded where the aim is some great deed
or if they can be obtained only by the sacrifice of liberty and personal independ-
ence. The notorious rejection of all emotions (as due to mistaken value judge-
ments) appears as a prerequisite to virtuous action, especially in the chapter on the

¹² Μόνον τὸ καλόν ἀγαθόν—more literally translated, ‘only what is honourable is good’, which
includes not only virtue, but also virtuous action.
232    

special decorum, where Cicero emphasizes that one’s impulses to action (appeti-
tus, used by Cicero to translate ὁρμή) must always be in accord with reason
(I 101–102), and notes that anger and excessive desire, as well as fear and excessive
pleasure, distort not only the mind but even the body. Needless to say, this is not
as clear as an explicit statement that any degree of emotion is dangerous and to be
avoided, just as the phrase ‘contempt for merely human things’ does not say that
material or physical advantages never contribute anything that could make one’s
life better. What is missing is not only explicitness, but mainly a clear explanation
and justification of the alleged contempt for merely human things, or of the
general dangerousness of emotion—such as one could find in more technical
Stoic treatises on ethics. The one Stoic claim that seems incompatible with
Panaetius’s exposition is precisely the thesis that the good, and hence the
honourable, coincides with the useful. Once virtue is identified with the pursuit
of the specifically rational aims and utility with the ‘comforts of life’ (I 9; II 11)
such as health, material possessions, power, and the support of others, this
coincidence can no longer be maintained, however much Cicero may protest
(II 9–10; III 11–12) that honestum and utile should never have been seen as
separable from one another. Sacrificing one’s life for one’s country, to take the
most extreme case, or even spending a lot of time and money on helping the poor,
are not useful activities in the sense of making one’s own life more comfortable, or
even advancing one’s career—the factor most prominent in book II. Nonetheless,
one could still maintain that conflicts between the honourable and the useful are
merely apparent by arguing that any alleged advantage that could be derived from
immoral action would be outweighed by the harm inflicted upon oneself by
committing a crime. In order to defend the view that nothing immoral can be
useful, one does not need to claim that every virtuous activity is useful in the
ordinary sense, nor, for that matter, that every useful activity must also be virtuous
or praiseworthy—it is enough that it should be appropriate and permissible.
Furthermore, the problems that Panaetius described as apparent conflicts between
virtue and utility could not be simply eliminated by terminological legislation—
they had, in fact, already been taken up by Panaetius’s predecessors, but presum-
ably as conflicts between the two natural impulses of self-preservation and
concern for others. One need not adopt the paradoxical claim of the Stoics that
the honourable coincides with the useful in order to interpret the Stoic formula
‘living in agreement with nature’ in the way that Cicero understands Panaetius:
‘always conform with virtue, and select the rest of natural things provided that
they do not conflict with virtue’ (III 13¹³).

¹³ Etenim quod summum bonum a Stoicis dicitur, convenienter naturae vivere, id habet hanc, ut
opinor, sententiam, cum virtute congruere semper, cetera autem, quae secundem naturam essent, ita
legere si ea virtuti non repugnarent.
       ’    233

But what about the most famous Stoic thesis, that virtue is the only good? One
could easily read the Off. as showing that virtue is the greatest good, but not what
the Stoics wanted to claim, namely that it is sufficient for happiness. On the other
hand, that thesis is also not ruled out—it is neither explicitly argued nor implicitly
denied by what one finds in Cicero’s book. The reason for this absence comes out,
I think, in the passage immediately after Cicero’s emphatic defence of Panaetius’s
Stoic orthodoxy, Off. III 13–15. It has often been pointed out by commentators
that this passage is not very well integrated in its context, and in fact it might have
been more helpful if it had appeared in the introduction to the entire work (which
Cicero had omitted). Cicero explains that perfect virtue, according to the Stoics,
can only be found in the wise, but that ordinary people can make moral progress
by acting in the appropriate way—and appropriate action is the same for sages as
for fools. He then somewhat abruptly tells us that ordinary people are likely to see
those who act in appropriate ways as shining examples of perfection and praise
them as paradigms of virtue. This is similar to the aesthetic judgements of the
inexperienced—they will recognize some admirable features in works of art but
overlook the flaws that experts might detect. The connection between these two
points can, I think, be understood by looking again at Seneca’s letter 120 where he
explains the analogy by which we arrive at our first notion of the good (ep. mor.
120.5). Given that real sages are as rare as the phoenix, ordinary people will, in
fact, arrive at their conception of the morally good by looking at less than perfect
examples. But this will be enough to inspire admiration and presumably the desire
to imitate them. However, where Seneca goes on to say that nature herself leads us
to overlook the defects and arrive at a conception of the good, as it were, by
idealization, Cicero—or, I would say, Panaetius—says that the inexperienced can
easily be made to change their judgement when instructed by experts. What this
suggests, I think, is that Panaetius saw his book about appropriate action not as a
kind of second-class morality for fools, but as an introduction to Stoic ethics that
leaves the more complicated theoretical parts for more advanced students.
A person who had read his book might be prepared to receive more precise
instructions from experts—that is, Stoic philosophers—and might then also see
that appropriate action is not all there is to virtue. By coming to a deeper
understanding of the inner state that underlies the conduct she has learned to
admire, such a person might also come to accept the claim that virtue is sufficient
for happiness. But this requires some effort and philosophical learning. It is one
thing to adopt a terminology that makes virtue and virtuous action the only things
that deserve to be called good, another to convince people that this is not just a
verbal manoeuvre—as Cicero has his Peripatetic speaker say at Fin. V 88–89. If
one believed that virtue alone and by itself is sufficient for happiness, one might
also come to believe that its value is not just greater, but different in kind from
other values, so that the terminology makes sense. But the specious syllogisms that
Cicero puts on display in Tusculans V (43–53) will hardly do the trick. Seneca’s
234    

lyrical descriptions of the freedom and serenity of the wise¹⁴ that comes with
thinking that nothing external can harm you any more may be better, but still not
quite enough—they only show the subjective side of happiness, not the thoughts
that lead to it. The more substantial arguments come from Stoic cosmic teleology
and the thesis that the universe is the work of a supremely rational and benevolent
divinity. This belief might well inspire admiration and the wish to be part of that
rational order. But a person who has reached this kind of faith would probably be
a Stoic sage—and that means that only perfect virtue is truly sufficient for
happiness. However, perfect virtue is beyond the reach of almost all ordinary
mortals, as the Stoics were the first to admit. Nevertheless, they also held that one
should not despise the moral achievements of ordinary decent people—what
Cicero calls the boni, true gentlemen who are, however, not philosophers (Off.
III 17).
To get back to Panaetius: we need not doubt Cicero’s emphatic assertions that
he never gave up the Stoic dogma that virtue is the only good. He could surely also
use the esoteric lingo of his school, including preferred indifferents, selection as
distinct from choice, and so on. But when addressing educated laymen—such as
Cicero’s son, whose philosophy teacher was after all a Peripatetic, not a Stoic—he
may not have found it necessary or helpful to insist on it.
Now, let us imagine an eighteenth-century reader brought up on Cicero’s
book—‘Tully’s Offices’, as it was often referred to, either as part of the philosophy
curriculum or simply as an edifying manual of advice for future leaders. Such
a person might well have found that this book contains all one needs from a
moral theory. It explains how and why people come to admire and strive for moral
virtue: they have a moral sense that enables them to appreciate it. The manual
then shows how the virtues are grounded in human nature and goes on to develop
detailed practical rules for justice and beneficence, warns us against excessive
ambition but praises selfless actions that require great courage, and finally advises
us about how to find the approval of others by polite conduct and good taste. The
second book, under the rubric of the useful, instructs the reader about the single
most important thing that will make life secure and comfortable, namely the help
and support of other men (here Cicero is clearly writing for young aristocrats).
Then, at least in Cicero’s book though not in Panaetius’s, there is advice about
decisions in difficult situations where one’s self-interest might conflict with
concern for others. What more should one expect from a moral theory?
Eighteenth-century readers were not Stoics. They did not hold the most famous
Stoic thesis, that virtue is the only good, nor did they expect a moral theory to have
the form of the ancient Greek theories, at least since Aristotle: starting from a
determination of what counts as the highest good in action—εὐδαιμονία—and

¹⁴ See, for example, at ep.mor. 59.16: Talis est sapientis animus qualis mundus super lunam: semper
illic serenum est.
       ’    235

then explaining how one should try to live in order to enjoy the best possible
human life. So these readers might not miss an account of happiness as the highest
good, nor expect it in a book of practical moral advice.
The famous claim that virtue is the only good and, therefore, sufficient for
happiness plays no role in the Off.; and I think it is significant that Cicero defends
Panaetius’s orthodoxy as a Stoic against other Stoics, who maintained that con-
flicts between virtue and utility could not arise, by repeatedly asserting that
Panaetius did hold this fundamental thesis, and that the conflicts between the
honourable and the useful would be only apparent. Later readers of the Off. could
well come away with the view that self-love and benevolence are equally natural
for humans, and that the ‘comforts of life’ such as health, wealth, good looks and,
of course, friends, would be quite relevant to human happiness. What Panaetius
treated in his book was—as the title indicates—only a part of Stoic ethics; it was
probably intended as an introduction for educated gentlemen that might perhaps
inspire some of them to go on to study with the experts. Judging by Cicero’s
version, it is quite remarkable to what extent the technicalities of the Stoic theory
could be left out, and how much of Stoic doctrine could still be at least suggested
without the Stoic terminology.
But in order to argue for the Stoic conception of the highest good—seeing
rational humans as members of a community that also included the gods, and
accepting the rational order of the universe as the only thing that really deserves to
be called good—more would have been needed, including the distinctions between
real goods and evils and so-called ‘preferred’ or ‘dispreferred’ indifferents: the
distinction between appropriate action and right or virtuous action, and also Stoic
theology, which describes the universe as a home for gods and men governed by a
supreme reason. Nevertheless, since Stoic theology also does not have an import-
ant place in the much more detailed exposition of Stoic ethics in Fin. III,¹⁵ some
scholars have argued that the theological part is not needed to justify the Stoic
view of happiness.¹⁶ In the de Finibus, the transition from the pursuit of the things
needed for self-preservation and self-development (οἰκείωσις) to the recognition
of rational order and harmony—that is, virtue—as the only real good is intro-
duced very early on, when only self-love, but not the instinct for communal living
has been described. Cicero says that after living in the natural way for a long time,
one eventually begins to understand what really deserves to be called good and
arrives ‘by insight and reasoning’ (cognitione et ratione collegit) at the conclusion
that the highest good is what the Greeks call ὁμολογία—consistency or

¹⁵ Cicero adds this point, almost like a footnote, at the end of his exposition, Fin. III 73, where he
explains that a person who is going to live in agreement with nature will have to start from the universe
and its administration, that is, Stoic cosmology.
¹⁶ See J. Annas, ‘Ethics in Stoic Philosophy’, Phronesis LII, 2007, 58–87. She agrees that Stoic ethics
can be connected with Stoic physics (which includes their theology), but simply because all parts of
Stoic philosophy are interconnected.
236    

convenientia, as Cicero proposes to translate it—and from then on one considers

virtue (here equated with ὁμολογία) as the only real good (Fin. III 20–21). The
reasoning that leads to this conclusion is not spelled out, and the change of values
therefore sounds rather mysterious. T. Irwin thinks that what underlies it is the
recognition of moral beauty that we find in the Off.,¹⁷ but surely this is not enough
to make one think that nothing but virtue is good at all. As far as I can see, the
conviction that agreement with both one’s own and universal nature—as
Chrysippus is said to have held (DL VII 87–88)—is the only real good can only
be motivated by Stoic theological cosmology. Seneca seems to agree. We should
remember that in ep.mor. 120 he only speaks about our first notion of the good; in
ep.mor. 124.14, he arrives at an explication of the human good:

There are four natures we should mention here: of the tree, animal, man, and
god. The last two, having reasoning power, are of the same nature, distinct only
by virtue of the immortality of the one and the mortality of the other. Of one of
these—namely god—it is nature that perfects the good; of the other—namely
man—pains and study do so. All other things are perfect only in their particular
nature, and not truly perfect, since they lack reason. Indeed, to sum up, that alone
is perfect which is perfect according to nature as a whole, and nature as a whole is
possessed of reason.¹⁸ (tr. R.M. Gummere, 1925)

So Panaetius’s book about appropriate action may well be read as an account of

ethics without metaphysics, but perhaps not quite of Stoic ethics without
metaphysics.

Decorum as a Virtue

Turning now to the specific virtue falling under decorum, let me begin with a
quotation from Hume:

It is the nature and, indeed, the definition of virtue that it is a quality of the mind
agreeable or approved of by everyone who considers or contemplates it.¹⁹

¹⁷ See T. Irwin, ‘Stoic Naturalism and its Critics’, in B. Inwood (ed.), The Cambridge Companion to
the Stoics, Cambridge 2003, 345–64.
¹⁸ quattuor hae naturae sunt, arboris, animalis, hominis, dei: haec duo, quae rationalia sunt, eandem
naturam habent; illo diversa sunt, quod alterum immortale, alterum mortale est. ex his ergo unius
bonum natura fecit, dei scilicet, alterius cura, hominis. cetera tantum in sua natura perfecta sunt, non
vere perfecta, a quibus abest ratio. hoc enim demum perfectum est, quod secundum universam
naturam perfectum. universa autem natura rationalis est.
¹⁹ David Hume, Enquiry Concerning the Principles of Morals, ed. Selby-Bigge. 2nd ed. Oxford
University Press 1902 , s. VIII, fn. 1, 261 and Appendix IV, 312–23. For a reference to Cicero see
p. 266, p. 318.
       ’    237

In an appendix, Hume later defends his wide definition against more restrictive
views by appealing to the ancients in general, and explicitly to Cicero’s Offices in
particular, insisting that the various ways that had been proposed to distinguish
virtues, for example, from talents, vices from defects etc., amounted to no more
than verbal disputes.
Whether or not we are willing to accept Hume’s definition, he is certainly right
about the wider conceptions of virtue among the ancients. Indeed, some recent
translators of Aristotle have switched to ‘excellence’ as a translation of Greek
ἀρετή, in recognition of the fact that theoretical knowledge, for instance, can
hardly be seen as a moral virtue. What Cicero and Panaetius described as a virtue
concerned with ‘the fitting’ or decorum, then, should be seen in this ancient
framework.
In the introduction of the four parts of the honourable, the word decorum itself
did not occur, though we find the verb decere, the adverb indecore, and finally the
noun decus. So the long chapter that deals with the fourth virtue (Off. I 93–151)
opens with an explanation of what is meant by decorum. Cicero begins with an
explanation of the sense that goes with the analogy between outer and inner
beauty to which he appealed in the earlier chapter by describing the uniquely
human sense of beauty that inspires it. But as we noted before, this does not fit a
particular virtue, since beauty of the soul is virtue as a whole. So now he adds the
distinction between two senses of the word, the second finally identifying the
specific virtue that will be treated in this chapter:

Now this has two senses: first, we understand a fittingness of a general kind,
involved with honorable behavior as a whole; secondly, something subordinate to
this, which relates to individual parts of what is honorable.
The first tends to be defined something like this: fitting is what agrees with
human excellence in the respect in which human nature is different from that of
other living things.
The part that is subordinate to the genus they define so as to say that fitting is
what agrees with nature in such a way that it shows moderation and temperance
together with the kind of appearance that belongs to a free man.²⁰ (Off. I, 96)

Unfortunately, this is still less than clear. We can find the first sense by going back
to the earlier passage in which the virtues were introduced: what distinguishes
humans from other animals is rationality, and its excellence is presumably

²⁰ Est autem eius descriptio duplex; nam et generale quoddam decorum intellegimus, quod in omni
honestate versatur, et aliud huic subiectum, quod pertinet ad singulas partes honestatis.
Atque illud superius sic fere definiri solet, decorum id esse quod consentaneum sit hominis
excellentiae in eo in quo natura eius a reliquis animalibus differat. Quae autem pars subiecta generi
est, eam sic definiunt ut id decorum velint essequod ita naturae consentaneum sit ut in eo moderatio et
temperantia appareat cum specie quadam liberali (Off. I 96).
238    

constancy and the harmony between thought and action that Cicero emphasized
before—what the Greek Stoics called ὁμολογία, the consistency of the Stoic Sage,
and what we might also, with a Latin term, describe as integrity. This is what one
will aim at once one has accepted the analogy between outer and inner beauty. But
for the particular virtue to be discussed here, we have once again moderation and
temperance, with the addition of ‘the kind of appearance that belongs to a free
man’.²¹ The addition provides, I think, a clue to the specific aim of the particular
virtue.
In 1.14, Cicero had concluded his initial description of the four parts of the
honestum by saying that ‘even if it is not accorded acclaim, it is still honorable and,
as we truly claim, even if no one praises it, it is by nature worthy of praise’. In the
later chapter, however, the accent is on the fact that what is fitting delights others
and attracts their applause, as in a play, where the poet’s words fit the character of
the person represented. Returning to the analogy with beauty, Cicero says

for just as the eye is aroused by the beauty of a body, and is delighted just because
all its parts are in graceful harmony, so this fittingness, shining out in one’s life,
arouses the approval of one’s fellows, because of the order and constancy and
moderation of every word and action.²²

Though virtue in general does not depend on praise or recognition by others,

moderation and the outward appearance of a free man are also what will make a
person attractive to others. The specific aim of the last virtue, then, is the approval
of one’s fellow humans:

Thus we must exercise a kind of reverence towards men, both towards the best of
them and also towards the rest. To neglect what others think of oneself is the
mark not only of arrogance, but also of utter laxity.²³

Though virtue itself does not depend on the praise of others, decorum in the
specific sense does. And we can now also see why Panaetius would have wanted to
keep the twofold sense of the word—the second goes with another aspect of
beauty, attractiveness, to which he had not appealed before.
This was not uncontroversial among Stoics, as we can see from a passing remark
in Fin. III 57:

²¹ Atkins translates this phrase pointedly as ‘along with the appearance of a gentleman’—no doubt
what Cicero had in mind.
²² Ut enim pulchritudo corporis apta compositione membrorum movet oculos et delectat hoc ipso,
quod inter se omnes partes cum quodam lepore consentiunt, sic hoc decorum, quod elucet in vita,
movet approbationem eorum, quibuscum vivitur, ordine et constantia et moderatione dictorum
omnium atque factorum (I 98).
²³ Adhibenda est igitur quaedam reverentia adversus homines, et optimi cuiusque et reliquorum.
Nam neglegere quid de se quisque sentiat non solum adrogantis est, sed etiam omnino dissoluti (I 99).
       ’    239

As far as good reputation is concerned (what they call eudoxia; it is more suitable
in the present context to translate it as good reputation than as glory),
Chrysippus himself and Diogenes used to say that, aside from any usefulness it
may have, it was not worth lifting a finger for. I absolutely agree. But their
successors, unable to withstand Carneades, declared that a good reputation is
advantageous and worthy of adoption in its own right. One who is free-born and
well educated would want to be well thought of by parents, relatives, and good
people in general, and this for its own sake, not just because it is useful.²⁴
(tr. R. Woolf with modifications)

Panaetius is not mentioned here—the speaker is Cato, not Cicero—but the Stoic
successors of Diogenes of Babylon were Antipater of Tarsus and Panaetius. The
debate about this point is not reported anywhere else, but it seems clear to me that
Panaetius, at least, would have been meant.
Among the four traditional cardinal virtues, it might indeed be moderation that
is most likely to make a person a welcome member of social gatherings, but of
course much more is added under the label of decorum. An important addition
was verecundia (Greek αἰδώς; respect or a sense of shame), which had not been
mentioned in the earlier passage, but appears in the first sentence of the chapter
on decorum. As is well known, the moral doctrines of the early Stoics, including
the founder Zeno, were quite close to those of the Cynics, whose hallmark was
precisely the opposite of a sense of shame, namely ἀναίδεια or ἀναισχυντία,
shamelessness. Panaetius was clearly eager to rule out such tendencies, and
explicitly declared that Cynicism was to be avoided—with due exception being
made for Socrates:

No one should be led into the error of thinking that because Socrates or
Aristippus did or said something contrary to custom or civic practice, that is
something he may do himself. For those men acquired such freedom on account
of great, indeed divine, goodness. But the reasoning of the Cynics must be
entirely rejected, for it is hostile to a sense of shame, and without that nothing
can be right, and nothing honorable.²⁵ (Off. I 148; see also 128)

²⁴ De bona autem fama (quam enim appellant εὐδοξίαν aptius est bonam famam hoc loco appellare
quam gloriam), Chrysippus quidem et Diogenes detracta utilitate ne digitum quidem porrigendum esse
dicebant; quibus ego vehementer adsentior. Qui autem post eos fuerunt, cum Carneadem sustinere non
possent, hanc quam dixi bonam famam ipsam propter se esse sumendam dixerunt, esseque hominis
ingenui et liberaliter educati velle bene audire a parentibus, a propinquis, a bonis etiam viris, idque
propter rem ipsam, non propter usum (Fin. III 57).
²⁵ Nec quemquam hoc errore duci oportet, ut, si quid Socrates aut Aristippus contra morem
consuetudinemque civilem fecerint locutive sint, idem sibi arbitrentur licere; magnis illi et divinis
bonis hanc licentiam adsequebantur. Cynicorum vero ratio tota est eicienda, est enim inimica ver-
ecundiae, sine qua nihil rectum esse potest, nihil honestum.
240    

Having introduced verecundia, respect for others and their judgements about
oneself, Cicero adds a significant qualification:

There is a difference between justice and respect in considering persons. The part
of justice is not to harm a man, that of respect not to offend him, which shows
most clearly the meaning of decorum.²⁶ (99)

This brings me to a point of translation: the traditional English rendering of

Cicero’s title De Officiis is ‘On Duties’. But, as Dyck correctly points out in the
introduction to his commentary (pp. 7f.), ‘duty’ is not a good translation for
officium in this book, let alone of the Greek καθῆκον. A more adequate rendering
of both would be ‘appropriate action’—the usual translation for the Greek term.
Cicero himself makes a distinction between what is obligatory and what is fitting
in the context of discussing what is fitting in the style of speeches, at Orator
22.73–74:

. . . there is a big difference between saying that something is fitting and saying
that it is obligatory; for ‘obligatory’ indicates a perfect duty (officium) that must
be followed always and by all persons, while ‘fitting’ means that something is as it
were appropriate and consistent with the occasion and the person.²⁷

Of course, some officia will be duties, as this quotation actually shows, but what
Cicero discusses in this treatise is the wider class of appropriate actions. So
another difference between justice and respect is that while rules of justice must
always be observed by all persons, lack of respect, rude remarks, tasteless jokes, or
excessive bragging may be offensive, but they are not prohibited; while immoral
actions are never allowed, offensive ones may even sometimes be commendable,
for instance as intentional provocations—witness Socrates. What Cicero offers in
this chapter under the label of decorum, then, is not a set of moral prescriptions,
but rules of good manners and advice for socially pleasing conduct. Disregarding
such advice brings no legal sanctions, but social ones, ranging from reproach, for
example, for a rude remark, to ridicule, as in the case of bad poetry or garish
clothing.
In setting out the appropriate actions associated with decorum, Panetius used
an image that had been around among the Stoics (and other philosophers) for a
long time: the comparison of a person’s conduct with the role played by an actor
in drama. This comparison could be used in two ways: either to illustrate the

²⁶ Est autem quod differat in hominum ratione habenda inter iustitiam et verecundiam. Iustitiae
partes sunt non violare homines, verecundiae non offendere, in quo maxime vis perspicitur decori.
²⁷ . . . aliudque totum sit, utrum decere an oportere dicas; oportere enim perfectionem declarat
officii, quo et semper utendum est et omnibus, decere quasi aptum esse consentaneumque tempori
et personae (Orator 22.73–74).
       ’    241

consistency between the words and actions of a person throughout his life,
corresponding to the consistent representation of the character of the person he
is playing that a good actor should provide; or the ability of a wise man to find the
right way of acting in different circumstances, similar to the ability of an actor to
adapt to different roles in different plays.²⁸ Panaetius made use of both these
options. Cicero explains that Nature has assigned two roles to all humans. The
first is common to all of them as rational beings, the second belongs to each
individual in particular, and is given by his character and talents (107). The first
and common role had already been mostly covered in the preceding chapters on
the other virtues; it remained here just to add the virtue of moderation by
emphasizing, as a Stoic would, that one’s impulses should always be controlled
by reason—that is, one ought to avoid emotions, all of which were considered by
the Stoics as excessive impulses. This clearly belongs to the role common to all
humans, and the accent lies on consistency (constantia); as Seneca puts it,
‘Consider it as a great thing to play the role of a single man’,²⁹ rather than adopting
many different characters from day to day, as we all allegedly do. Cicero adds a
point derived from the superiority of rational over irrational animals by warning
against succumbing to the allure of pleasures—this may be natural for cattle and
other beasts who are guided only by pleasure and pain, but human dignity
requires that one should look down upon bodily pleasures and care for one’s
body only with regard to health and strength (101–106). Then he goes on to the
particular roles, dealing with individual personality and talents. The many vari-
ations in personality are illustrated with examples of famous people both Greek
and Roman to show that what fits one type of character may not be appropriate
for another. Some of the Greek examples (probably taken from Panaetius) are
morally rather dubious, but Cicero does not consider the question of possible
conflicts between personal traits and the common role of man, except for remark-
ing that one should hold on to one’s personal talents and dispositions ‘so long as
they are not vicious’ (Off. I 110). In this way one will avoid the awkwardness that
may come from trying to play roles for which one has no talent, and also preserve
consistency. He then adds yet another two roles, given by social status and by
one’s choice of occupation in life—a topic obviously important for a young man at
the beginning of his career. There follow some more aspects of social life,
including age, external appearance in public, and the like, and even a nice little
excursion on the right style of conversational speech as opposed to public orations
(132–135). This section is not very well organized—after all, this was Cicero’s last
book, which he never had a chance to revise. It might be amusing, but also a little

²⁸ For the use of the actor simile by earlier Stoics, see A.M. Ioppolo, Aristone di Chio e lo Stoicismo
antico, Naples 1980, 188–207.
²⁹ magnam rem puta unum hominem agere. Praeter sapientem autem nemo unum agit, ceteri
multiformes sumus’ (ep. mor. 120.22).
242    

tedious to go through all the various sections here that add up to the detailed
portrait of a Roman gentleman. Good manners and good taste, as Hume puts it—
the kind of conduct that will bring the approval and perhaps also the affection of
others. But would adherence to the decorum in this sense also add up to a virtue?
One might agree with Panaetius that the desire to be appreciated by others is
natural for humans as social animals, exceptions like Diogenes the Cynic not-
withstanding. But the problem with good manners seems to lie in the fact that,
unlike inner beauty and integrity, winning the approval of one’s fellow humans is
not in itself a noble aim. In this case one is actually aiming at the appearance of
good character; and just as healthy good looks may be an indication of bodily
health, but can also be imitated by cosmetics, so good manners may be an
expression of good character, but they can also be acquired without it. To cite
an example close to Cicero himself: Julius Caesar, for all we know, may well have
been the perfect example of a Roman gentleman as far as good manners are
concerned, and he evidently managed at least to attract the admiration and loyalty
of his troops. But Cicero would certainly not want to recognize him as an
honourable man. In this chapter, Cicero considers good manners only as the
expression of what might be considered as its legitimate source (and, of course,
also what impostors want to suggest). But then, good manners could also be used
to manipulate or to deceive others. What this shows, I think, is that some of the
starting points for virtue on Panaetius’s list were indeed no more than that—
starting points or resources, not reliable guides to virtue. A philosopher less
optimistic about human psychology—such as Hobbes—might, for instance,
have described the desires for ‘a certain kind of leadership’, or for great deeds,
as desires for power and glory. These desires, natural as they may be for some
‘minds well endowed by nature’, will only lead to virtue if they arise within the
context of a commitment to justice. In other words, morality comes in only with
justice, not from the desire itself. This is why we find Cicero in the chapter on
magnanimity insisting that courage and the willingness to take great risks or make
heroic efforts must always be linked to justice:

However, if the loftiness of spirit that reveals itself amid danger and toil is empty
of justice, if it fights not for the common safety but for its own advantages, it is a
vice. It is not merely not virtuous; it is rather a savagery which repels all civilized
feeling. Therefore the Stoics define courage well when they call it the virtue which
fights on behalf of fairness. (62ff.)³⁰

³⁰ Sed ea animi elatio quae cernitur in periculis et laboribus, si iustitia vacat pugnatque non pro
salute communi sed pro suis commodis, in vitio est; non modo enim id virtutis non est, sed est potius
immanitatis omnem humanitatem repellentis. Itaque probe definitur a Stoicis fortitudo cum eam
virtutem esse dicunt propugnantem pro aequitate (Off. I 62).
       ’    243

The special status of justice among the virtues is emphasized by Cicero in many
places, also and especially in book II, where he talks about winning the love and
support of other people. In book III (28) he calls justice domina et regina virtutum,
‘the mistress and queen of virtues’. But its role as the source of moral obligations is
actually somewhat obscured by Panaetius’s introduction of the concern for the
community of men as simply one of four resources for humans as rational beings.
The older Stoic theory of οἰκείωσις had only two natural instincts, common to
humans but also to most other animals. Justice was founded on the social instinct,
and justice in the sense of obedience to the laws of Nature could be seen as the
basis of all moral virtues. But Panaetius was writing for an audience of educated
laymen, not for philosophers. The theory of the laws of nature that we can find in
Cicero’s De Legibus I would, no doubt, have been too much for his readers, and the
Stoic account of wisdom as the knowledge of things divine and human which
includes the account of the community of gods and men is only briefly mentioned
in Off. I 153—mainly to insist that action is more important than the theoretical
study of nature. Panaetius avoided the complicated terminology of Stoic ethics
and used the commonsensical distinction between what is useful and what is
honourable, to the horror of some fellow Stoics—but then we are told that even
Chrysippus sometimes accepted the language of the vulgar.³¹ The ‘natural
resources’ for magnanimity and the desire for beauty in human nature could be
easily understood. But the desire for leadership and great deeds, and also the
desire for a good reputation, need to be restrained by justice, while in the older
theory courage could be defined as based on justice. Also, by replacing courage by
magnanimity, temperance by decorum, Panaetius gave the Stoic theory a
decidedly male and aristocratic turn. Women were definitely not expected to try
for great deeds, let alone self-determination, and in the long chapter on the
decorum they are mentioned only once, in a clause that states that loveliness in
outer appearance is appropriate for women, while men should show dignity
(Off. I 130). Panaetius’s ‘starting points’ should be compared to what Aristotle
describes as ‘natural virtues’—character traits that dispose some people especially
to develop certain virtues, such as bravery, but that need to be perfected by
practical wisdom (EN VI ch. 13)—or, in the Stoic context, by justice—to make a
person truly virtuous.
To return to the decorum as the expression of a virtue: what Cicero, and
presumably also Panaetius (in a less specifically Roman version) lists under the
label of decorum/πρέπον is in fact a rather mixed bag. Arguably, one might say
that respect for others is not just fitting, but required or obligatory—a requirement
that stems from justice as much as from the desire to be liked by others. But much
of the advice for aspiring gentlemen, whom Cicero includes in this chapter, can

³¹ See, for example, Plutarch, De Stoicorum repugnantiis 1048A.

244    

hardly count as moral instruction, though it may be quite sensible in terms of

advancing one’s social career. So, for example, choosing an occupation for which
one has no talent is surely not a good idea, but it is simply a mistake, not a crime
and not even an offence (119–120). Elegant style and good taste, again, are no
doubt desirable advantages, but lacking them does not make one an outcast.
Striving for the ideal of integrity or consistency between one’s words and actions
may help one to become virtuous, but the desire, however natural, to please others
is, as it were, morally neutral in itself. It may lead to politeness, tact, and patience,
but it may also turn into flattery. In other words, the two forms of decorum that
Cicero distinguishes, general and special, need not necessarily go together. It may
be no accident that Seneca in his epistle adopts the analogy between outer and
inner beauty in explaining how we arrive at our first idea of the good that is virtue,
but not, as far as I can see, the special version of decorum as based on the desire for
a good reputation.
Where does this leave us with respect to decorum as a virtue? We might
perhaps say that integrity is indeed a moral virtue, but though good manners
are pleasant and attractive, they cannot by themselves count as a virtue. Or we
might accept Hume’s more liberal view and count good manners and good taste as
a virtue—just not a moral one.
 Index of Names

Ackrill, John Lloyd 104 Caujolle-Zaslawsky, Françoise 186 n.9

Aenesidemus 177, 180, 186, 188–9 Chrysippus 152, 153, 168 n.12, 171, 174, 214,
Agrippa 181, 185, 186, 190, 191 236, 239, 243
Alexander of Aphrodisias 4, 10 n.18, 11, 12, 14 n.22, Cicero x, 129 n.2, 152, 153, 159
16, 34, 39, 40, 49 n.5, 62, 70, 72, 90 Academica (Varro) 162, 188
Ammonius 56, 59 Catulus 163
Annas, Julia 235 n.16 De Finibus 154, 155
Antiochus 162–9 passim, 173, 175, 177 De Natura Deorum 197
Antipater of Tarsus 157, 230, 239 De Officiis 143, 154–8, 218, 222–44
Antipho 135 n.5 Lucullus 162–74 passim, 176 n.19, 177, 190
Apellas 186 n.9 Tusculan Disputations 211, 212
Apollodorus of Seleucia 153 Clitomachus 179
Arcesilaus 162, 169, 172, 173, 177, 180, 188, Cooper, John 31 n.21, 113
189–90 Corcoran, John 6 n.10, 36 n.7, 50, 60 n.1
Aristo of Chios 153, 171–2, 173 Crates 153
Aristotle ix, x, 179, 182, 185, 194, 198, 206, 218, Crubellier, Michel 89 n.1
219, 237, 243
Analytica Posteriora 19, 36, 37, 38, 71, 72, Dardanus 163
98–101, 114 Democritus 181, 193, 194, 196, 199, 202, 209
Analytica Priora 1–77 passim, 84–101, 114 Descartes, René 173, 178, 185
Categories 102–11 Diodorus Cronus 216
De Anima 112, 131 Diogenes Laertius 179, 187, 194–9 passim
De Interpretatione 99 Diogenes of Babylon 157, 230, 239
De Motu Animalium 147 Dyck, Andrew 223, 240
Ethics (Eudemian and Nicomachean) 117,
121, 128–41, 142–51, 159 Ebbinghaus, Kurt 9
Metaphysics 109, 110 Epictetus 153, 218, 220
Physics 30 Epicurus 179, 182, 193–209, 221
Politics 128–9, 135–7, 141, 149 Euclid 55 n.16
Rhetoric 71, 73, 77, 81–4, 97, 112–27, 148–9, Eudemus 22, 35, 62
150, 158, 159
Sophistici Elenchi 74, 80, 109 Fine, Gail 103
Topics 41, 42, 48, 71, 72, 77, 78–81, 88, 96, Fortenbaugh, William W. 114 n.7
102, 110, 112, 114, 144 Frede, Dorothea 117
Frede, Michael 10 n.17, 169 n.13, 184 n.7
Barnes, Jonathan 19, 21, 23 n.8, 29 n.19, 32
Becker, Albrecht 22, 24, 35–6, 44 Galen 180
Bochenski, Józef Maria 3, 35 Gauthier, René Antoine 143, 145, 146
Boethus of Sidon 47, 56–9 passim Geach, Peter Thomas 45
Bolton, Robert 70 n.15 Glucker, John 162
Brunschwig, Jacques 79
Burnyeat, Myles 75, 82 n.3 Hecato 157
Byrne, Patrick H. 89 n.1 Heraclides 180
Heraclitus of Tyre 164
Carneades 157, 162, 165, 168, 173–7 passim, 179, Hermias 56
180, 188–91 passim, 230, 239 Hintikka, Jaakko 22
Cato 230, 231, 239 Hirzel, Rudolf 164 n.4
252   

Hobbes, Thomas 242 Laches 229

Hume, David 236–7, 242, 244 Meno 2, 3 n.4, 66, 67–8, 80, 197–8
Parmenides 105, 106, 144
Iamblichus 56 Phaedo 105, 106, 107
Ioli, Roberta 182 n.4 Philebus 106, 114, 115, 116, 117, 132
Ioppolo, Anna Maria 172 Protagoras 151
Irwin, Terence 136 n.7, 236 Republic 115, 116, 122, 135, 142
Theaetetus 196
James, William 192 Plutarch 124, 188 n.11, 189, 201, 216 n.7
Joachim, Harold H. 19 Porphyry 56
Jolif, Jean-Yves 143, 145, 146 Proclus 56
Protagoras 151, 186 n.10
Kant, Immanuel 133 Pseudo-Galen 168 n.12
Kenny, Anthony 145 Pyrrho of Elis 180, 184, 193
Kneale, Martha 5 n.7, 8, 37 n.9, 42, 45
Kneale, William 37 n.9 Reid, James S. 166 n.8
Ross, William David 4, 8, 9
Lear, Jonathan 61
Lee, Tae-Soo 48 n.3, 57 n.20 Saturninus 180
Leighton, Stephen 116 n.10 Schofield, Malcolm 207
Locke, John 107 Scholz, Heinrich 36
Lucretius 196, 199, 201–2, 206 Sedley, David 164 n.4, 216
Łukasiewicz, Jan 35, 47, 86 Seneca 210–21, 228, 233–4, 236, 241, 244
Sextus Empiricus 164, 166 n.9, 167, 174, 175,
McCall, Storrs 24 178–92 passim, 200, 204
Maier, Heinrich 8, 16 Shorey, Paul 2, 8
Mansfeld, Jaap 189 n.13 Sigwart, Christoph von 8
Maximus 56–9 passim Simplicius 104
Mendell, Henry 67 n.12, 73 n.20 Smiley, Timothy 1 n.1
Metrodorus 164 Smith, Robin 36 n.7, 50, 53 n.13,
Mignucci, Mario 8, 26 n.16 55 n.16, 61
Mnesarchus 163 Socrates 66, 67, 79, 80, 105, 107, 188, 218,
Mueller, Ian 51 n.9, 56 n.16 229, 239
Solmsen, Friedrich 19, 78 n.1
Nortmann, Ulrich 36 n.7, 37, 38, 39, 44 n.18 Sphaerus 171, 172
Nussbaum, Martha 210 Stobaeus 225
Syrianus 56
Pacius, Julius 8
Panaetius 143, 153–9 passim, 212 n.3, 217, 222–44 Tarrant, Harold 162, 165 n.6
Patterson, Richard 37–8 Themistius 56–7, 59
Patzig, Günther 9 n.15, 48, 49, 53 Theophrastus 16 n.25, 22, 35, 62, 94
Philo of Larissa 162–6 passim, 175–7 Thom, Paul 94, 111 n.19
Philodemus 167, 207 n.5 Timon of Phlius 180
Philophonus 4, 11, 12, 35, 39, 89 n.1, 96, 99 n.11
Places, Édouard des 144 van Rijen, Jeroen 37, 38, 42 n.15, 45
Plasberg, Otto 166 n.9
Plato ix, 79, 101, 102–11 passim, 123, 131, 132, Wesely, Michael 89 n.1
136, 144, 180, 181, 182, 194, 199, 209, 219 Williams, Bernard 136 n.7
Gorgias 212
Hippias Maior 229 Zeno of Citium 153, 164, 169, 177, 239
 Index of Passages

As these passages appear in essays that were published in a number of different journals over many
years, different styles of citation have been used in the text. However, for simplicity and ease of use, all
citations have been consolidated under one style in this index.

Alexander of Aphrodisias 31,13–25 56

In Aristotelis Analyticorum Priorum 31,18–22 59
Librum I Commentarium 31,28–33 51
1,1–6 61 n.4 33,18–21 56
1,12 61 n.4 45,42–46,2 22
2,33–4,29 62 n.6
Antipho
7,11–22 90
On Truth
21,12–18 55
DK 87B44 135 n.5
24,12–18 12
113,5–9 48 Aristotle
124,8–17 35 Categories/Categoriae
124,8–125,2 35 n.2 2 102, 103, 110
140,3 39 2.1a20–b6 103
144,18 39 3.1b11–12 104
144,32–145,4 34 n.1 5 102, 103, 104, 107, 110
171,30 40 5.2a19–21 106
173,33–174,3 35 n.2 5.2a34–2b2 107
207,28–36 34 n.1 5.3a21–29 108
213,11–27 34 n.1 5.3b10–21 109
215,8–10 40 5.4a10–22 108
216,24–26 34 n.1 Eudemian Ethics/Ethica Eudemia
220,9–16 35 n.2 1.2,1214b10–11 140
234,4–14 35 n.2 1.6,1217a10–17 75 n.22
255,1–17 49 n.5 2.2.1220b13 118
256,18–25 7 n.11 2.10 143
265,30 14 n.22 2.10,1226b17 144
265,30ff. 72 n.19 3.1,1229a31–1229b22 122
301,12ff. 70 8.2,1247a32–35 27
318,17ff. 12 Metaphysics/Metaphysica
324,19f. 14 n.22 A 1,981a12–24 139
324,22–24 10 n.18 Γ 2,1004b24–25 146
379,9–11 94 Γ 2,1004b27–1005a2 53
387,1–5 4 Γ 3,1005b7 54 n.15
387,1–11 67 Γ 4,1005b26–29 53
387,5–11 4 Γ 5,1009b12 150 n.7
In Aristotelis Topica commentaria Δ 2,1013b20 55
324,24–31 16 Δ 2,1013b20–21 89 n.1
324,29–30 16 Δ 12,1019b23–24 28 n.18
325,3–8 16 Δ 28 111
Ammonius Δ 28,1024a29–b9 109
In Aristotelis Analytica Priora Commentarium Δ 28,1024b9–16 53 n.12
14,28–29 56 Δ 30,1025a6 19
254   

Aristotle (cont.) 5.1,1129a34 134

E 2,1027a19–28 18 5.1,1129b11–25 134
E 3,1027b14–15 54 n.14 5.1,1129b19–25 136
Z 1 110 5.1,1130b22–24 136
Z 1,1028a33–34 111 5.8,1135b19–27 146
Z 1,1028a34–36 110 5.8,1136a1 146
Z 3,1029a9–30 107 n.11 6 130, 132, 147, 148, 149,
Z 3,1029a30–32 110 n.17 150, 158
Z 13 111 6.1,1138b32–34 132
H 1 111 6.2 144
H 1–2 110 6.2,1139a23 144
H 1,1042a26–31 110 6.2,1139a29–31 132
Nicomachean Ethics/Ethica Nicomachea 6.2,1139b5 144
1.1,1094a23–24 225 n.5 6.5,1140a25–28 148
1.2,1094b10–11 128 6.5,1140b4–5 148 n.5
1.3,1094b11–27 129 6.5,1140b7–11 148
1.4,1095a5–6 130 6.7,1141b14–22 139
1.4,1095a31–b8 130 6.8,1141b23ff. 128, 148
1.4,1095b4–7 130 6.8,1141b27 150
1.5 133 6.8,1141b29–1142a10 137
1.7,1097b8–11 132 6.8,1142a10 149
1.7,1098a16–18 132 6.8,1142a10–11 137
1.13,1102a7–10 136 6.8,1142a14–30 147
1.13,1102a7–13 128 6.8,1142a20–23 134
1.13,1102a23–26 114 n.5 6.8,1142a23–30 133
1.13,1102b25–1103a3 126 6.9,1142a312 149 n.6
1.13,1103a3–10 112 6.11,1143a32–b5 148
2.1,1103a18–26 27 n.17 6.11,1143b11–14 149
2.2,1103b34–1104a11 129 6.11,1143b13–14 139
2.3,1104b8–27 131 6.11,1143a35–b14 133
2.5,1105b23 118 6.12,1144a23–29 159
2.5.1105b19–1106a4 122 n.17 6.12,1144a23–b1 151
2.5,1106a2–4 147 6.12,1144a29–1144b1 125
2.6,1106b21–22 139 6.12,1144a31–36 125
2.6,1107a1 126 6.13 243
3 150 6.13,1145a6–9 137
3.2 143 7 117
3.2,1111b5–6 143 7.12,1152b33–1153a2 117
3.2,1112a16 150 7.12.1152b36 117
3.3 143 8.1,1155a22–28 138
3.3,1112b20–25 149 10 117
3.3.1112b21–23 149 n.6 10.9,1179a33–35 132, 146
3.3,1112b33–1113a2 147 10.9,1179b23–31 131
3.3,1113a3–7 144 10.9,1179b34 129
3.3,1113a11 144 10.9,1180a4–5 138
3.5,1113b19–20 54 n.14 10.9,1180a34–b2 138
3.6,1115a10ff. 122 10.9,1181b15 141
3.8,1117a17–22 148 On the Generation of Animals/De
3.10,1118a24–25 117 Generatione Animalium
3.11.1119a4 117 Δ 8.777a18–20 27
4 135 E 8.789a8–b8 31
4.9,1128b12–13 122 On Generation and Corruption/De
5 134–5, 137, 159 Generatione et Corruptione
5.1 128 2.2,329b15ff. 53
5.1,1129a26–1130a9 134–5 2.6,333b5 27
    255

On Interpretation/De Interpretatione A 12,78a6–13 75 n.22

9 18 A 22,83a1–18 34, 41
On the Motion of Animals/De Motu A 22,83a1–23 85
Animalium A 22,83a5 71 n.16
7,701a7–12 147 A 22,83a14–17 71 n.16
7,701a32–33 147 A 26,87a1–17 68 n.13
On the Parts of Animals/De Partibus A 26,87a14–17 6
Animalium A 26,87a20–30 69
2.1,642a34–36 30 n.21 A 32,88b7 20, 38
2.13,657a25ff. 31 B 5,91b12–14 101
On Sophistical Refutations/Sophistici Elenchi B 5,91b13 98
6,168a17–20 53 B 6,92a7ff. 15
6,168a21–23 80 B 6,92a20ff. 15
22,178b36–179a10 109 B 6,92a21 15 n.23
On the Soul/De Anima B 11,94b37–95a2 32 n.22
1.1,403a29ff. 112, 114 n.5 B 11,95a2 31
2.1,412a6–9 110 B 12,96a10–11 26
2.2,413b10–13 131 Prior Analytics/Analytica Priora
2.3,414b19–415a13 131 A 1–2 47
Physics/Physica A 1–7 63
B 3,194b22–23 54 n.14 A 1–22 62
B 3,195a16–19 89 n.1 A 1,24a13 58
B 3,195a18–19 55 A 1,24a16–17 100
B 5,197a19–20 18 A 1,24a16–22 84
Β 8,198b36 27 A 1,24a22–b15 62
B 8,199a10 29 A 1,24a25 11
B 8,199a33–35 177 n.21 A 1,24a28–b3 84
B 8,199b18 29 A 1,24a30–b15 3
B 8,199b26 29 A 1,24b16–20 85
B 9,199b34–35 30 n.21 A 1,24b18–20 1 n.1, 63
B 9,200a15–30 55 A 1,24b20 66
Θ 4,255b7 29 A 1,24b21 13
Θ 4,255b10–12 29 A 1,24b21–22 51
Θ 4,255b19–21 29 A 1,24b23–24 49, 50
Politics/Politica A 1,24b24 11
1.2 136, 137 A 1,24b27–29 58
1.2,1253a9ff. 226 A 1,24b28–30 56 n.16
3.4,1277b7–32 136 A 2 63, 65
3.4,1277b25–30 149 A 2,25a5–14 85
3.9,1280b7–12 137 A 2,25a14–17 13
3.9,1280b35–39 137 A 3 40
3.9,1280b38 147 A 3,25b14–15 25
Posterior Analytics/Analytica Posteriora A 4 47, 49, 57, 58, 63, 70, 89, 98–9
A 1–4 37 A 4–6 11, 48, 70, 89 n.1, 99
A 1,71a1–2 185 A 4–7 47
A 2,71b22 55 A 4,25b9–40 58
A 3 190 A 4,25b26–31 89
A 3,73a7–11 64 A 4 25b28–31 62
A 4 38 A 4,25b28–31 84
A 4–6 37 A 4,25b32–35 56 n.16, 57
A 6,75a1–10 36 n.7 A 4,25b39–40 11
A 10,76a37–b2 16 A 4,26a24 11
A 10,76b10–22 13 n.20 A 4,26a27 11
A 11,77a22 9 n.15 A 4,26a27–28 58
A 11,77a22–24 16 A 4,26b24–25 42
256   

Aristotle (cont.) A 15,34a17–18 63

A 4,26b30 50 A 15,34b5–18 34
A 4,26b31–33 41 A 15,34b7–18 24, 45
A 5 6, 14, 63, 65, 69 A 15,34b9 25
A 5–6 50, 51, 52 A 15,34b12 25
A 5,27 11 A 15,35a2 41
A 5,27a16–18 50 A 16,35b23–26 44
A 5,27a17 11, 51 A 16,36a6–7 50
A 5,27a36–b1 5, 50 A 16,36a29–30 43
A 5,28a4–9 51 n.9 A 16,36b12–18 43
A 5,28a6 11 A 17.37a7–9 40 n.12
A 5,28a6–7 11 A 17.37a9–31 34
A 6 6, 14, 51, 63, 65, 69, 89 n.1 A 20,39b4–6 40 n.13
A 6,28a24–26 51 A 23 1–17 passim, 63, 64–5, 67, 70 n.15, 72,
A 6,28b14–15 57 n.21 95, 97, 99
A 6.28b17–20 5 A 23–46 91
A 6,28b20 12 A 23,40b17 49 n.5
A 6,29a9–10 42 A 23,40b20–22 62
A 6,29a14–18 51 n.9 A 23,40b23 2
A 6,29a15–16 50 A 23,40b23–24 10
A 7 2, 49 n.5, 51 n.9, 54, 55, 63, 86 A 23,40b25 2
A 7,29a30–31 48, 51 n.9 A 23,40b35 63
A 7,29a30–34 11 A 23,40b35–36 86
A 7,29a32 2 A 23,41a23–41b1 7
A 7,29b1 48 A 23,41a24 8
A 7,29b7–8 57 n.21 A 23,41a24–25 66 n.11
A 8 36 n.7 A 23,41a28 8
A 8,30a9–14 52 A 23,41a32 7, 8
A 9 36, 38 A 23,41a32–34 66 n.11
A 9,30b2 59 n.26 A 23,41a38 10 n.18
A 9,30b5–6 39 A 23,41b3–5 49 n.5
A 9,30b35 39 A 24 1, 6
A 10 41 A 25 51
A 10,30b37–38 44 A 25,42a32 50
A 10,31a14–17 39 A 27–29 86
A 11,31b8–10 41 A 27–31 70, 72
A 13 18–33 passim, 38 A 28,44a37 70
A 13,32a16–b23 20 A 29,45a23–28 68
A 13,32b4–18 43 A 29,45b1–3 6 n.10
A 13,32b4–23 20 A 29,45b8 2
A 13,32b6 23 A 29,45b14 2
A 13,32b7 26 A 29,45b17 14 n.22
A 13,32b13–18 22 A 29,45b18 10 n.18
A 13,32b18–20 38 A 29,45b19–20 80
A 13,32b18–22 18, 22, 36 n.7 A 29,45b36–46a2 71
A 13,32b21–22 43 A 30 72
A 13,32b23–32 24 A 30,46a17–30 87
A 13,32b25–32 44 A 30,46a22–28 72
A 14 25 n.13 A 31 98 n.10, 100
A 14,32b40–33a5 58 A 31,46a31 70
A 14,33a24–25 58 A 32 52, 89, 96, 97
A 14,33a27 59 n.26 A 32–45 52
I 14,33b6–9 40 A 32–46 72, 88, 97
A 15 30, 38 A 32,46b38–47a2 62, 74
    257

A 32,46b40–47a4 52 B 14,62b29 2
A 32,47a1–9 89 B 14,62b29–31 68
A 32,47a5–8 72 B 14,62b32–35 8
A 32,47a22–35 80 B 14,62b33–37 5
A 32,47a28–30 91 B 14,62b35–37 8
A 32,47a37–40 90 B 14,62b38 2
A 33 94 B 14,62b38–40 69
A 33–46 91 B 14,63b12–13 69
A 34 91, 94 B 14,63b12–16 68
A 34,47b40–48a15 108 n.13 B 15 69
A 34,48a1–15 92 B 15,64b17–27 69
A 34,48a20–22 92 n.4 B 17 7
A 35 93 B 22 72
A 36 71 n.16, 73, 92, 94 B 23 97
A 36,48a40–b4 92 B 23,68b9–13 74, 97
A 37 92 B 25 68, 73
A 37,48b4–9 92 B 25,69a23–24 68
A 37,48b10–14 93 B 25,69a24–27 68
A 38 94 B 26.69b13ff. 4
A 41 94 B 27 73, 83 n.4
A 42,50a5–10 99 B 27,70a3–38 97
A 42,50a8 97 n.7 Rhetoric/Rhetorica
A 44 1–17 passim, 32, 64–6, 94, 95, 96 1 112, 115–21
A 44,50a16–28 95 1.1,1354a1 81
A 44,50a16–b4 80 1.1,1354b31 81
A 44,50a19–20 14 1.1,1355b17–21 146
A 44,50a19–26 4 1.2,1355b 81
A 44,50a23–28 7 1.2,1355b16–17 71 n.18
A 44,50a24–26 4, 80 1.2,1356a 81
A 44,50a25–26 66 1.2,1356a1–20 115
A 44,50a30–32 7 1.2,1356a22 71 n.18
A 44,50a31 8 1.2,1356a25–27 71 n.18
A 44,50a34–35 10, 14 1.2,1356a35–b7 82
A 44,50a35–38 8 1.2,1356b9 114 n.4
A 44,50a39–b4 95–6 1.2,1356b12 114 n.4
A 44,50b2–3 32, 62, 81 1.2,1356b16–19 82
A 45 52, 53, 54, 57, 90 1.2,1356b22–24 75
A 45,50b8–9 52 1.2,1357a22–b5 43 n.17
A 45,51b3–5 96 1.2,1357a29–b20 82
A 46 94, 96 1.2,1357a30 114 n.4
B 100 1.2,1357a34–b1 26
B 2,53b18–20 63 1.2,1357a36 26
B 2–4 75, 86, 101 1.2,1357b1–1358a2 98 n.8
B 8–10 75 1.2,1357b10–20 83
B 11 6 1.2,1357b22–25 114 n.4
B 11–14 5, 7, 14, 68–9 1.2,1357b31–35 82
B 11,61a24 5 1.2,1358a10–17 72 n.19
B 11,61a25 5 1.2,1358a10–28 83
B 11,61a27–31 5 1.2,1358b20–25 150
B 11,62a13 9 n.15 1.3,1358b8–10 150
B 11,62a13–15 5 1.4,1359a38 54 n.14
B 11,62a14–15 9 n.15 1.4,1359b8–12 71 n.18
B 11,62a15–17 68 1.4,1359b10 55 n.16, 98
B 11,61b14 5 1.4,1359b20ff. 148
258   

Aristotle (cont.) 1.18 15

1.4,1360a37 149 1.18,108b12–19 79
1.9 115 n.9 1.18,108b12 2
1.9,1366b3–7 159 1.18,108b12ff. 14 n.22, 15
1.9,1367a6–8 159 1.18.108b13 15 n.23
1.10–11 115 1.18,108b16–17 10
1.10,1368b32–1369a7 116 2.2,109b35–110a4 114 n.6
1.10,1369a2–3 116 2.6,112b5–9 26
1.10,1369a18 116 2.6,112b11–12 27 n.17
1.10,1369b15–16 119 n.14 2.7,113a35–113b3 114 n.6
1.10,1369b31–32 114 3.6,119b21–24 14
1.10,1369a34–b2 27 3.6, 119b35 14, 15 n.23
1.11 115, 116, 117, 119, 121 3.6, 119b35ff. 15
1.11,1370a14–15 119 4.1,121a11–12 106
1.11,1370a16–18 117 4.5,125b20–27 114 n.6
1.11,1370a17–18 119 n.14 4.5,125b28ff. 114 n.6
1.11,1370a18–27 115 n.8 4.5,126a6–12 114 nn.6 and 7
1.11,1370a32–1370b10 119 4.5,126a17–25 106
1.11,1370b8 119 n.14 4.5,127b27–32 114 n.6
1.11,1370b13 119 n.14 5.4,132b35–133a11 106
1.11,1370b33 119 n.14 6.13,151a14–19 114 n.6
1.11,1370b33–34 117 7.5.155a11–16 41
1.11,1371a9 119 n.14 7.5.155a31–36 41
1.11,1371a19 119 n.14 8.1,156a32–33 114 n.6
1.11,1371a23–24 119 n.14 8.2,157b34–158a2 6, 70 n.15
1.13,1374a9–16 146 8.11,162a8–11 75 n.22
2 112, 113, 115–21 8.14,163a32–36 75
2.1 116 n.10 Cicero
2.1,1378a4 116 Academica
2.1,1378a16–19 115 n.9 1.13 165 n.5
2.1,1378a20ff. 115 1.32 166 n.9
2.1,1378a20–23 119 1.35 162
2.1,1378a21 118 I 45 188
2.2–11 115 De Finibus
2.5,1383a6–7 125 1.30–31 199 n.3
2.7 119 n.15 3.33 152
2.7,1385a18 119 3.58 152
2.9 114 n.6 3 154, 155, 223, 225, 230,
2.25 73, 98 231, 235
2.25,1402b13ff. 26, 36 n.7 3.6–19 225
2.25,1402b13–1403a5 43 n.17 3.20–21 236
2.25,1402b25–1403a2 73 3.21ff. 230
2.25,1402b35–1403a1 27 3.57 238–9
2.25,1403a4–5 114 n.4 3.60–61 153
2.25,1403a12 114 n.4 3.62–68 225
Topics/Topica 3.73 235 n.15
1.1,100a18–21 78 4 230
1.1,100a25–26 78 4.76 177 n.20
1.4,101b13–25 71 n.16 4.26–28 230
1.6,102b27–103a5 71 n.17, 79 4.79 224 n.2
1.6,102b36–38 71 n.17 5.88–89 233
1.7,103a23–31 109 n.15 De Legibus
1.7,103a30 109 I 18 152
1.8,103b2–19 71 n.16 De Natura Deorum
1.14,105b20–25 88 1.44 197
    259

De Officiis 3.13–16 218

1 154, 161 n.16, 231 3.17 234
1.7 223, 224 3.28 243
1.8 152, 224 3.50–53 157
1.9 155, 224, 232 3.89–90 157
1.10 224 3.90 157
1.11–17 225 De Re Publica
1.13 155 III 51 157 n.13
1.14 212 n.3, 225, 226, 231, 238 Epistulae ad familiares
1.15 154 9.8 164 n.5
1.18–19 155 n.11 Lucullus
1.20 156 11 164
1.21 156 12 162, 163
1.38–41 156 18 163
1.42 156 23–24 165 n.7
1.46 218 32–33 165
1.61ff. 227 32–36 165, 166
1.61–91 155 n.11 33–34 167
1.62ff. 242 33–36 166 n.9
1.66–67 231 36 165 n.7, 166 n.9, 167
1.68 227 40 173
1.69–73 129 n.2 46 173
1.93 227 47 173
1.93–151 237 48 170
1.94–95 228 49 173
1.96 229, 237 50 170, 173
1.97–98 229 51 173
1.98 238 n.22 51–52 171
1.99 238 n.23, 240 51–53 170
1.101–102 232 54 170, 171 n.15
1.101–106 241 56–57 170
1.107 241 57 170, 172
1.110 241 58 166 nn. 8, 9, 173
1.119–120 244 60 190
1.128 153, 239 69 163
1.130 243 77 169, 177
1.132–135 241 78 164, 169
1.145 233 84 166 n.9, 173 n.16
1.148 153, 239 84–86 170
1.153 243 85 170, 173
2 154, 227, 232, 243 86 173
2.1 158 88–90 170
2.9 156 93 174
2.9–10 232 99 173
2.11 232 105 175 n.18
2.19 158 108 190
2.33–34 158 111 166, 173
2.35 159 112–113 164 n.5
2.86 158 113 177
3 155, 157 146 165 n.5, 176 n.19
3.7 155 Orator
3.7–10 155 22.73–74 240
3.11–12 231, 232 Tusculan Disputations
3.13 232 IV 29–32 212
3.13–15 233 IV 30 211
260   

Cicero (cont.) Lucretius

IV 31 212 De rerum natura
V 43–53 233 4.256–258 201
Clement of Alexandria 4.362–363 201
Stromata 4.379–387 196
II 21 230 n.11 4.385 199
4.466–467 202
Democritus 4.486–499 196
fr. B 11 202 6.703–711 206
fr. B 125 193
Philodemus
Diogenes Laertius De Signis
Lives and Opinions of Eminent Philosophers I 1–17 167
1.16 179
7.87 152 Philoponus
7.87–88 236 In Aristotelis Analytica Posteriora
7.89 152 n.8 commentaria
7.94 152 334–5 99 n.11
7.107 152 In Aristotelis Analytica Priora
7.121 153 Commentaria
7.127 171 n.14 5,15–6,1 96
7.162 171 37,16–38 12
7.177 171 37,29 12
9.88 186 n.9 358,27–31 4
9.106 186 n.9 Photius
9.107 187 Bibliotheca
9.115–116 180 cod. 212, 170a16 177
10.28 208 Plato
10.31 193, 203 Gorgias
10.31–32 194–5 464B–465D 212
10.32 196 Hippias Maior
10.33 196–7 293Dff. 229
10.35 194 Laches
Letter to Herodotus 194Dff. 229
10.37–38 197 Meno
10.38 194, 203 75B–76A 198
10.39 205 80D 197
10.40 205 86Eff. 2
10.46–48 202 86E–87A 67
10.51 203 87B–96C 80
10.67 206 Parmenides
10.72–73 207 129A 106
10.82 194, 205 129C 106
Letter to Pythocles 130C4–D8 105
10.113 206 143C2 144
Letter to Menoeceus Phaedo
10.132 199 74A9–C10 105
10.133 208 102B–103A 92 n.4
10.147 195 103A4–22 105
Epictetus Philebus
Dissertationes 16C5–E2 106
III 22 153 21D6–E5 132
IV 2.10 213 31Dff. 116
Encheiridion 47C–50C 117
48.2–3 218 47C–50E 114
    261

Protagoras 7.163 168

327E–328A 151 7.163–164 168
Republic 7.168 174
2,358B–367A 135 7.184 168
10,596A6–7 105 n.6 7.203–216 200
Theaetetus 7.211–216 203–4
184B–186E 196 7.206–209 200
Plutarch 7.252 169, 174
Against Colotes/Adversus Colotem 7.408 174
1121B–D 201 7.410 174 n.17
1122A–F 189 7.427–429 174
De cohibenda ira 8.473–475 191
453D–E 188 n.11 11.66 153
De Stoicorum repugnantiis Outlines of Pyrrhonism
1043A 216 n.7 1.1–4 178–9
1048A 243 n.31 1.8 182
De Virtute Morali 1.9.21–24 181
445A–447C 124 1.9.24 182 n.4
449A–449B 124 1.9.25–30 181, 183
Moralia 1.16–17 181
75D (SVF III 539) 216 n.7 1.18 181
1.22 184
Seneca 1.23 184
De vita beata 1.29 187
3.3–4 216 1.33–34 191
4.4–5 216
1.145 182 n.4
15.2 216 1.164–177 185
17.3–4 219 n.9 1.180–185 188
Epistulae morales ad Lucilium
1.203 187
9.21 216
1.206 187
52.2–3 221 1.232–233 189
59.16 234 n.14 1.235 164
75 214–15
1.236 185, 188
75.9–14 214 2.10 184 n.7
75.10 215, 216 3.246 153
75.13 215 n.5 4 182
75.16 215
85.10ff. 215 n.5 Simplicius
120 212, 228, 233, 236 In Aristotelis Categorias Commentarium
120.4–5 212 n.3 45,19–23 104
120.5 233 402,22–26 171 n.14
120.11 213 n.4, 228 Stobaeus
120.12–13 213 Eclogae
120.19 213 II p. 63, 10–64, 12 Wachsmuth-Hense
120.22 213, 241 n.29 225 n.5
124.14 236 Themistius
Sextus Empiricus Reply to Maximus, French translation
Against the Mathematicians/Adversus 166,28–32 57
Mathematicos 167,2 56
1.61 180 167,40–168,5 57
7.139 202 168,30–36 56
7.154 173 169,13 57
7.158 172, 189 169,16–17 57
7.159–189 168 170,4–5 57
262   

Themistius (cont.) 175.34 57 n.21

170,27–171,26 57 176,9–10 57
171,3–5 57 176,11–12 57
171,38–172.8 59 177,2–3 57
172,8–10 58 n.23 177,4 57
172,25–26 57 177,5–7 59
173,34–35 57 177,7ff. 57 n.21
173,35–38 59 178,11–12 57 n.22
173,41–44 57 n.20 179,22 57
174,17–18 57 Theophrastus
174,35–39 57 Topics
175,18–29 57 n.20 5.21–26 16 n.25