Barnes-The Toils of Scepticism

Epistemology
EDITED DY Stephen E\'crson
This volume, the first in the Companions to Ancient Thought series, deals
with Epistemology. The period from the sixth century B.c. to the second and
third centuries A.O. was one of the most fertile for the theory of knowledge, and
the range of'epistemic states' explored in the ancient texts is much wider than
those to be found in contemporary discussions of epistemology or cognition.
Greek philosophers approached these problems in a great variety of ways,
from the extreme relativism of Protagoras to the scepticism of the Pyrrhonists,
and the contributors demonstrate both the familiarity and novelty of this
range of views in their critical essays.
Contents: The beginnings of epistemology: from Homer to Philolaus,
EDWARD HUSSEY; Protagoras and self-refutation in Plato's Theaetetus, MYLES
BURNYEAT; Plato's early theory of knowledge, PAUL WOODRUFF; Knowledge
and belief in Republic v-vu, GA! L FINE; Aristotle's epistemology, c. c. w.
TAYLOR; The problem of the criterion, GISELA STRIKER; Epicurus on the
truth of the senses, STEPHEN EVERSON; Stoic epistemology, JULIA ANNAS;
Some ways of scepticism, JONATHAN BARNES; An empiricist view of know-
ledge: memorism, MICHAEL FREDE
The Modes of Scepticism

Ancitnl Texts and 1\lodem !11ttr/Jrt/a/1011j
Julia Annas and Jonathan Barnes
The Ten Modes of Scepticism are one of the most important and influential of
all ancient philosophical texts. They made an enormous impact on Western
thought when they were rediscovered in the sixteenth century and they have
shaped the whole future course of Western philosophy. This book translates
the texts and supplies them with a discursive commentary, concentrating on
philosophical issues but also including historical material.
'The book opens up to debate numerous new or inadequately aired issues in
the early history of scepticism, and for a far wider audience than has hitherto
had access to them. What is more, the immense learning underlying it has
been distilled into a fluent and untechnical style of presentation which makes
it a joy to read.'
The Times Literary Supplement
'A lively and accessible introduction'

The London Review of Books
CAMBRIDGE UNIVERSITY PRESS
0-521-38339-0
I
9 780521 383394
II
The Toils of Scepticism
JONATHAN BARNES
Professor of Ancient Philosophy in the University of Oxford,
and Fellow of Balliol College
�CAMBRIDGE
� UNIVERSITY PRESS
Published by the Press Syndicate of the University of Cambridge
The Pitt Building, Trumpington Street, Cambridge CB2 1RP
40 West 20ch Street, New York, NY 10011-4211, USA
Contents
10 Stamford Road, Oakleigh, Melbourne 3166, Australia
© Cambridge University Press 1990
First published 1990

Reprinted 1994
Printed in Great Britain by

Antony Rowe Ltd, Chippenham, Wiltshire
British Library catnlog11i11g i11 p11l1/icatio11 data
Barnes, Jonathan, 1942- Introduction page Vtt

The toils of scepticism.
1. Greek philosophy, ancient period. Scepticism
I. Title
I Disagreement I
186
ISBN 0-521-38339-0 2 Infinite regression 36
Library of Congress cntalog11i11g i11 publication data 3 Reciprocity 58
Barnes, Jonathan. 4 Hypotheses 90
The toils of scepticism/Jonathan Barnes.
p. cm. 5 The sceptic's net IIJ
ISBN 0-521-38339-0
I. Scxtus, Empiricus. 2. Skepticism-History. I. Title.
B623.B37 1990
186'.1-dc20 89-27951 CIP
Note on the ancient authors
ISBN 0 521 38339 0 hardback Index of passages
Index of persons
Index of Greek terms
Index of subjects
SE
v
Introduction
Greek philosophy knew two main vanet1es of scepticism, one

taking its name from Pyrrho of Elis and the other associated with a
particular phase in the history of Plato's Academy. My concern in
this book is with Pyrrhonian scepticism, and I say nothing about
the Academic variety. For the sake of brevity, I usually leave the
words 'sceptic' and 'scepticism' without a qualifying adjective; it
should be understood that it is the Pyrrhonian sceptic and
Pyrrhonian scepticism to which I mean to refer.
Pyrrhonian scepticism had a long career, in the course of which it
assumed significantly different forms. For us, its chief representa
tive is Sextus Empiricus, the only Pyrrhonian whose works have
survived. I shall be concerned, almost exclusively, with the scepti
cism which is expressed in the pages of Sextus. Thus I ignore the
earlier phases of Pyrrhonism, which pose special problems both of
an historical and of a philosophical nature;1 and I ignore the few
texts outside Sextus which offer us additional evidence for the later
history of the Pyrrhonian philosophy.
Of Sextus himself we know little. 2 He was a doctor by pro
fession. He lived in the second century AD. He was a prolific author,
but not an original thinker. We possess three of his writings. One
offers an outline of scepticism, giving first a general account of the
nature of Pyrrhonism and then a survey of the arguments which the
Pyrrhonians advanced against the Dogmatists. ('The Dogmatists',
ol 5oyµcxTJKol, is the title by which the sceptics referred generically
1For Pyrrho himself (c.360-270 nc) sec the fundamental study by

Fernanda Decleva Caizzi, Pirrone - testimonia11ze (Naples, 1981).
2 See, most recently, D.K. House, 'The Life of Scxrus Empiricus',
Classical Quarterly 30, 1980, 227-38.
VII
Introduction Introduction
to non-sceptical philosophers.) The destructive arguments are somewhat less obscure figure of Aenesidemus.6) Again, my interest
arranged in three sections, corresponding to the three traditional focusses on a single facet of the Agrippan aspect of Sextus'
parts of Dogmatic philosophy, logic and physics and ethics. A Pyrrhonism. For it is the general form which Agrippan argumenta
second work collects a larger quantity of these destructive argu tion characteristically takes and the general structures which
ments, similarly organized in three sections. A third work consists Agrippan scepticism characteristically erects which constitute the
of six essays directed against six Dogmatic arts - grammar, rhet central theme of this book. And I pay little attention, except by way
oric, geometry, arithmetic, astrology, music. 3 In all three of his of occasional illustration, to the numerous particular instances of
works Sextus is largely concerned to assemble and arrange existing these forms and structures which occur throughout Sextus'
material: he draws.on - and, sometimes at least, actually copies writings.
from - earlier Pyrrhonian sources.4
It is a difficult question whether the three works present a single After such disclaimers, the reader may pardonably wonder if he has
and coherent form of Pyrrhonism. Some scholars find important not opened a book of piddling breadth, a learned monograph - or,
differences among the works, and some scholars find important at least, a monograph - which prescinds from everything which
differences within the works - differences which are perhaps to be made ancient scepticism a subtle and living philosophy, and which
explained by the interesting hypothesis that Sextus' own views limits its outlook by the close and narrowing blinkers of leathery
underwent some change or development, or else by the dispiriting scholarship.
hypothesis that Sextus nonchalantly drew on different sources in But I claim three things for my circumscribed subject. First, and
different parts of his writings. Such suppositions and hypotheses exegetically: that the forms and structures I discuss were among the
raise issues of scholarly significance, but in this book I pass them most important aspects of Pyrrhonism, so that to study them is to
by. study the soul of ancient scepticism. Secondly, and historically:
.
My particular subject is what I may call the Agrippan aspect of that these same forms and structures have had a unique influence
Sextus' scepticism, the aspect which in some fashion derives from on the subsequent history of sceptical enquiry, and hence, more
the shadowy figure of Agrippa. 5 Much of Sextus' work is Agrippan .generally, on the history of epistemology or the enquiry into the
in inspiration or colouring, but by no means all of his thought was nature and scope of human knowledge: the Agrippan forms lie at
moulded by Agrippa. (And so I shall say nothing about some of the the heart of the western philosophical tradition. Thirdly, and
most celebrated parts of Sextus' writings which derive from the philosophically: that these forms and structures remain today
among the central issues in the theory of knowledge; that every
3 The three works are: (1) Outlines of Pyrrhonism, in three Books (the
title is customarily abbreviated co PH); (2) a work in five Books, modern epistemologist must take notice of them; and that they still
normally known as Against the Mathematicians v11-x1 (the title is provide the subject of epistemology with some of its most cunning
abbreviated co M); and (3) the six Books of Against the puzzles and most obdurate problems.
Mathematicians 1-v1 (also abbreviated to M). As the cities indicate,
M v11-x1 and M 1-v1 used to be printed as parts of a single treatise,
My book has both exegetical and philosophical pretensions.
but they are two perfectly distinct works, and the modern Exegetically, it ceni:res, as I have said, on Sextus. I have attempted
nomenclature is unfortunate.
to give a comprehensive treatment of my topics insofar as they
4 For Sexcus as a copyist see Jonathan Barnes, 'Diogenc Laerzio e ii
Pirronismo', Elenchos 7, 1986, 385-427, with references to the appear in Sextus' pages. In addition, I have from time to time
pioneering studies by Karel Janacek. (Add now K. Janacek, "O �� referred to other ancient authors and ancient texts. Here I have not,
V1To6foEc.:is Tp61Tos', Eirene 24, 1987, 55-65.)
Agrippa is referred to by Diogenes Lacrcius (1x 88); and he is of course, been exhaustive in my citations or references; rather, I
presumably the cponym of the Agrippa, written by an unknown
sceptic called Apellas (sec Diogenes Laertius, 1x 106). Otherwise he is 6 Aenesidemus flourished in the first half of the first century ec. But
never mentioned in the ancient texts. We may reasonably conjecture little enough is reported about him - and most of that is
that he flourished at the end of the first century nc (sec below, controversial. See e.g. John Glucker, Antiochus and the Late
pp.121-2), and hence a century and a half before Scxtus. Academy (Gottingen, 1978), pp.II6-19.
VIII IX
have picked a small sample of texts in order to indicate that Sextus sometimes set out arguments in a modestly formal manner. In a
and the Pyrrhonists were, from an intellectual point of view, very few places I have used a few of the logical symbols which
neither hermits nor pariahs - their concerns and interests con pepper modern philosophical writings. I do not see why symbols,
cerned and interested other philosophers and scientists. once explained, should deter. But numerous readers are allegedly
The book is not an introduction to ancient Pyrrhonism.7 But I unsettled by them, and whenever I have used them I have also given
have tried to make my historical remarks elementary and I have a paraphrase in ordinary language.
tried to avoid esoteric scholarship. I hope that readers with no The five chapters of the book derive from five lectures which I
antecedent knowledge of Greek philosophy will find the book gave in Naples in April 1988. The written text does not present
intelligible. Historical allusions in the text call upon no pre exactly what I said; but it remains the text of a set of lectures, and it
existing acquaintance. (And the Note on the ancient authors offers retains some of the looseness and informality which lecturers are
a minimal mise en scene.) All quotations from ancient works are customarily allowed.
done into English; and those few Greek words which appear in the I have added a few footnotes. (Many of them are simply lists of
body of the book are all explained on their first appearance. (The references, which look ugly if set out in the text.} It is easy enough
footnotes sometimes cite Greek without translation.) Greek words to clog the bottom of a page with annotations; but more often than
are written in the Greek alphabet. I was once a champion of not such stuff rather displays the author's learning than forwards
transliteration, but I now find it both aesthetically displeasing and the reader's understanding. At any rate, that is my current excuse
pedagogically fatuous - I cannot believe that an intelligent and for idleness.
Greekless reader who has had the Greek term for disagreement Again, the book contains no bibliography and few bibliograph
Englished for him will somehow find it easier to understand the ical references. There are several excellent bibliographies of
sign 'diaphonia' than the sign '51mpwvia'. Pyrrhonism in print,8 and there is no need to publish another. As
The philosophical pretensions of the book are modest. All the for references, I have used them for one purpose only, viz. to direct
issues I discuss arise from the work of Sextus. I have, of course, the reader to discussions of matters which are not dealt with in my
selected those points which I - and, I hope, other philosophers - own text. (Even here I have been selective, and in particular, I have
find exciting; and I have on occasion developed a point somewhat not thought it appropriate to give running references to the vast
further than the Greek texts do. But my ambitions are essentially modern literature in the theory of knowledge.9) Lest this practice
determined by the texts. I have been concerned primarily to de 8 There is an introductory bibliography on Pyrrhonism in Annas and
scribe and present the real philosophical difficulties which the texts Barnes, op. cit. n.7; and a full one by Luciana Fcrraria and
raise. I do not claim to have resolved any of the difficulties. I do not Giuseppina Santese, in Gabriele Giannantoni (ed.), Lo scetticismo
antico (Naples, 1981). For an excellent bibliographical introduction
even claim to have contributed to their resolution - except insofar to Hellenistic philosophy in general see volume z of A.A. Long and
as a plain description may itself make such a contribution. D.N. Sedley, The Hellenistic Philosophers (Cambridge, 1987). The
bibliographical record published each year in the journal Elenchos
I hope that the philosophical parts of the book will be intelligible will keep the brazen-bowelled up to date.
to readers who may be interested in ancient thought but have no 9 Of the modern literature l think I have gained most from P.F.
antecedent knowledge of modern philosophy. I have tried to avoid Strawson, Skepticism and Naturalism - some Varieties (London,
1985); Laurence Bonjour, The Structure of Empirical Knowledge
jargon, and I have tried to avoid covert allusions to modern issues (Cambridge Mass, 1985); Alvin I. Goldman, Epistemology and
or modern authors. I have also tried to write plainly and to write Cognition (Cambridge Mass, 1986); John Pollock, Contemporary
Theories of Knowledge (London, 1986); and from the articles
with lucidity. But the issues are difficult - or so I find -and I am sure collected by G.S. Pappas and Marshall Swain, Essays on Knowledge
that I have left some things vexatiously dark and obscure. I have and Justification (Ithaca NY, 1978), and by Hilary Kornblith,
Naturalizing Epistemology (Cambridge Mass, 1985). There is a
7 For that l may refer to Julia Annas and Jonathan Barnes, The Modes helpful recent survey by Ernest Sosa, 'Beyond Scepticism, to the Best
of Scepticism (Cambridge, 1985). of our Knowledge', Mind 97, 1988, 153-88.
x XI
seem intolerably immodest, I should perhaps say - what will be Galen. But I do think that his works contain a quantity of good
evident to any professional who may chance upon this book - that I philosophy. His reputation is now once more on the rise. If this
do indeed owe a very great deal to earlier philosophers and schol book helps to persuade one or two readers to look seriously at
ars. Like all other students of Pyrrhonism, I have, for example, Sextus and at ancient Pyrrhonism, it will have achieved all it can
learned much from Victor Brochard's Les sceptiques grecs. If I do decently hope for.
not write 'cf. Brochard' at the foot of every fifth page, that does not
mean that I am not indebted to Brochard at least so often. Oxford JO NATHAN BARNES
June r989
The five lectures were given at the invitation of the lstituto Italiano
per gli Studi Filosofici. I am deeply grateful to the President of the
Institute, Avvocato Gerardo Marotta, and to its Director, Profes
sor Giovanni Pugliese Carrattelli, for inviting me to speak on
scepticism in the elegant and learned surroundings of the Palazzo
Serra di Cassano. I am grateful, too, to the Secretary General of the
Institute, Professor Antonio Gargano, for his generous and unob
trusive aid.
I thank my Naples audiences for the helpful comments and
criticisms which they offered, and also for the courtesy and pa
tience with which they suffered my vile Italian.
Earlier versions of some of the lectures were delivered as papers
to various groups and gatherings: at the Universities of GOttingen,
Oxford, Ziirich, Bern, York, Alberta, Princeton, Budapest and
Pees, and at the London School of Economics. On each of these
occasions I suspect that I learned more from my audience than they
learned from me.
The final typescript was scrutinized by two referees for the
Cambridge University Press. Their anonymous remarks enabled
me to make a number of substantial improvements to the text.
Over the years I have accumulated more debts to friends and
colleagues than I can readily recall. Several of the ideas which I here
put forward as my own were certainly suggested, implicitly or
explicitly, by others. I hope that the true begetters will be content
with a general expression of gratitude - and more flattered than
vexed if I have ploughed with their heifers. I cannot name them all
individually, and to select some would be invidious.
Sextus once had the status of an influential thinker, and he was

once read as an important philosopher. I do not think that he was a
great philosopher, of the rank of an Aristotle or a Chrysippus or a
XII Xlll
I
Disagreement
ln a dialogue written by the satirist Lucian, Menippus relates how

he determined to abandon all worldly pursuits - wealth and office
and power - and resolved to study science and the nature of things.
At first he was overwhelmed by doubt and perplexity. In order to
defeat the doubt and disentangle the perplexity, he sought out the
best philosophers and scientists of the age - and he paid vast sums
for his tuition. The fees, he complained, were wasted.
So far from releasing me from my original ignorance, they
actually plunged m·e into greater perplexities, drowning me
every day with principles and ends and atoms and voids and
matters and ideas and the like. And what seemed hardest of
all to me was that, although no one of them said the same
thing as any other but all uttered conflicting and contrary
statements, yet each thought to persuade me and tried to
convert me to his own view.
To which Menippus' companion ironically replies:

How extraordinary that these wise men should fight with one
another over the facts and not hold the same opinions on the
same matters. (lcaromenippus 5)
Faction and disagreement, conflict and dispute, were endemic

among scientists and philosophers in the second century AD. Plus <;a
change.
Galen, Lucian's contemporary and the leading medical scientist
of his age, was impressed and troubled by the disagreement he
found among his fellow-scientists. He refers repeatedly to the
disputes which exercised the doctors. His own technical writings
often start by rehearsing a dispute, which they then attempt to
I
Disagreement Disagreement
resolve. 1 He composed separate essays about disagreements and Again, there was disagreement within the schools. That the Stoics
2 disagreed with one another was a commonplace. As for the Aca
about ways of solving disagreements.
Disagreement - or 51acpwv{a, to use the commonest of several demics, Numenius wrote a petulant essay on their dissensions
Greeks words for the phenomenon - was a normal feature of the (Eusebius, PE XIV iv 16). Even the Epicureans, who were tradition
medical world of Galen's day; indeed, it was an institutionalized ally regarded as uncommonly harmonious and uniform in their
feature. For medical men would usually belong to a sect or school views, indulged in domestic strife. The writings of Philodemus,
(a a1pea1s, whence the English 'heresy'); and the sects were distin preserved on the Herculaneum papyri, record several squabbles.
guished one from another precisely by their doctrinal differences. Some members of the school, for example,
Nor was that all: even within a sect there was likely to be interne refuse to say that sophistical rhetoric is an art of other things,
cine strife, and whatever school you subscribed to - whether you as it in fact is, and want to prove that it is an art of the useless,
were a Pneumatic or an Erasistratean or a Herophilean or a and on each point they disagree (5tacpwvoVvTES) with the
Methodic - you would find enemies at home as well as abroad.' Masters. (rhet [PHerc 1674 } xxix 13-2 1; cf. Iii 19-2 1)
If disagreement was normal among the scientists, it was notori
These renegades, as Philodemus characterizes them, were in their
ous among the philosophers. And it was notorious a century before
own eyes orthodox followers of the Masters of the Epicurean
Galen complained. Philosophical 51mpc.vvia was a commonplace in
school.
Seneca's day, and allowed him a little joke: facilius inter
Some philosophical disputes were no doubt trifling and termino
philosophos quam inter horologia conveniet - You're more likely
logical (or so Galen insists); but many were substantial and signifi
to find two philosophers in agreement than two clocks (apocol n 2).
cant. Deep disagreement was a philosophical fact.
Earlier still, this outrageous dissensio had encouraged a Roman
proconsul who was passing through Athens to summon the disput
But if disagreement was a fact, what - if anything - did the fact
ing philosophers to a conference at which, whether ingenuously or
imply? What attitude should a philosopher take to the disputes
disingenuously, he offered to arbitrate among their views and bring
which certainly separated his school from other schools and which
the scandal of philosophy to an end.'
probably divided his own school within itself?
As with the doctors, so with the philosophers, disagreement was
One attitude or reaction to dispute was an industrious resolve.
institutionally enshrined. Most philosophers thought of them
Disagreement could be seen as a challenge and a spur. If you and I
selves as belonging to a school or sect.5 The main schools, Stoics
disagree we cannot both be right, so let us strive to determine who
and Epicureans, Peripatetics and Academics, defined themselves
(if either of us) actually has truth on his side. In this way 51aq>wvia
by their doctrines. And their doctrines, of course, conflicted.
may stimulate philosophical research. A familiar aspect of
I See e.g. nat fac 1193 K; us part Ell 17 K; PHP V 288 K; alim fac VI 454 K;
syn puls ix 443 K. Aristotle's philosophical procedure involves the collection of 'rep
2 See e.g. lib prop XIX 38 K {On the Disagreement a1nong the utable opinions (1v5o�a)': these opinions will normally conflict
Empirics, three Books; Against the Objections to 'On the with one another; hut analysis and critical revision will eventually
Disagreement among the Empirics' and the Summaries of Theodas,
three essays); 45 K (On judging between those who Disagree in their reveal the truths behind the disagreement and provide for the
Doctrines); ord lib prop XIX 55 K (On Disagreement in Anatomy). establishment of undisputed doctrine. And thus 51aq>c.vvia leads to
These works have not survived.
Sec e.g. meth med X 35, 53, 125 K, on disagreements among the
knowledge.'
Mcthodics; Soranus, gyn 111 2.
4 The man was L. Gcllius Poplicola, the date 93 BC: sec Cicero, leg 1 6 See e.g. Jonathan Barnes, 'Aristotle and the Methods of Ethics',
xx 53. Revue lnternationafe de Philosophie 133/134, 1980, 490-51I. (For a
See D.N. Scdley, 'Philosophical Allegiance in the Greco-Roman later Peripatetic example see Jaap Mansfeld, 'Diaphonia: the
World', in Miriam Griffin and Jonathan Barnes (edd.), Philosophia Argument of Alexander De Fata Chs. 1-2', Phronesis 33, 19 88, 1 8 1-
Tagata (Oxford, 1 989). 207. )
2 3
Galen, too, construed disagreement as a challenge - a challenge infinite . . . And so with many other philosophical enquiries
to uncover some reliable scientific and philosophical method. too - some disputes we cannot decide at all, and others
require much research. But when it comes to what helps or
Drowning in the sea of disagreement among doctors, I turned
harms the sick, matters are different; yet even so discovery
to judge the matter, and recognized that I must first train
still requires much time - and men of exceptional talent.
myself in the methods of proof. This I did for many years.
(in Hipp morb acut xv 434-5 K; cf. PHP v 766 K)
(meth med x 469 K)
But if Galen thought that philosophical disagreements were harder
Logic, for Galen, is the key to method (cf. subfemp 62.2-6 B), and it
to resolve than medical disagreements, he was not wholly pessimis
must be backed by arduous training. There is reason to think that it
tic even on the philosophical front; and he certainly did not hold
will succeed.
that the existence of disagreement was in itself a sign that the
Do not be downcast by the mass of disputing doctors and discovery of truth is beyond us. On the contrary, he supposed, in
philosophers. If they all possessed the wherewithal to learn Aristotelian fashion, that disagreement should be a spur to indus
the truth and yet did not discover it, then it would indeed be try, that Blaq:iwvia should be an incentive to q:itf..o1Tovia.
reasonable for us to renounce any hope of discovery. But Not everyone agreed with Galen. There was - not unexpectedly
some of the prerequisites they do not possess (as they them - disagreement about the proper attitude to disagreement. Thus
selves actually admit), and in the case of others it is unclear
Galen believed that we could discover the composition of the non
whether they possess them or not. As for us, if we are aware
organic parts of the body. But
that we possess all the wherewithal, then we may tackle the
enquiry with confidence. (canst art med I 243-4 K) we see the philosophers who embark on these enquiries - and
some doctors too - differing widely among themselves. That,
What are these prerequisites to discovery? They are seven: natural I think, is why most doctors seem to have given up such
acumen; an early education in mathematics; submission to the best enquiries, supposing that what they are investigating cannot
teachers of the age; indefatigable industry; a longing for the truth be discovered �and some have abandoned them on the
('which very few have possessed'); a grasp of logical method; and grounds that they are not only impossible but also useless.
constant practice in the method. 'If you possess all these things, (const art med I 243 K)
what prevents you from enquiring into the truth with good hopes?'
The doctors who abandoned such enquiries were the Empirics,
(canst art med I 244-5 K).
members of the medical school which denounced all theorizing and
Galen thought that knowledge was attainable. He was an episte-
rejected any investigation into things imperceptible, and which
mological optimist. But his optimism was not inordinate.
maintained that medical science required nothing but perception
If we escape from the disagreements of the Dogmatists, we and the collation of perceptions.' For the Empirics observed that
shall still be overcome on many matters; but on some issues
from the same phenomena different people infer different
we shall produce coherent theories, as the geometers and the
conclusions. Alld here they lay hold of the undecided dis
arithmeticians do. (sub( emp 6 7;22 -68.3 B)
agreement (6:veiriKpt-ros 61aq:iwvia) which they say is a sign of
And he was decidedly less sanguine about philosophy than about unknowability . . . Unknowability, they say, is a cause of
medicine. For undecided disagreement, and conversely disagreement is a
sign of unknowability. It is disagreement about unclear
in philosophical disagreements we have no sensory evidence
we cannot use sensory evidence to determine if the world is 7 On the Empirics see e.g. the introduction to Michael Frede, Galen:
generated and destroyed, if there is void beyond it, if it is Three Treatises on the Nature of Science (Indianapolis, 1985).
4 5
things, not about the phenomena, which is undecided. For Since the pagans themselves stood in diametrical opposition
with the phenomena, each thing appears as it is and bears to one another and kindled useless conflicts and wars against
witness to those whose views are true, refuting those whose themselves, surely absolutely anyone would reasonably al
views are false. (sect ingred I 78-9 K) low that for us suspension of judgement (ETioxtl) about these
matters is the prudent course. (xv xxxii 9)
The same report can be read, a century before Galen, in the medical
writer Celsus. The medical Empirics were sceptics in matters hidden to percep
Those who are called Empirics . . . claim that enquiry into tion. Eusebius urges Erroxfi or suspension of judgement in matters
hidden causes and natural activities is pointless, since nature philosophical. In each case, disagreement has generated a sceptical
is not knowable. That it cannot be known is plain from the conclusion.
disagreements among those who have disputed over it - for
on these points there is agreement neither among philo Eusebius was not a philosophical sceptic; nor were the Empirics
sophers nor among the doctors themselves. philosophers by profession (though there were interesting links,
(rned proem 27-8)
both theoretical and historical, between medical Empiricism and
Thus the Empirics in effect regarded 81mpwvia as a sort of philosophical scepticism). But both the Christian bishop and the
disease, and (rather against their own theoretical position) they pagan quacks were prepared to infer a sceptical conclusion from
proposed an explanation or aetiology for it. If there is undecided the 81a<pwvia1 they observed. Now the inference from 81a<pwvia to
disagreement, then that is a sign of unknowability: if there is scepticism was held to be especially characteristic of the
undecided disagreement over some question (over the eternity of Pyrrhonian philosophers. In the following passage, Galen ad
the universe, or the composition of the blood, or the physiological dresses Julianus, a Methodic doctor who claimed to be Stoic in
location of psychological functions), then we cannot possibly philosophical matters but who professed scepticism about the
know the answer to the question. The disagreement thus points to chemical composition of bodies. Galen asks:
the unknowability. And the unknowability in turn is the cause of Why should we not have precise knowledge of this, so long as
the disagreement. That is the aetiology of the disease. And the we can provide a proof of the thesis and so long as the best
therapy? Ignore the thing: if you can't know the answer to a philosophers - whom you yourself admire - agree with
question, then don't bother to ask it. The disagreement will not Hippocrates and with one another? Unless, of course, you
disappear but it will cease to irritate you. think that the disagreement (8laqiwvla) is sufficient evidence
A similar reaction to philosophical S1a<pwv{a is found in a writer of our ignorance of the thesis, thus suddenly becoming a
of a different profession and character. The disagreement of the sceptic (0TTOPflTJK6S) instead of a Stoic. For then you accept a
pagan philosophers is a theme which runs through Eusebius' theory which says that nothing which is a subject of disagree
ment among all the philosophers can possibly be available for
Preparation for the Gospel. Early in the work, Eusebius draws
human know�edge. (adv lul XVIIIA 268 K)
attention to 'their opposition to one another � for they agree in
nothing and have filled everything with conflict and disagreement' If Julianus bases his confession of ignorance on the fact of
(PE I viii 14); and at the end ofBookxv he closes with tart references 8ta<pwvia, then he shows himself a sceptic. (The word cmopflTIKOI
to 'their massive disagreement', to 'their rivalry with one another is a standard denomination of the Pyrrhonian sceptic: Sextus, PH
and their conflicts and their dissensions' (xv lxii 1 3 , 1 5 ) . The I 7.) For the idea that disagreement is a sign of ignorance and a
disagreement or Oia<pwvia among the Greeks and tl.eir philos reason for suspension of judgement, is, according to Galen, the
ophies contrasts with the agreement or crvµ<pwvia of the Hebrews characteristic mark of Pyrrhonian scepticism.
and their scriptures (e.g. XIV ii l "' iii l ) . The moral is plain. And Galen is right. .6.tcxq:>wvia1 are referred to on countless
6 7
occasions8 in the writings of Sextus Empiricus. For Sextus, dis Suspension of judgement is a standstill of the intellect, be
cause of which we neither reject nor accept anything.
agreement is a fount and origin of scepticism, and this use and
(PH t ro)
understanding of disagreement is utterly characteristic of
Pyrrhonism. It is the Pyrrhonian use of 8taq>wvla which I want to Thus:
discuss. If I have none the less begun this chapter by citing Galen
We take 'I suspend judgement' in the sense of 'I cannot say
and the Empirics and Eusebius, that is because I also want to insist
which of the offered views I should believe or disbelieve', thus
that the argument from disagreement is not peculiar to showing that the matters seem equal to us with regard to
Pyrrhonism. Some moderns regard the ancient sceptics as philo warranty and lack of warranty. (PH I r95)
sophical lightweights or dilettantes - as men whose arguments are
profoundly Sl:'perficial. Their outlandish views may amuse us, but In other words, I suspend judgement on, say, the immortality of the
they cannot enlighten us and they should not engage us. On one soul if, having considered the matter, I neither reject nor accept the
point at least - namely, the argument from disagreement - the soul's i mmortality, if I neither believe nor disbelieve that the soul is
Pyrrhonians developed a line of thought which commended itself immortal. (Why add the qualification 'having considered the mat
to men of bottom: to sober doctors and to reverend bishops. Let us ter'? Well, Sextus knew nothing about the fauna of South America.
dare to treat Pyrrhonism seriously (which is no reason why it He neither believed nor disbelieved that there are pangolins in
should not also be diverting). Patagonia. Did he then suspend judgement on the matter? No:
The main question is this: What is the connexion between 1noxfi is something which comes about 'after the enquiry' (PH I 7),
disagreement on the one hand and sceptical suspension of judge and Sextus has not- could not have - made any investigation of the
ment on the other? What is the connexion between 81acpc.vvia and fauna of the Americas.)
E-iroxfi? Before tackling the question, I had better say a little more In schematic terms, _I suspend judgement with regard to a propo
about the two concepts which it links. sition P if, having considered the matter, I neither believe that P nor
'ETioxi} was a term of art in ancient Pyrrhonism. Sextus explains believe that not-P. More generally, I suspend judgement with
it formally as follows: regard to the question ?Q if, having considered the matter, I neither
8 'Countless occasions'? Well, I have counted about 120 uses of
accept nor reject any answer to ?Q. My scepticism can be, of
6Jmpwvla: and its cognates; in addition, Sextus often uses O'T6cr1s or course, more or less extensive. Every rational being is sceptical on
816:crTa:cr1s and �heir cognates, or 6:µqi1a�i)TT)OtS and its cognates, or some issues ('When exactly was Sextus born?'). Some rational
<Xvwµa:;>da: and its cognates. There are occasional uses of µCcxTJ and
n�Aeµos, and the like. Often the positive counterparts of 81a:qiwv!a beings are sceptical with regard to general areas of enquiry (as the
(viz. avµqiwvla, 6µoqiwvia, OµoAoyla etc.) indicate the presence of an Empirics were sceptical about the underlying causes of diseases). In
argument from disagreement. (See Karel Jan3.Cek, Sextus Empiricus' the most extreme case, I might suspend judgement on every ques
Sceptical Methods (Prague, 1972), pp.73-80. Throughout this
invaluable study, JaniiCek emphasizes Sextus' predilection for variatio tion I consider: I shall then have no considered beliefs or disbeliefs;
in his choice of words. I suppose that in Sextus, and also in Galen I shall be - as we may put it - a radical sceptic.
(who �s another lover of variety), the predilection is not merely
.
stylistic: Sextus, like Galen, insists that it is fatuous to fuss over 'Enoxfi is the lieart of Pyrrhonian scepticism. It is clear that
words oU qicuvoµaxoVµev,) In all, there are more than 200
- Pyrrhonian scepticism is somewhat different from the scepticism
references, direct or oblique, to disagreement in Sextus' works; and which Galen ascribes to the Empirical doctors. For the Empirics
there must be several hundred sections (i.e. numbered paragraphs in
t�e modern editions) which are specifically concerned with using hold that nothing can be known, they aver unknowability or
disagreement to sceptical ends. I have generally tried to give aKoTOA11'f' IO. There is a difference between unknowability and
comprehensive documentation for my remarks about Sextus. Here I
have been content with a sample set of references. Those who want suspension of judgement. Since I am in a state of unknowability
more should turn to one of the most useful books yet written on with regard to ?Q if I hold that the answer to ?Q is not and cannot
ancient Pyrrhonism, viz. Karel JanftCek, Sexti Empirici Opera: JV be known, it is evident that I may suspend judgement over ?Q
Indices (Leipzig, 19621).
9
8
without being in an attitude of aKaTaA'lljJia towards ?Q. For I may But we should not take this too seriously. Whatever Sextus may
neither believe that ?Q is answerable nor believe that ?Q is not say, the Pyrrhonists did not - in any normal sense - prosecute
answerable. (I may - and, of course, a radical sceptic must - philosophical and scientific researches; nor did they construe dis
suspend judgement over the question of aKaTaA111jJia.) Thus broxfi agreement as a positive incentive and challenge in the Galenian
with regard to ?Q does not imply aKaTaAf1\jlla with regard to ?Q. spirit. None the less, it remains true that Pyrrhonian scepticism -
Nor, in the other direction, does aKaTaA'lljJia imply broxfi ; for I Pyrrhonian ETIOXTJ - is, formally speaking, open-minded and in
may believe that P is the answer to ?Q while holding that no one can principle tolerant of future progress. 'AKaTaAf1\jlta is different: it
know the answer to ?Q. (Perhaps we cannot know whether Galen locks the door to the temple of wisdom.
had read Sextus; but I myself believe that he had not.) Thus the As for disagreement or Slaq>wvia, Sextus offers no official defini
difference between suspension and unknowability is clear in princi tion. It is, I suppose, worth saying first that Ola<pwvia is not mere
ple. Moreover, Sextus himself, whether rightly or wrongly,' distin difference of opinion. You and I may have different opinions
guishes Pyrrhonian from Academic scepticism (and also from the without disagreeing on anything. Perhaps I believe that the toad is a
philosophy of the Cyrenaics) precisely by the fact that Pyrrhonians mammal and you believe that it has magical properties. We are
stick with suspension while Academics (and also Cyrenaics) main both wrong, but our opinions do not conflict. For disagreement to
tain unknowability (PH I r-3 and 215; cf. 226). Often, it is true, exist, there must be conflict: my opinion and your opinion must be
Sextus says - or seems to say - that Pyrrhonism embraces incompatible or in conflict with one another." Perhaps I hold that
aKaTaA'lljJia.'0 These passages are admittedly puzzling. But most the toad is a mammal and you hold that it is an insect. (Note that if
of them can, I think, be explained away (Sextus does not actually two opinions conflict, it does not follow that one of them must be
mean what he appears to say); and the rest may be put down to correct. One, at least, must be wrong - and both may be wrong.)
carelessness. However that may be, I shall speak as though Sextus Again, disagreement need not involve only two parties: if I take
consistently took suspension to be the sceptic's state of mind. the toad to be a mammal and you take it to be an insect and the
Thus the Pyrrhonists did not, officially, infer from Oia<pwvia to Zoologer Royal believes it to be a bird, then there is a 6taq>wvla a
the particular sceptical conclusion which we have seen in the trois. Each of the three opinions is incompatible with each of the
Empirics. Indeed, Sextus' official view is closer, in a way, to other two. Many of the disagreements Sextus chronicles are in fact
Galen's; for he insists that the sceptics continue to investigate (PH of this multilateral sort: there were, after all, many different
I 3). The Greek word crKETTT1K6s, which I translate as 'sceptical',
philosophers and several different philosophical schools, and each
means literally 'enquiring'; and Sextus explains that liked to disagree with all the others. But multilateral 61a<pwvia1 can
always be analysed in terms of bilateral 61a<pwvia1. We may say
the enquiring (crKETIT!Kfi) school of thought is also called that x and y and z are in disagreement with one another if and only
investigative (�f]Tf]TiKfi) from the activity of investigating and
enquiring. {PH 1 ?) 12 What exactly is conflict here? It is more than 'non-compossibility';
that is to say, if two propositions cannot both hold together, it docs
And he often remarks that the sceptic's broxfi holds good only 'up not follow that they conflict with one another. For suppose that P is
in itself impossible {let it be 'There is a highest prime number' or
to now (µexp• or axpi vuv) "11 thereby hinting that future resolution 'Scxtus is both older and younger than Galen'); then P cannot hold
of the doubt and future knowledge are not formally excluded. together with any other proposition. For since P is impossible, the
conjunction 'P and Q' is impossible for any proposition Q. But we
should not want to say that P conflicts with every other proposition
9 On this see Gisela Stricker, 'Ueber den Untcrschied zwischen den
Pyrrhoncern und den Akadcmikern', Phronesis 26, 1981, 153-71, ('There is a highest prime number' docs not conflict with 'Vesuvius
is on the bay of Naples'), It is difficult to say precisely what conflict
10 Sec e.g. M I 320: 'what is disagreed upon without decision is does consist in: it is, as it were, non-compossibility plus something
unknowable (0J<aT6:i\rprTov)'; and cf. janaCck, Methods, pp.27-s.
else, but the something else is elusive. Sec e.g. M.R. Stopper, 'Schizzi
11 Sec e.g. PH 1 25, 200, 201; 111 70; M vn 380; vm 118, 257, 401, 427/8; Pirroniani', Phronesis 28, 1983, 265�7, at pp.285--6.
Xl 229.
II
IO
if x is in disagreement with y and y is in disagreement with z and x is be; for it is no part of the concept of 81a<poovla that the parties to it
in disagreement with z. (The account is easily generalized to cover be actively in dispute with one another. Indeed, there may be
any number of disputing parties.) disagreement among the dead, who are certainly not still in debate
The parties to the disagreements which Sextus rehearses - the and who in many caseS did not even know of one another's
'parts of the 81acpoovia', as he often puts it 13 - are usually disputing existence. You and I are in disagreement about the taxonomy of the
philosophers; and the disagreements are usually disputes among toad even if neither of us has heard of the other's foolish opinion.
the rival schools. (Sextus also notices some 81acpoovia1 within the Disagreement is essentially a logical relation between our opinions
schools: e.g. M vu 228; VIII 400.) But disagreement is not, of course, and not a personal relation between us.
limited by law to the professionals, and Sextus sometimes explic But disagreement is still a relation between opinions, between
itly refers to disputes in ordinary life (�ios) or among laymen." propositions held or opined. It is not simply a relation between
Laymen, like professionals, are people; and in Sextus' writings propositions. The proposition that the toad is a mammal is not
the parties to a disagreement are usually persons or groups of itself in disagreement with the proposition that the toad is an
people. Sometimes, however, Sextus will speak of a 6taqiwvia insect; it is the two opinions on the matter which show 81a<poov{a.
among the senses or among the phenomena or even among the Thus we might think to define disagreement in roughly the
'things'." If my sense of sight reports that a surface is smooth and following manner:
my sense of touch reports that it is rough, then there is a disagree
(D) x and y disagree (5icxq>wvo0cri) over some issue ?Q,
ment between these two senses. If the Russian flag appears red to
whenever x offers P1 in answer to ?Q and y offers P2 in
humans and grey to bulls, then there is a disagreement between
answer to ?Q, and P1 and P2 are incompatible with one
these two appearances, between these two cpavTacriat. It is no another.
doubt natural and intelligible to extend the concept of 6taqiwvia in
this way; but the extension introduces new issues and new prob So there is disagreement (61aqiwvia) over a question ?Q if and only
lems. In what follows I shall ignore this extended form of disagree if at least two incompatible propositions have been offered in
ment, which is not an essential part of Agrippan scepticism, and answer to it. For example,
restrict my attention to 81acpoovia1 among people.
Disagreements among the ancient philosophers often took the there have been disagreements among the philosophers over
the material elements. Pherecydes of Syrus said that earth was
form of debates or disputes. Sextus, like Eusebius, uses political
the first principle of everything; Thales of Miletus said water;
and military metaphors to describe the business. He speaks of
Anaximander, his pupil, the infinite; Anaximenes and Dioge
factions and parties (crT6:cr1s and 816:crTacr1s), of wars and battles nes of Apollonia, air . . . (PH III 30)
(n611Eµos and µaxri), 16 and he twice speaks in heightened terms of a
TI6Aeµos &crTielcrTos, a war without truce. 17 These metaphors sug There is disagreement over the question: What are the material
gest actual debate, indeed heated dispute. But although disagree elements of the universe? For in answer to this question, Pherecydes
ments may be manifested in debate and discussion, they need not gave one opinion (P1 }, Thales another (P,), Anaximander a third
13 See PH I 50, 90, 98, 113; II 60, 67; Ill 182; M VII 318, 351; VIII 298. (P,) and so on; and P,, P,, P, . . . are all mutually incompatible. Some
14 See PH I 165, 185; Ill 65, 218, 233; M VIII 355; IX 60, 191; XI 43· of the parties to this multilateral dispute actually debated with one
15 Sec e.g. PH 11 52; M vu 345/6; v111 182 (o:ioB11cns); PH 1 59, 112; M another. Most of them did not. But they were all in disagreement
VII 177, 430 (tpo:VTacrkx); PH I 12, 163; III 235 (-rrp&yµo:).
16 For references see Jan:lCek's Indices: more than 25 occurrences of with one another.
factional words and about a dozen military terms. This preliminary characterization of disagreement requires two
17 See PH III 175; M XJ 36 (where the MSS offer &-rr1crTOS and Fabricius
further comments. First, suppose that you and I are discussing the
emends to &<nTElO"TOS though Mutschmann perversely determines to
-
read 0:TTJO"TOS in both places). authorship of the Magna Moralia, a treatise on ethics which is
I2
transmitted to us as part of the Aristotelian corpus. You hold that the sceptics are themselves 'part' of the 81mpwvia insofar as they
Aristotle wrote the Magna Moralia. I deny that he did, while suspend judgement on the matter, 1 8
offering no positive theory of my own. It is plain that there is a Neither (D) nor its successor, (Dr), will accommodate this sort
81mpwvia between us; and yet we might be reluctant to say that the of 8taq>wvia. We need a further modification to the definition. Let
two of us have given conflicting answers to the question 'Who us say that someone 'takes an attitude' to a question ?Q if, having
wrote the Magna Moralia?', for we might be reluctant to say that I considered the matter, he either accepts some proposition as the
had answered the question at all. answer to ?Q or rejects some proposition as an answer to ?Q or
The preliminary characterization of disagreement must be suspends judgement over ?Q. Such attitudes may be said to conflict
modified to account for cases of this sort. Of several possible with one another in a variety of ways. Accepting P 1 will conflict
modifications, I prefer the following: add a second clause to the with accepting P2 when P 1 and P2 are incompatible with one
definition of disagreement, and write it thus: another. Rejecting P will conflict with accepting P. And suspending
judgement over ?Q will conflict with any other attitude to ?Q. Then
(DI) x and y disagree over some issue ?Q whenever either x
we may define disagreement simply enough, as follows:
offers P1 in answer to ?Q and y offers Pi in answer to ?Q
and P1 and Pi are incompatible, or else x offers P in answer (D2) x and y disagree over some issue ?Q whenever x and y
to ?Q and y rejects P as an answer to ?Q (or vice versa) . hold conflicting attitudes to ?Q.
This is a little cumbersome, but its import is plain: it amounts to no The successive characterizations of 81aq>wvia, (D), and (Dr) and
more than a pedantically precise way of saying that a 51ac.pwvia is a (D2), are related in the following way: anything which counts as a
conflict of opinion on some topic. disagreement according to (D) also counts as a disagreement
The second comment required by the preliminary characteriza according to (Dr), but not vice versa; anything which counts as a
tion of disagreement is this. Sextus sometimes includes the sceptics disagreement according to (Dr) also counts as a disagreement
themselves as parties to a disagreement- and the sceptics do not, of according to (D2), but not vice versa. If we call the notions defined
course, participate in a 51ac.pwvia by virtue of holding an opinion on by (D), (Dr) and (D2) respectively positive disagreement, disagree
the issue, whatever it may be. I have in mind here not those several ment in opinion and disagreement in attitude, then we may say that
passages in which Sextus himself fictionally assumes the role of a all positive disagreements are also disagreements in opinion, and
non-sceptical thinker, producing opinions for the sake of argument all disagreements in opinion are also disagreements in attitude; but
and not in propria persona. Rather, I am thinking of those texts not the other way about.
where ETioxrl is itself an option in the disagreement. For example, Now although (D2) gives the largest - and, I suppose, the
opening his account of the 'criterion of truth', Sextus says: authoritative - explanation of what Sextus understands by a
5tacpwvia, as a matter of fact most of the disagreements which he
Let us begin with the disagreement: of those who have dis reports are disagreements in opinion. My own discussion of
cussed the criterion, some - e.g. the Stoics and others - have 8taq>wvia will therefore usually be couched in terms of such dis
asserted that there is a criterion, others - including Xeniades agreements. But we should remain aware of the possibility of
of Corinth and Xenophanes of Colophon, who says
disagreements in attitude which are not disagreements in opinion -
but belief is found over all -
and at one point in the discussion this possibility will have an
have claimed that there is not, and we suspend judgement as
to whether there is or is not. (PH 11 17-18) interesting consequence.
r!s See PH I r85; II r8, 31, 180, 259; III 23, 65, 119; M VIII 327, 334, 380;
Is there a criterion of truth? Some say yes. Some say no. Some IX 60, 195; and cf. M T 28, where one Dogmatic party to a
suspend judgement. There is disagreement, and it is trilateral. For disagreement suspends judgement.
we know next to nothing about Agrippa, the identification is not
Well then, how does disagreement connect with suspension of particularly informative.
judgement? How does 51acpwvia lead to E-rroxt']? Some passages in The first of the Five Modes is called the mode 'from disagree
Sextus have suggested that disagreement is a necessary condition ment (cm6 51acpwvla1)', and Sextus explains it as follows:
for suspension, in the sense that we cannot or shall not suspend
judgement on an issue ?Q unless there is-or, perhaps, unless we are The mode from disagreement is the one in virtue of which we
discover that on any matter proposed there has been
aware that there is - a Sta<pwvia over ?Q. And it is true that Sextus
undecided faction (6:veTIIKp1Tos crT6:cr1s) both among laymen
himself frequently introduces and describes a disagreement, some % (rro:pO: Tcf> �1'1') and among the philosophers. And because of
2
times at considerable length, before drawing any sceptical con the dispute we cannot choose or reject anything, and so end in
l
clusion on a topic. 1 9 Now Sextus no doubt thinks that drawing suspension of judgement. (PH I 165)
attention to disagreement is a useful preliminary to other sceptical
argumentation - it softens the reader up. But no text explicitly * The mode of disagreement thus moves from crTO:cr1s, i.e. from
states, and none clearly implies, that disagreement is, in the strict I 51acpwvla, to ETIOXtl· How and why does it do so?
l
A key term in the text is the adjective c'xvrniKpnos. The verb
sense, a necessary condition for suspension of judgement. Nor, of
course, should Sextus have held that 51acpwvla was an obligatory I
;:: E1TtKpive1v, in the sense of 'decide' or 'determine' (a dispute), is
precondition for suspending judgement. For it would be absurd to I ubiquitous in Sextus; and its standard use is precisely in connexion
imagine that I cannot suspend judgement on a topic unless at least ± with disagreements - ETI1Kpive1v Ti)v S1aq>wviav is to settle or decide
two other people have held conflicting opinions on it; it would be the disagreement, to bring it to an end. Thus a disagreement is
absurd to suggest that I cannot be doubtful myself unless others ETitKptT6S if there is a decision for it; and it is 6:veTiiKptTOS (the
have been sure - and sure in different directions. (I return briefly to adjective is formed with an 'alpha privative')i if there is no decision
this point in the final chapter.) for it. But what is the precise force of the adjective? What exactly
In any event, it is the converse relation which is the important does it mean to say that 'there is no decision' for a dispute? There
one: Is S 1a<pwv ia a sufficient condition for ETioxf}? If there is a are two questions here. One, which I shall postponei concerns the
disagreement over ?Q, must we suspend judgement on the matter, precise force of the verbal root ETI1Kp1v-; the otheri which I address
neither accepting nor rejecting any proffered answer to the at once, concerns the precise force of the adjectival termination
question? -TOS.
At PH I 164-77 Sextus explains the five 'modes of suspension Verbal adjectives ending in -TOS are very common in Greek.
(TpOTIOl Tfil ETioxfis)' which 'the more recent sceptics hand down' They are generally held to be ambiguous: V-TO\ may carry the
(PH I 164). A 'mode' or TPOTIOI is, roughly speaking, an argument notion of possibility and have a 'modal' sense ('can be V-ed') and it
form; and a mode of suspension is an argument-form which leads may also have an ordinary non-modal sense ('is V-ed'). The ambi
to suspension of judgement. The 'more recent' sceptics are identi guity was familiar to the Greeks themselves. Thus Galen carefully
fied for us by Diogenes Laertius, whose Life of Pyrrho contains an explains that the adjective ala611T61, which derives from the verb
account of the Five Modes which closely parallels the account in aicr66:vecr6a1 ('to perceive'), sometimes means 'perceivable' and
Sextus (Diogenes, IX 88-9). 20 According to Diogenes, the Five sometimes 'perceived' (diff puts VIII 7ro K). Hence we might
Modes were presented by Agrippa (ol mpi "AypmTiav). But since conclude that 6:veTiiKp!TOS may mean either 'ttndecidable' or
'undecided' - 'there is no decision' for a dispute either if the dispute
r9 See e.g. Pll 11 17; 111 30; M VII 46, 261; vm 3, II. (These are passages
where Sextus expressly remarks upon the fact that he is starting the cannot be decided or if it has not been decided.
sceptical discussion by setting out a B1aqiwvla.) Yet this conclusion is too swift, and it ignores a further subtlety.
20 See Jonathan Barnes, 'Diogene Lacrzio e ii Pirronismo', Elenchos 7,
1986, 385-427.
Suppose there is a jewel at the bottom of the sea: can it be
16 17
perceived? In a sense, plainly not. But this is precisely Galen's not thereby determine the sense of the word 6:velTiKpt1os. For,
example of something which is alcr6T)T6S in the sense of 'perceiv evidently, one could hold that a disagreement was undecidable or
able'. So the jewel is perceivable even though it cannot be per not the sort of thing to be decided, and yet, in calling it 6vmiKp1Tos,
ceived. It can (in one sense) be perceived; it cannot (in another mean only that it was undecided. Your attitude to 5taq>wvia does
sense) be perceived. It is the sort of thing which, given the right not in itself fix the sense you assign to the word avmiKpnos. And
conditions, is perceived; but now, the conditions being wrong, it there is in fact some reason to suppose that in Sextus' writings the
cannot be perceived. In short, we must distinguish between some adjective avmiKpnos should not be interpreted in sense (iii).
thing's being in principle perceivable and something's being in the For on more than one occasion Sextus says that a dispute is
circumstances perceivable. O:ve1TiKpt1os up to now, or that it is still 6:vE1TiKptTos, or that it
Aristotle, like Galen, was alive to the ambiguities of adjectives awaits decision. 22 The implication is that, for all the sceptic knows,
terminating in -10s. Thus he comments on the adjective &cp6ap1os, the disagreement might, in the future, be resolved. Now it would be
which derives from the verb q>6EipE1V ('to destroy'): at best mildly odd to say that a disagreement is 'still' or 'up to now'
'It is now &cp6apTos' is ambiguous: it means either that it has not the sort of thing that can be decided. If something is ever not the
not now been destroyed, or that it cannot be destroyed now, sort of thing to be decided, it is surely always not the sort of thing to
or that it is now such as never to be destroyed. be decided. Modal status of this sort does not change. (Similarly, it
(Top 145b24-7) would be odd to say 'It is, up to now, impossible to be at the same
time both taller and shorter than someone else'.) Hence in these
Here Aristotle notes a third sense in addition to the two distin
passages we should be inclined to take avmiKpnos in sense (i) or
guished by Galen. For 'It is &q>6apTos' ('There is no destruction for
sense (ii); an 6:veTiiKptTOS 81a<pwvia is a disagreement which re
it') may mean either (i) 'It has not been destroyed', or (ii) 'It cannot
mains undecided or which no one has yet managed or been able to
(in the circumstances) be destroyed', or (iii) 'It is not the sort of
decide. And if in these passages, why not in all passages?
thing to admit destruction'.
Sense (i) or sense (ii)? Some passages seem to me to favour (i) and
Thus if we said of a problem that it was &f-vTos ('There is no
others to favour (ii), and I cannot decide if it is right to fix upon one
solution for it'), we might mean either (i) that it has not been
of these senses for all the occurrences of the word in Sextus. None
solved, or (ii} that it cannot at the moment be solved, or (iii) that it is
the less, I have chosen to translate the adjective by 'undecided', as
genuinely insoluble. And a 01acpwvla may be 6:verriKp11os inasmuch
though it always bore sense (i). Even if this is mistaken, it is not an
as (i) it has not been decided, or (ii) it cannot at the moment be
important mistake; for as a matter of fact the difference between (i)
decided, or (iii) it is not in principle decidable.
and (ii) does not affect any of the arguments in which the adjective
Most Sextan scholars seem to take the word 6:vETiiKp1TOS in the
crucially occurs. Thus if we think of an 6:veTiiKptTOS Ota<pwvfa as an
strongest sense, sense (iii). Now there are passages in which Sextus
undecided dispute, we shall avoid the serious error of imagining
explicitly says that a disagreement cannot be decided, or that it is
that such disagreements in principle cannot be resolved, and at the
not possible to decide a disagreement. 2 1 These passages could
same time we shall not falsify the logic of any Sextan arguments.
indeed be read in sense (iii). Moreover, any readerof Sextus is likely
to derive the impression that, if the Pyrrhonists are right, disagree
ments are simply and in principle undecidable. But the impression 22 For µExpl (6xpi) vOv {Sto:Crpo) see: PH 111 70; M v111 I77, 257, 400, 427;
Xl 22.9; for lTL: M XJ 230, 231; M I 28; IT 95; for µEVEIV: M VIII 177,
may be misleading; and the texts, important though they are for
187. Note also the phrases with !:qi' Ooov at PH Ill 3; M VII 380; VIII
determining the Pyrrhonian attitude to certain disagreements, do 118· and further the texts referred to in n.29 below. For what it is
wo;th, I note that Caelius Aurelianus uses iniudicatus as
21 See PH 1 26, 2.9, 59, 178; n 85, 113, 116, 181; 111 54. (presumably) a translation of CxvETTIKpnos: morb acut n 8.
r8
principle. And clearly - or so we might suppose - the appropriate
Sextus does not always qualify the noun 81aqiwvia with the adjec
modification will refer not to possibility but to rationality, not to
tive 6:verrlxp1Tos; but he does so very frequently,23 and he also uses a
what we can or cannot say but to what we should or should not say.
number of other qualifying adjectives with the same general
Then let us replace 'It is not possible' by 'One should not' (where
force." It is plain, I think, that the mode from disagreement
the 'should' bears not a moral but an epistemological sense: 'It is
depends - in one of its forms, at least - on undecided disagreement.
not rational, or warranted, or justifiable . . . ').
In other words, scepticism is supposed to follow not from the fact
Sextus himself might not have been happy with this modifica
of disagreement as such but rather from the fact of undecided
tion; for in general he prefers- and has reason to prefer - 'can' and
disagreement. Thus:
'cannot' to 'should' and 'should not'. (He will rather say that
The question proposed . . . whatever it may be, is a subject of 81cxcpwvicx causes the rr6:6os of €noxr1 than that 81cxcpwvicx makes it
disagreement. Now will they say that the disagreement is rational for us to suspend judgement.) This preference raises
decided or undecided? If undecided, we have it that we must difficult questions, both exegetical and philosophical. I ignore
suspend judgement. For it is not possible to make assertions
them here; for it is plain that, whatever Sextus might have said, we
about matters which are disagreed upon without decision
ourselves shall get no further with the principle enunciated at PH I
(lxvemKpiTWS). (PH l I70)
170 unless we reformulate it in terms of a rational 'should not'.
And so, explicitly, in some twenty-five further passages. 25 Thus 'as Even so, the reformulation is not enough to save the principle.
long as (E<p' Ocrov) the dispute is undecided, we must remain in For it seems clear that I may sometimes rationally make assertions
suspension of judgement' (M vm n8). where there is an undecided dispute. A student of ancient philo
Is this a good argument? It rests on the principle that, in Sextus' sophy might rationally assert that the Magna Moralia was not
words, 'It is not possible to make assertions about matters upon written by Aristotle, even though there is (let us allow) an
which there is undecided disagreement'. But this principle cannot undecided disagreement about the matter. For, first, the student
be literally true. For it is perfectly possible to make assertions on might be perfectly unaware of the disagreement and have no reason
such matters. People do make positive assertions in the face of at all to suspect its existence. (He might possess good evidence for
undecided disagreements (indeed the parties to the disagreements inauthenticity and no indication that the matter was controver
must do so); and if they do, they can. So we must modify Sextus' sial.) Or again, he might be aware of the disagreement and yet
suppose - falsely, but for good reasons - that it had been decided.
23 See PH I 26, 59, 88, 98, 112, 114, 165, 170, 175, 178, .t85; II 19, 29, (He might, for example, have been falsely assured by a reputable
32, 33, 49, 57, 58, 59, 67, 85, 112, 113, 114, 116, 121, 145, 168, 181,
222, 259; I l l 3, 5, 6, 54, 56, 70, 108, 182, 254; M V I I 341, 380; VIII 118, scholar that the old 81aqiwvia had at last been &ettled.)
177, 257, 265, 266, 288, 297; X I 230; M I 27; II 95, 102. {This list Thus we need a second modification to Sextus' principle: the
includes any use of 6:vETTIKp1TOS, adjectival or adverbial, in connexion
with any word for disagreement.) principle should not connect suspension of judgement with
24 E.g. O:vfivvTOS (PH II 8, 31, 212; Ill 56; M VIII 262; M I 91, 156}; 81cxcpwvicx tout court - it should connect suspension with known
O::rrE1pos {PH 11 48, cf. 51' cd&vos: M VIII 186; M 1 170); O:crrrEl<YTOS 81aqiwvia. Thus: ·
(PH HJ 175; M XI 36, see above, n.17); 6:516:KpLTOS (M VIII 214); note
also 6µV6T)TOS (PH II 21), rrolK1i\os (M XI 217; M I 9), icrocreevfis (PH l
26; 1u 65, 139; M 1 6), oVx Ti TvxoOcra (M v J7), µeTEwpos (M 1 28). If sOmeone is aware that there is an undecided dispute about
25 See PH I 26, 29, 59, 88, 112; II 19, 32, 57, 145, 259; Ill 5, 56, 108, 186, ?Q, then he ought not to accept or reject any proposed
254; M VII 380; VIII 177, 257, 265, 427/8; M ' 170/1, 320; II 102; and answer to ?Q.
cf. PH 111 65, 139; M 1 6 (51mpwvia lcrocr6evi)s leads to �rroxfi). There
is a twist to the story: often Sextus says that an undecided I suppose that this thesis represents what the Pyrrhonists ought to
disagreement induces suspension of judgement; but sometimes he
says that it leads to unknowability, 6KaTai\111f1la. On this, sec above, say, even if they do not actually express themselves in this way. I
pp.9�10. shall refer to it as the Principle of Disagreement.
20 2I
Now the question raised by the mode of disagreement is this: Is induce suspension of judgement. If I recognize undecided dispute
the Principle of Disagreement true? And the answer is surely: Yes, over ?Q, then I must - I rationally must - suspend judgement over
the Principle is true. For suppose that it were not true. Then the the matter.
following could be the case. I recognize that there is a dispute about
the authenticity of the Magna Moralia, some holding that the work Thus Agrippa's mode ano 01aqiwvias embodies a sound principle.
was written by Aristotle himself and <?thers holding that it is a later Agrippa, here at least, is right. Now while conceding victory to
counterfeit. I believe, further, that the dispute is still undecided: the Agrippa, we might at the same time wonder how significant his
parties have not come to any agreement, and no decisive argument victory really is - but before I air that notion, let me, by way of a
or consideration for or against authenticity has yet been advanced. diversion, mention two corollaries of the Principle of Disagree
Nevertheless (if the Principle is false) it is rational for me to hold ment. First, recall the distinction between (Dr) and (D2), between
that the work is not authentic. Now it seems clear to me that this is disagreement in opinion and disagreement in attitude. The Prin
incoherent; for how could it possibly be rational for me to plump ciple of Disagreement allows us to prove a curious thesis about
for authenticity, thus opting for one side to the dispute, and yet still disagreement in attitude: any 01aqiwvla in attitude to which a
to maintain that the dispute is undecided? If it is rational or sceptic is a party is always decided. For suppose that there is an
warranted for me to decide against authenticity, then I must undecided disagreement over ?Q, and that Sextus himself is a party
suppose that whatever warrants my decision also and thereby to it. (In answer to the question ?Q, some say P and Sextus, of
decides the dispute, which I can therefore no longer hold to be course, suspends judgement.) Since the Oiaqiwvia is undecided, it
undecided. If, on the contrary, I insist that the disagreement follows by the Principle that we should suspend judgement over
remains undecided, then I cannot consistently suppose that my ?Q. Yet if we suspend judgement over ?Q, we are siding with one of
inclination to reject authenticity, whatever it may be founded the parties to the 01aqiwvia and therefore deciding it. In other
upon, has any satisfactory justification; and hence it is not rational words, the Principle requires us to resolve the dispute in Sextus'
for me to reject authenticity. favour. And so, in this particular sort of case, the Principle leads to
Of course, I may adopt it as a 'working hypothesis' that the a decision.
Magna Moralia is a counterfeit. I may act as if the work is spurious Sextus sometimes speaks as if all Oiaqiwvim are undecided -
-say, by excluding it from my translation of the collected works of indeed, as I have said, he is often taken to suppose that all
Aristotle. But in so acting I am not manifesting any belief that the 01aqiwvim are in principle undecidable. But this cannot be right.
work is spurious. I am not putting money on the horse. (Moreover, Some disagreements of attitude are decidable, and are implicitly
I may perhaps also hold that it is likely or probable that the work recognized as decidable - indeed, as decided - by Sextus himself.
will turn out to be spurious. Then I shall indeed hold a belief on the The thesis we may find in Sextus (if we may properly speak of
matter - but not a belief which is, in any straightforward way, a 'theses' in connexion with a Pyrrhonist) is not the thesis that all
party to the disagreement. For the disagreement is not over prob disagreements are undecided but rather, and at best, the thesis that
abilities but over authenticity.) Thus while recognizing the all disagreements 'of opinion are undecided.
existence of an unresolved dispute over authenticity, I may yet act The second diversionary point is this. The 11!0de ano 01mpwvias
as if the work is spurious (and perhaps even take it to be probably does not really depend on 01aqiwvia, and the Principle of Disagree
spurious); but I cannot rationally believe that it is spurious. ment is not really a principle of disagreement. The Principle states
If this putative counter-example to the Principle of Disagree that I must suspend judgement over ?Q if I am aware of an
ment is incoherent, then any putative counter-example is incoher undecided dispute over ?Q, that is to say, if I am aware (i) that
ent. And thus the Principle is true. Then since the Principle on different people have taken different attitudes to ?Q, and (ii) that
which the mode of disagreement rests is true, the mode does indeed no decisive reasons have yet been adduced in favour of any answer
22 23
to ?Q. Now it is plain, I think, that the Principle would remain true at all. Yet he is undisturbed by this seemingly damaging admission;
even if clause (i} were omitted; for what grounds an_d warrants the for in such cases, he explains, we should reflect that there might be
Principle is the connexion between suspension of judgement and disagreement.
the lack of decisive reasons - and it is clause (ii) which makes this
If on some issues we cannot immediately produce an anom
connexion. Yet clause (ii) does not invoke the notion of dispute. aly, then we should say that in some countries unknown to us
Dispute is invoked only in clause (i), which is otiose. it is possible that there is a disagreement even on these
Thus the mode from disagreement depends on the Principle of matters. (PH 111 233)
Disagreement. The Principle of Disagreement makes reference to
61acpwvia, but it does so, as it were, accidentally. Hence it is only a And we might, I suppose, be prepared to grant that on any issue
superficial feature of Agrippa's first mode that it deals with dis whatever, there may, for all we know, be a disagreement. For even
agreement. Essentially, the mode has nothing to do with 5tcx<pwvicx. if everyone I have ever come across takes a uniform attitude to
(But I shall, on good conservative principles, continue to talk of the some question or other, nevertheless other people in other places
mode of dispute or disagreement.) may just possibly, for all I know, take a different line.
Let us allow to Sextus, then, that on every issue there is, for all we
Let me return to the earlier question. Agrippa, I said, wins a victory know, a 81acpwvia. Yet this concession is of no value to the
but how much territory does he actually gain? There is a Pyrrhonists. In granting it, we are only granting that there may be
difficulty with the argument which I based on the Principle of disagreement over every issue, that there is, if you like, potential
Disagreement. The difficulty is not that the conclusion of the disagreement over every issue. We are not granting that there is
argument is false. Rather, it is that the conclusion is too obvious actual disagreement over every issue. Still less, of course, are we
and commonsensical a truth. The Pyrrhonists advanced the mode granting that there is undecided disagreement over every issue. But
from dispute as a way of inducing a radical scepticism; for it the Principle of Disagreement involves undecided dispute. Hence
supposedly applies to 'any matter proposed' (PH I 165). But the we shall not and should not allow that, by way of the Principle, the
mode, as we have thus far understood it, seems to offer a sober and mode of 81acpwvia yields universal scepticism. There are (we may
cool-headed account of rational assent, and to be far removed from still maintain) topics on which no dispute actually arises, and there
any extreme form of Pyrrhonism. Why, in that case, did the are also topics on which any dispute that has arisen has already
Pyrrhonists make such a fuss over the 6tacpwvia mode? The answer been decided. On these topics at least, so far as the mode of
lies in their thought that there is 'undecided faction' on absolutely 8taq>wvia is concerned, we are at liberty to make non-sceptical
every topic (PH I 165), or, as Sextus puts it a little later, that 'things assertions. The mode of disagreement may indeed, as we have seen,
plain and things unclear are all of them disputed' (PH I 185), which lead to local scepticisms. But it seems that it will at best produce a
is to say that everything is disputed. If there is undecided dispute modest, urbane, bourgeois little scepticism. It will not yield the red
about ?Q, l shall suspend judgement over ?Q. If there is undecided radicalism of the Pyrrhonists.
dispute over every topic, I shall suspend judgement over every
thing. Thus the Principle of Disagreement generates universal If Sextus is to reach a red radicalism, he must somehow start from
scepticism - provided that there is undecided disagreement the assumption of disagreement simpliciter (or even of potential
everywhere. disagreement) and not from the assumption - implausble if ex
But whyever concede that undecided dispute is ubiquitous? Why tended to 'any matter proposed' - of undecided disagreement. And
concede even that there has been dispute of any sort on literally yet he must somehow also manage to make use of the Principle of
every topic? After all, Sextus himself appears to allow that, in some Disagreement, which refers only to undecided disagreement. He
cases at least, there may not, so far as we yet know, be any 8tacpwvla must contrive to have it both ways. Now there are in fact several
arguments in which Sextus contrives, or purports to contrive, to or proof; we require a mode or means of deciding the issue: we
have it both ways. Some of these arguments involve other Agrippan require, as I shall put it, a yardstick. And this claim must seem
modes, and in the last chapter of the book I shall consider, gener highly plausible. For what is the alternative? Only, it seems, an
ally, the ways in which the Agrippan modes may work in collabo arbitrary decision to opt for one side of the dispute or another- and
ration to a sceptical end. But even if these arguments are successful, that can hardly count as a resolution of the dispute. There is
they will not show that the mode of disagreement by itself can disagreement over ?Q. Some hold thatP,, some thatP,. We wish to
'
induce radical scepticism . So for the present, let us restrict our determine who, if anyone, is in the right; we want to decide which
attention to the mode of disagreement, and ask whether this mode, answer, if either, is true. And for that we need more than an
by itself, can possibly produce large-scale suspension of judgement. arbitrary and capricious act of plumping. We require to master
I shall look at one particular argument which purports to have just some means of deciding the issue. In Sextan terms, we need a
this outcome. criterion or a sign or a proof. We need a yardstick.
Fairly often, the Pyrrhonists do seem to think that they can argue But, as Sextus himself insists at length in PH II and M vn-vm,
from mere disagreement (rather than from undecided disagree every attempt to produce a yardstick has been subject to deep
ment) to suspension of judgement, and that they can do this disagreement. Our primary problem, ?Q, thus raises a secondary
without invoking any other sceptical argument-forms. 26 The texts problem (or 'meta-problem'), viz.: How are we to decide ?Q itself?
are not exactly probative - sometimes it may seem reasonable to What is an appropriate yardstick in the case of ?Q? And this meta
'understand' a qualifying avmiKplTOS with the word 510<pwvla, problem is, according to Sextus, itself involved in undecided dis
sometimes it may seem reasonable to think that Sextus is tacitly pute. Thus every disagreement leads to a further disagreement,
alluding to the adjunction of further argument-forms. But I think namely a disagreement over yardsticks, and this disagreement is
that in many of these texts it is most plausible to suppose that itself undecided. But if the disagreement over yardsticks is
Sextus does wish to move from disagreement to suspension of undecided, then we surely cannot properly use a yardstick. Hence
judgement without qualification or ado. And Galen, in the passage every dispute is undecided.
from Against julianus I cited earlier, implies that it is characteristic Sextus expresses the point like this: 'Surely the dispute among
of the anopT)TIKOS to take exactly this leap. the philosophers over the highest things does away with knowledge
Can we explain, or even justify, the leap? Sextus frequently of truth?' (M vn 369). The 'highest things' include, of course,
reminds us that if an issue is under dispute, then it is unclear or criteria and signs and proofs. Sextus means that if you dispute over
&B11i\ov,27 and that we therefore need some criterion or sign or the procedures for judging disagreements, then everything else is
proof if we are to decide it. 2 8 It is puzzling that Sextus should indirectly involved in the dispute and hence a matter for suspension
profess that all matters subject to dispute are thereby 'unclear' or of judgement. The argument is given fairly explicitly in the follow
&811i\a; for he also expressly says, at PH I 185, that cpa1v6µeva, i.e. ing passage.
things which are clear, are subject to dispute. But Sextus' use of the
term &5T)AOV is notoriously perplexing - and fortunately the point There has been disagreement among the natural philosophers
over all things, I suppose, whether objects of perception or
does not matter greatly here. For here the crucial claim is that
objects of thought. And this disagreement is undecided, since
where there is disagreement, there we require some criterion or sign
we cannot use a criterion, whether an object of perception or
26 See PH I 163, 177; 11 153; 111 13, 197, 235, 238; M VIII 2, 353, 356; IX an object of thought - for everything, whatever we take, is a
191. matter of disagreement and hence lacks warranty. (PH I 178)
27 See e.g. PH II lI6, 168, 182; M VIII 328, 334, 335; M 11 108.
28 See e.g. PH II 172, 182; M VII 341, 346; VIII 178/82, 317, 341, 346,
351, 365, 430; XI 177· Also PH 111 6, where an undecided
Note two things about this argument. First, it takes as its premiss a
disagreement is said to lead to 6:5rii\6TTJ5 and to require proof. claim about the prevalence of disagreement, and not a claim about
the prevalence of undecided disagreement. It supposes only that almost all important issues, as well as most minor issues, have in
everything has been disputed - and perhaps the argument need only fact been subject to disagreement. Thus even if we rejected (r), we
assume that everything is, for all we know, disputed. The argument should still, as far as the rest of the argument goes, find ourselves
starts not from undecided disagreement, but from ordinary dis reaching a disquietingly Pyrrhonian conclusion. For we should
agreement - and perhaps it may start from potential disagreement. have to replace (r) by a modification which would be strong
Secondly, the only sceptical mode which the argument requires is enough to induce scepticism over almost all major and most minor
the mode from disagreement, a mode which, I have argued, relies, issues. We should not end up as radical sceptics, but we should be
in one of its applications, on an indisputably sound Principle. Thus left believing nothing of interest or importance.
if the argument works, then it establishes a radical, general scepti As for premiss (2), I have already argued for it; and I think we
cism from disagreement alone, without invoking either undecided should accept it as a truth about rationality.
disagreement or any supplementary Agrippan modes. Again, the logic of the argument seems impeccable. From step (4)
Sextus' argument at PH 1 r78 is, I think, interesting and impor onwards there is nothing to question.
tant. I want to set it out as formally and as rigorously as I can. In Hence in order to avoid something close to radical Pyrrhonism
doing so, I shall tacitly modify or emend Sextus' statements at a few we must reject premiss (3) or premiss (4) or both. And indeed both
points; but I think the version I offer can decently be called an these premisses are questionable. I shall end this chapter with some
interpretation of Sextus' words. Here, then, is the Sextan argument inconclusive reflections upon them.
in white tie and tails.
(1) On every issue ?Q there has been (or might be) disagreement First, what of premiss (4)? Are there really undecided disagree
(2) If a disagreement is to be decided, then we need a yardstick to decide ments about every yardstick? Sextus himself argues at considerable
it length to the effect that there are undecided disputes about the
(3) If we are to use yardstick Y for issue ?Q, we must be justified in criterion of truth and about signs and proofs; but these arguments I
holding that Y is appropriate for ?Q shall ignore here, for they adduce other sceptical argument-forms,
(4) On any question of the form 'ls Y appropriate for ?Q?' there is and the interest of the present argument lies precisely in the fact
undecided disagreement that it appears to argue from disagreement to scepticism without
Hence, from (4) by the Principle of Disagreement: adducing any other forms of sceptical argumentation. Of course,
(5) For no Y and no ?Q are we justified in holding that Y is appropriate Sextus' other arguments may be sound; and then he will be able to
to ?Q defend (4). But that is not my present concern.
In any case, (4 ) looks as though it is meant to be construed as an
Hence, by (3) and (s):
empirical premiss. The people who have thought about yardsticks
(6) For no Y and no ?Q may we use Y for ?Q
are - or, in Sextus' time, were - philosophers; and they habitually
Hence, by (r) and (2) and (6): disagree among themselves, never more so than over the criterion
(7) No issue is decided of truth and the theory of signs and proofs. So I think we may allow
it to be true that on every question of the form 'Is Y appropriate to
Hence, by the Principle of Disagreement again:
?Q ?' - on every question of this form which had been raised - there
(8) On every issue ?Q, we should suspend judgement.
had been philosophical disagreement.
When an argument is set out as fussily as this, it should be easy But was the disagreement undecided? (For premiss (4 ) essentially
enough to see where its faults are - if it has any faults. demands that the disputes over yardsticks be undecided.) Here we
We could always reject premiss (r), but this would be a tedious need to attend more closely to the notion of decision, which I have
option. For we shall surely grant, as a plain empirical truth, that so far used without any particular explanation. We need to answer
the question which I earlier postponed: What is the force of the school has ever persuaded all the members of every rival school to
verbal root EirtKptv- in the adjective clveiriKptTos? accept its views. (And in Sextus' time, as I indicated at the begin
In supporting premiss (4), Sextus might think to offer us the ning of this chapter, it was a famous fact that philosophers never
following simple reflection: Epicureans and Stoics have been in came to any agreements with one another.)
dispute for years over the criterion of truth, the Epicureans But it is also clear that historical decisions are not the sort of
advancing and the Stoics rejecting the sovereign claims of sense decision which bears upon the mode of disagreement and its
perception. The dispute still rages, neither party being prepared to sceptical claims - they are not the sort of decision which the
concede an inch to the other. Thus plainly the disagreement is Principle of Disagreement invokes. Suppose I am interested- again
undecided - were it decided (say, in the Stoics' favour), then one -in the problems of the Magna Moralia. I know that there has been
party (here the Epicureans) would have thrown in the sponge. a serious disagreement over the matter, and I know that the
The notion of decision involved in this little argument is closely partisans of authenticity still write articles against the partisans of
connected with the notion of persuasion. In this sense of 'decide', a spuriousness, and vice versa. The disagreement has received no
dispute is decided when and only when the parties to it come to an historical decision. No peace treaty has been signed. Recognizing
agreement or form a consensus- when, instead of a 'truceless war', this fact, am I rationally obliged to suspend judgement on the
there is a peace treaty signed and ratified by all the belligerents. question? Evidently not. For I may well say something like this:
There is, I think, little doubt that Sextus sometimes has this notion 'The two sides are still quarrelling. But the issue has been decided -
of decision in mind. For example (the context is irrelevant): for someone has in fact produced decisive arguments in favour of
spuriousness, whether or not the partisans of authenticity are ready
If this is to be agreed upon, there must first be a consensus and
an agreement among all the natural philosophers about the to recognize it.'
existence of perceptible objects . . . But there has been no In thinking in this way I employ a different notion of decision
consensus. Rather, from the one I sketched before. A disagreement can be decided, in
as long as water flows and tall trees flourish29 this second sense, even if all parties to it continue the fight. (And
the natural philosophers will never cease from warring with conversely, a disagreement may remain undecided in this sense
one another on the matter. (M VIII I83-4) even if the parties to it have all signed a binding treaty of agree
ment.} For the 81aqiwvia is decided (in this sense) just in case
Here the disagreement is plainly construed as an actual, historical
sufficient reason has been produced to determine that P1 (say) is the
dispute among the philosophers; and agreement or decision - the
correct answer to ?Q. In considering whether a dispute is, in this
settlement of the disagreement - is construed as a sort of peace
sense, decided, I shall not be particularly interested in the attitudes
treaty to which each warring party is a signatory and by which each
of the parties (except insofar as these attitudes may be evidence for
party binds itself to a common view. Let us call this sort of decision
the existence of plausible arguments on each of the conflicting
- for want of a better phrase historical decision.
-
sides); rather, I shall be concerned with the arguments for and

Now if we think of decisions in this way, it seems clear that there
against rival answ�rs to ?Q. And I shall regard the dispute over ?Q
is undecided dispute over pretty well every issue about yardst:icks.
as decided - whatever the parties to it may think - just in case
Perhaps some disputes have been, in this �.-.�nse, settled - this is an
among these arguments I find one which determines the issue in one
historical question to which I do not know the answer. But it is
direction or another. Call such decisions rational decisions.
plain that all the chiefproblems-I mean, all the chief philosophical
It is evidently rational decision which is relevant to scepticism
problems - remain historically unresolved, in the sense that no
(and which I implicitly used in stating and arguing for the Principle
29 The line, from an anonymous epigram on Midas {see Plato, Phaedrus of Disagreement). And it is clear, too, that it is rational decision
2640), is quoted again by Sextus, in similar contexts, at Pfl 11 37 and
M 1 28. which Sextus often intends when he uses the verb Err1KplvEtV and its
30 JI
cognates. Should we then say that En1Kpive1v is ambiguous, some because I read it in the tea-leaves, or because I fallaciously inferred
times meaning historical decision and sometimes rational? I am it from a set of false premisses, or because it was suggested to me
unsure; but I incline to think that ETrtKpivetv always means 'decide' under hypnosis by a malignant hypnotist or . . . None of this need
in the rational sense, but that Sextus sometimes - and misleadingly matter a hang (we are supposing) so long as I have the belief. But
- pictures or dramatizes a rational decision as an historical plainly it does matter a hang. Well, that is rhetoric, not argument;
decision. but I find it persuasive rhetoric, and I do suppose that to maintain
However that may be, it is very far from plain that, in this sense (3*a) and deny (3''b) would be paradoxical.
of 'decide', premiss (4) of our Sextan argument is true. Have no But what of (3*a)? Should we accept this part of (3) as a truth?
disagreem�nts about yardsticks been rationally decided? Has no Plainly, someone might in fact use Y for ?Q without holding that Y
determining reason ever been produc:ed for any dispute over yard is appropriate for ?Q. He might use Y without realizing that he is
sticks? At the very least, we shall need some good reason to affirm using Y, and a fortiori without having any opinions about its
premiss (4) - it cannot simply be accepted as a highly plausible appropriateness for ?Q. Or he might be aware that he is using Y and
empirical conjecture. Now Sextus does, as I have said, argue for (4); yet never have considered the question of whether Y is appropriate
but his arguments appeal to other sceptical modes, and to the to ?Q. A commonplace example of the first kind of case: we all use
extent that they are needed to support (4), to that extent the certain basic logical schemata in testing ordinary arguments; but
argument before us does not show that 61mpwvia by itself can few of us are aware that we are using these procedures; few of us
induce scepticism. use them selfconsciously. (As John Locke said, God did not make
I shall not positively assert that premiss (4) is false. But I do say men bare two-legged creatures and leave it to Aristotle to make
that it demands special support from other Pyrrhonian arguments them rational. Aristotle codified, for the first time, certain logical
if it is to be made at all plausible. And in that case, as I have said, the procedures; yet men had been using such procedures, correctly but
argument loses its special appeal. unconsciously, for centuries.) An example of the second kind:
when I wanted to get an outside line on my old college telephone, I
Finally, what of premiss ( 3 ) ? This is, I think, the most interesting used regularly to depress a button marked 'Recall' before dialling
part of the argument. Let me repeat the proposition: the number. I do not know why I followed this procedure, but I
(3) If we are to use yardstick Y for issue ?Q, we must be justified in followed it consciously. A colleague once saw me phoning and
holding that Y is appropriate to ?Q asked me why I was depressing the button. I realized that I had
never really thought of the matter - I certainly did not actually
We can usefully distinguish two 'parts' to this premiss, thus:
believe that depressing the button was an appropriate (part of) the
(3'�) If I am to use Y for ?Q, then procedure for getting an outside line. (And in fact it was not:)
(a) I must believe that Y is appropriate to ?Q, and Thus it seems. clear that believing Y to be appropriate for ?Q is
{b) that belief must be justified not a necessary condition for using Y for ?Q. ls it a necessary
condition for justifi"ably using Y for ?Q? It is not clear that it is. For
Now we shall surely agree that if part (a) of (3") is required, then suppose that Y is in fact the appropriate procedure for ?Q, then -
part (b) must be required too. Otherwise we should be maintaining even if I have no beliefs about Y at all - may I not be justified in
that, if we are to use Y, then (a) we absolutely must believe that Y is using Y for ?Q, and justified precisely because Y is appropriate? Let
appropriate and yet (b'") we need not be justified in that belief. That me try to make the point absolutely plain. Sextus supposes, in his
seems absurd. How could my using Y depend essentially on a belief premiss (3), that
about Y but yet be indifferent to the justifiability of the belief?
If you are to use Y for ?Q, you must have reason to believe that
Suppose I do believe that Y is appropriate to ?Q, but hold this belief
Y is appropriate to ?Q
32
33
suppose that I am justified in using stylometrical methods to
The counter-suggestion, which we may imagine being advanced by determine the authenticity of the Magna Moralia.
a Dogmatist, is this:
If this last thought is right, then even if premiss (J) of the Sextan
You may be justified in using Y for ?Q provided that Y is in argument is false, we can readily find a suffect which is true,
fact the correct means of deciding ?Q; you do not also need namely:
to believe that Y is appropriate to ?Q.
(3 + ) If we are to use Y for ?Q, then it must not be the case that, having
ls the counter-suggestion correct? If it is, then premiss (3) is false, considered the question 'Is Y appropriate to ?Q?', we return a
negative or a sceptical answer to it
and the Sextan argument from Stmpwvia fails.
The Sextan supposition and the Dogmatic counter-suggestion Now proposition (3 + ) is weaker than (3). lt is not strong enough to
together raise - or so I believe- a problem which lies at the heart of sustain the Sextan argument from disagreement as it stands - I
all epistemological reflection. I shall say no more about the prob mean, it cannot replace (3) in the argument and provide the same
lem here, but it will re-emerge at the end of the final chapter. logical power as (3) provided. So I do not offer it to the Pyrrhonist
One further turn to the story. Suppose that - prompted by as a simple way of repairing the argument from disagreement. But
Sextus' argument - I do consider the question of the appropriate (3 + ) has an interest of its own. And it lies here as a warning to the
ness of Y to ?Q. And suppose that either I decide that Y is not Dogmatist; for it exemplifies a form of reflection which will turn
appropriate to ?Q or else I find myself in the sceptical position of out to be the final cunning thought in the Pyrrhonian philosophy.
not knowing whether Y is appropriate to ?Q. In these circum
stances would I be justified in using Y for ?Q? Surely not. Or rather,
surely I myself would not feel justified in using Y. Even if the
Dogmatic counter-suggestion is correct, so that we may justifiably
use Y without having reflected upon it, none the less once we have
reflected on Y and found its credentials dubious, we shall surely not
feel ourselves entitled to rely upon it.
Return, once more, to the Magna Moralia. This time suppose
that I engage in stylometrical studies to determine the question of
authenticity. Stylometry, I tacitly imagine, supplies a suitable
yardstick for such questions of authenticity. But suppose next
(what is, of course, highly likely) that someone raises the question
of the appropriateness of current stylometrical methods to prob
lems of authenticity. Suppose further (what is not implausible) that
I decide that stylometrical methods are inappropriate to the prob
lem, or else (more modestly) that I find it simply unclear whether or
not stylometrical methods are appropriate. In these cases it seems
plain that I shall not, and rationally cannot, continue to attack the
problem of authenticity with stylometrical strategies. No doubt I
may go on making stylometric tests - for fun, perhaps, or because I
can think of nothing else to do, or in the hope that they may
eventually turn out to be appropriate after all. But I can hardly
34
35
Infinite regression
follows. Were the extracosmic void finite in extent, it would be
2 bounded by something. Not by body, since by definition all body is
included in the cosmos. Hence by something incorporeal; and the
only possible candidate among incorporeal things is void.
Infinite regression Thus there will have to be another void containing the
extracosmic void. And this, not being infinite, will have to be
contained by another; and that by another, ad infinitum.
Thus if we are unwilling to allow that the void outside the
cosmos is infinite, we shall be reduced to the necessity of
supposing infinitely many different voids - and that is utterly
absurd. Hence we must agree that the void outside the cosmos
is infinite. (disc eye! r i 8)
Zeno of Elea was the first philosopher to spin puzzles from the
The example is pretty; for Cleomedes adduces the absurdity of a
notion of infinity. Thereafter, thinkers of every persuasion were
regressive infinity to show the necessity of a non�regressive infinity
obliged to perpend the paradoxes of the infinite. They were also
- since there cannot be an infinite sequence of distinct vacuums, the
ready to use the concept of infinity in destructive arguments. And
one vacuum must be of infinite extent.
they were particularly fond of threatening any view opposed to
Destructive arguments from infinite regression work in the
their own with an infinite regression. Here are two characteristic
following way. (r) We hypothesize the claim which the argument is
examples.
intended to destroy (the claim that souls are corporeal or that the
Nemesius ascribes the following argument to the Platonist philo
extracosmic void is finite). (2) From this hypothesis we next
sopher, Ammonius Saccas. Ammonius has argued that all bodies,
generate, by some process of inference, an infinite sequence of
being essentially divisible, require something to conserve them or
objects (an infinite sequence of souls or of voids). (3) We deny the
hold them together, and that this conserving or containing force is
possibility of such a sequence. And hence (4) we reject the
properly called 'soul'. Well then, Ammonius asks, can souls them
hyothesized claim.
selves be corporeal?
Note that the sequence generated at stage (2) is literally infinite:
If the soul is a body of any sort, however rarefied, there will it is not merely very long, nor merely unimaginably long - it is
again be something to hold it together (for it was proved that infinitely long. (It contains as many members as there are positive
every body requires something to hold it together). And so ad integers: its members may be paired one to one with the integers in
infinitum until we come to something incorporeal.
the infinite sequence ( 1 , 2, 3, . . . ).) Now it is not true - nor did any
(nat hom 70 M )
ancient philosopher suppose - that any infinite sequence will
If the soul i s a body, i t must itself be conserved by another soul. suffice to ground a.n argument from infinite regression; for there
This second soul, being ex hypothesi corporeal, will again be are infinite sequences against which no objection can be raised.
conserved by yet a third corporeal soul. And so on. Thus we shall (The sequence of positive integers is infinite. But not even Aristotle,
have an infinite sequence of corporeal containing souls. But that - the most notable ancient opponent of the infinite, thought that
or so Ammonius assumes - is absurd. Therefore the soul is there was anything disreputable about this sequence.) Yet equally,
incorporeal. not all infinite sequences are tolerable, and where a sequence is
Again, in his Elementary Theory the astronomer Cleomedes intolerable a reductio ad infinitum - I mean, an argument of the
argues that there is a vacuum or void outside the cosmos, and wants form I have just described - may prove a sound and powerful
to prove that it is of infinite extension. One of his arguments runs as device.
37
In-finite regression Infinite regression
In ancient texts, the claim which is to succumb to an infinite display the form of the reasoning used by the two Ammoniuses.
regression often has an epistemological content- it is a claim to the For this form of reasoning was especially dear to the Pyrrhonians.
effect that a certain proposition or type of proposition can be Regression ad infinitum, 1) eis Cirreipov EKrr1wcr1s, grounded the
established or proved or known in a certain way. Thus a later second of the Five Modes of Agrippa - the second of the set of five
Ammonius employs an epistemological regression in arguing that argument-forms by the application of which the Pyrrhonians
'hypothetical' syllogisms cannot by themselves suffice for the pre claimed to introduce universal irroxiJ or scepticism (PH 1 r66). And
sentation of arro8ei�EIS or proofs. By a 'hypothetical' syllogism Sextus appeals, time and again, to infinite regressions in his argu
Ammonius here means an argument with two premisses, of which ments against the Dogmatic philosophers. The Dogmatists them
one is a compound proposition (a conditional or a disjunction) selves use regression for Dogmatic ends - to refute an hypothesis
and the other (called the 'extra assumption') a simple proposition. and thereby establish its contradictory. The Pyrrhonians, of
Now course, have no such designs: their regressions serve a sceptical
hypothetical syllogisms assume without proof what is called purpose and introduce only irroxiJ.
their further or extra assumption (and sometimes too their
The regressions in Sextus are not all of them epistemological; and
conditional or disjunctive component needs argument). Thus
they need premisses to warrant their original premisses. Now some of Sextus' comments on his non-epistemological examples
1
consider how to establish these premisses: if you use another may aid our understanding of the epistemological cases. But
hypothetical syllogism, then you will need another way to epistemological examples are - unsurprisingly - by far the more
establish the warrantability of its premisses, and another for frequent. Sextus turns eagerly to �egressions of proofs and regres
that, and so ad infinitum, if you want to corroborate sions of criteria. If a philosopher offers a proof, he will then be
premisses by way of premisses. required to prove the premisses of his proof, and the premisses of
(in Int 3.19-26; cf. [Ammonius], in APr 67.11-15) that proof, and so ad infinitum; or else he must prove the soundness
The claim which Ammonius examines is the claim that hypotheti of his first proof, and then the soundness of that proof, and so ad
cal syllogistic alone suffices to prove certain sorts of proposition. infinitum.2 If a philosopher suggests a criterion of truth, he must
This claim, Ammonius argues, generates an infinite sequence of produce a second criterion by which to show that the first criterion
syllogisms in any putative proof; for each syllogism in the proof is adequate, and a third for the second, and so ad infinitum.3 In a
will contain at least one premiss which requires further support - similar way, Dogmatic philosophers are entangled in infinite se
and hence a further hypothetical syllogism. But such infinite se quences of signs or of explanations or of definitions.4 Sometimes
quences of syllogisms are impossible. Hence hypothetical syllogis the Dogmatic philosopher or his philosophy 'falls into infinity
tic alone is not enough and a different form of argument - a (EKrrirrTetv eis O:rretpov)'; sometimes the Pyrrhonist 'tosses' him
'categorical' syllogism - must be introduced in order to complete there (iKj3aMetv). The result is the same: the Dogmatist disappears
the proof and to block the threatening regression. and the sceptic triumphs.
Ammonius' argument is formally parallel to that of his earlier 1 For non-epistemological regressions see: PH II 40; 111 44, 67, 68, 76,
162; M VII 312; IX 221, 261, 435; X 20, 76, .r39, 256; M I 180, 242/3;
namesake. Just as an incorporeal soul is required if an intolerable Ill Sr. As the references suggest, and as we should in any case have
infinity of bodies is to be avoided, so a non-hypothetical syllogism guessed, such regressions are most common in the Pyrrhonian attacks
is required if an intolerable infinity of hypothetical syllogisms is to on aspects of qiucruo'J.
2 Sequences of proofs: PH 1 122; 11 85, .r82; 1 1 1 8, 36, 5 3 ; M VII 339; VIII
be avoided. 16, 21, 347; M II .ro9, I.12.
Neither of the Ammonian arguments should seem immediately 3 Sequences of criteria: Pil 11 20, 36, 78, 89, 90, 92/3; 111 36, 24.r; M v11
340, 429, 441; VIII 19, 78; cf. PH I .r71, 172, 176; M VIII 28.
convincing - each makes some pretty questionable assumptions.
4 Sequences of signs: PH 11 124, r28; of explanations: PH 1 186; 111 24;
But here I am concerned not to assess the soundness but only to of definitions: PH n 207. See also PH 1 .r79; M VIII 4_9-50.
39
Infinite regression lnfinite regression
There is a pleasing parody of the Pyrrhonian use of regression in appearances and hold that it represents the real nature of the
Lucian. object. Now suppose that someone contests this, alleging that he
Imagine that we find someone who claims to know a proof can prove one appearance to be preferable to the others. What will
and to be able to teach it to someone else-I suppose we shan't Sextus say in reply?
believe him without ado but shall look for someone who can
judge if our man is speaking the truth. And if we are lucky If he wants to use a proof, . . . then if he says that the proof is
enough to find such a person, it will still be unclear to us true he will be asked for a proofthat it is true, and for another
whether this arbiter knows how to distinguish good judges proof for that proof (for it, too, must be true), and so ad
from bad; and we shall, I suppose, need another arbiter for infinitum. (PH I 122)
this arbiter - for how could we know how to pick out a man
Thus the Dogmatist holds that P (say, that Mount Vesuvius actu
who is good at judging? Do you see where this is leading? It's
ally has the shape it appears to have from the Palazzo Tarsia), and
getting infinite and can't ever come to a stop and be grasped.
(Hermotimus 70) he offers some reason R 1 in proof of P (say, that our eyes function
most accurately when they are at a certain distance from the objects
You are puzzled whether the first argument in Euclid's Elements is of vision). His claim is that R 1 establishes P. But this claim - or so
a sound proof. You call in an expert to pronounce on the question. Sextus alleges - opens up an infinite regression. For the dogmatist
But is the expert competent? Well, you had better call in a second must next offer some further reason, R2, to prove that R1 is a true
expert to assess the competence of the first. And a third for the proof of P (he must show that the claim about our eyes really does
second. And so ad infinitum. Since you cannot consult an infinite support the claim about Mount Vesuvius); and then he must find a
series of experts, you cannot determine whether or not Euclid's further reason, R3, for R2 and so on. Thus he must offer a sequence
argument is sound. Hence ETioxfi. of reasons, (R 1 , Ri. . . . , R0, • •) , where each Ri shows that its
•
In his account of the Five Modes Sextus gives a brief description predecessor is a true proof. The sequence is infinitely long. But
of his epistemological use of infinite regressions: there cannot be such an infinite sequence. Therefore the Dogmatist
cannot prove that P by adducing R 1•
In the mode deriving from infinite regression, we say that
I shall not comment on the soundness of Sextus' argument at PH
what is brought forward as a warrant for the matter in
I 122. I cite it to illustrate the regressive mode in action. It is, from a
question itself needs another warrant, which itself needs
another, and so ad infinitum. Thus we have nowhere whence formal and stylistic point of view, a typical example; and there is no
to begin to establish anything, and suspension of judgement need to produce further illustrations. Rather, I shall concentrate in
follows. (PH 1 166) what follows on the two main philosophical questions which the
mode of infinite regression raises. First, what exactly is wrong with
(The description in Diogenes Laertius' parallel account, at IX 88, is infinite epistemological regressions? (Indeed, is there anything
briefer, and it adds nothing to what we learn from Sextus.) wrong with them 0s such?) Secondly, why should the discovery of
The first application in Outlines of Pyrrhonism of the mode of such a regression lead us, as Sextus thinks it must, to a sceptical
regression is actually made before the mode has been officially suspension of judgement? I take the questions in reverse order.
introduced. According to the fifth of the set of Ten Modes ascribed
to Aenesidemus, we shall be led to scepticism once we reflect on the Let us grant, for the sake of the next stretch of argument, that
fact that objects look different when viewed from different dis infinite epistemological regresses are indeed unacceptable. Sup
tances or in different places or in different postures. For we are thus pose that a philosopher makes some claim - say, that things really
confronted by a number of mutually incompatible appearances, are the way they seem from nearby - and that he then supports this
and - so Sextus maintains - we cannot prefer one of the variant claim by a procedure which admittedly opens up an infinite
Infinite regression In'{inite regression
regression. Should we therefore suspend judgement about his Sextus' writings. The Greek for it is Oaov Errl aoVTcp (Ocrov Err\ aolc:;
claim? Should we decline to believe that it is true and also decline to J\eyoµEvotc:;, Tc{:> J\6ycp, etc.).5 And I suppose that Sextus' general
believe that it is false? Sextus does indeed sometimes appear to view about the regressive mode is this: if you are given a regressive
suggest that in such circumstances we should (or shall) end in argument for a claim, then Oaov Errl ToVTcp you will suspend
scepticism. And that, after all, is surely what it means to say that the judgement about that claim. Admittedly, I can produce no text in
regressive mode is a mode of Erroxi\, a mode which induces suspen which Sextus actually says this; but his normal sceptical procedure
sion of judgement. implicitly shows that it was in fact his view. For normally the
But it is evident that, in the supposed circumstances, we should regressive mode is used not alone but in concert with other argu
not and would not automatically suspend judgement - and for a mentative strategies. Schematically: someone urges that P. Sextus
trivial enough reason. Although our philosopher may have failed replies: 'If your claim that P is warranted, then either X or Y or Z.
to make out a case for his claim, there may yet be a case, and But X leads to an infinite regression, and Y and Z are unacceptable
someone else may have argued for the claim in a way which avoids for different reasons. Hence your claim is unwarranted, and we
any regression. The fact that one man produces a bad argument for shall suspend judgement.' Plainly, within such a strategy, the mode
the claim that P is in itselfno reason at all for suspending judgement of regression has hypothetical force: as far as it is concerned,
about the claim. It is a reason for criticizing the claimant, but it is no suspension of judgement must follow; if nothing else is to hand, we
reason for doubting his claim. (If Sextus' argument at PH I 122 is must suspend judgement.
sound, then we shall infer that a Dogmatist cannot prove P by We can thus see - and it is worth insisting - that the mode of
adducing R,. Evidently, we cannot make the further inference that regression and the mode of disagreement relate to suspension of
we must suspend judgement over P. For the fact that one Dogmatist judgement in rather different ways. And on two counts. First, as I
has made a hash of his defence of P does not show that there is no have just explained, the regressive mode will rarely, if ever,
reason to believe P. Equally, of course, it does not show that there is produce Erroxf} on its own; for we shall rarely, if ever, be con
no reason to reject P and to believe not-P .) fronted by a claim for which the only argument (or counter
Hence it would be absurd to think that the mode of regression in argument) is an infinite sequence of reasons. By contrast, the mode
and by itself is sufficient to produce broxfi. I do not like to think of disagreement, as I explained in the previous chapter, is supposed
that Sextus really entertained this absurd thought. Rather, we to be capable of inducing broxi\ by itself. Li1cxq>wvia is, both in
should make, and we should suppose that Sextus makes, a hypo principle and in practice, often an independent worker. Regres
thetical or conditional claim for the sceptical powers of the regres sion, in principle normally and in practice always, works as one of a
sive mode. What I mean is this. If the only consideration offered in gang.
support of a given claim leads to an unacceptable epistemological Secondly, and more interestingly, the regressive mode collabo
regression, then we must suspend judgement on the claim. The rates to induce scepticism not in virtue of its peculiar regressive
regression induces scepticism not absolutely but hypothetically; character but rather . in virtue of the fact that infinitely regressive
for it induces scepticism on the hypothesis that no other, non arguments are (we are supposing) bad arguments. For if I am right,
regressive, mode of support is available. the principle behind the regressive mode is in fact a wholly general
There is another way of expressing more or less the same point. principle. It is the principle that if the only arguments for or against
If our Dogmatic philosopher, maintaining that P, offers us an P are bad arguments, then we must suspend judgement over P.
argument which opens up a regression, then as far as that argument
And the mode of regression helps to induce suspension just inso
goes we shall and should suspend judgement. This particular far as regressive arguments are bad arguments. Contrast the
argument cannot move us from Erroxi\ �as far as it is concerned, we
mode of disagreement: here it is the fact that there is undecided
must be sceptical.
See Karel Jan£Cek, Sextus Emf1iricus' Sceptical Methods (Prague,
The phrase I have just used, 'as far as this goes', is common in 1972), pp.13-20.
42 43
Infinite regression Infinite regression
disagreement- and not some more general fact of which undecided Sextus puts the point most clearly in connexion with a non
disagreement is one particular exa1nple - which of itself leads to epistemological regression. In PH 111 he considers the suggestion
ETioxi\. In these two ways, the mode of disagreement and the mode that bodies move insofar as each of their parts moves in sequence.
of regression are not on a level with one another. (When a train starts to move, first the locomotive moves and then
In the final chapter I will say something about the way in which each of the carriages moves in sequence.) Against this suggestion he
the mode of regression may collaborate with other sceptical proce says:
dures. Here I content myself with a highly hypothetical statement
If bodies, and the places and times in which bodies are said to
about the relationship between regression and suspension of judge
move, are infinitely divisible, then there will be no movement;
ment: if (i) arguments which generate an epistemological regres
for in infinite sequences (Ev 6:TTelpo1s) it is impossible to find
sion are unacceptable arguments, and (ii) the only argument we
any first element, from which what is said to move will first
have for a given claim is such a regressive argument, and (iii) the move.7 (PH III 76)
claim should not be accepted without argument, then we should
suspend judgement and remain staunchly sceptical about the In an infinite sequence there is no first element. But if a body moves
claim. I believe that this highly hypothetical statement represents by virtue of the sequential movement of its parts, then its first part
what Sextus ought to say, and perhaps what he means to say, about must move first, and so it must have a first part. Similarly, we may
the connexion between infinite regression and scepticism. I also suppose, if a Dogmatic philosopher is to establish a claim by virtue
believe that the highly hypothetical statement is true. (Indeed, I of producing a sequence of supporting claims, then he must first
suppose that it is little more than a tautology.) produce the first such claim - and hence there must be a first claim.
But in an infinite sequence of claims there is no first claim. Thus, on
But what of the first of the two questions I posed a while ago? Are these suppositions, bodies will not move and claims will not be
regressive arguments unacceptable? And if so, why are they established.
unacceptable? In other words, is the first hypothesis - clause (i) - in The objection to Sextus' argument is plain: it is simply false that
my highly hypothetical statement a true hypothesis? And if so, why an infinite sequence must lack a first member. The sequence of
is it true? natural numbers, (1, 2, 3, . . . ) , is infinite; its first member is r. A
When he employs the regressive mode and points to an infinite train of infinitely many carriages none the less has a first part, viz.
regression, Sextus usually leaves unspoken the thought that infinite the locomotive, and the locomotive will be the first part of it to
regressions are Bad Things. Occasionally he will note explicitly move. And our philosopher's putative sequence of arguments will
that a regress is 'impossible' without explaining why (e.g. PH 1 r22; also have a first member, namely the argument which has as its
II 182). But one objection to epistemological regressions is in fact conclusion his claim that P. Thus in the case imagined at PH I 122,
briefly stated in his description of the regressive mode itself: the first element in the infinite sequence of reasons is R1•
Sextus might reply that this is to invert the natural order of
Thus we have nowhere whence to begin to establish anything things: if I am to 'establish P by way of R., then I must settle R,
(µTj Ex6vTwv TfµWv TT66ev O:p�6µe8a Tils KaTacrKevils), and
before I assert that P; and similarly I must settle R, before I settle R,,
suspension of judgement follows. (PH I 166)
and so on. Hence the first shall be last: the apxfi of the sequence, the
In an infinite sequence we have no starting-point, nowhere TI66Ev item I must start from, is the last element in the sequence - but the
6:p�6µE6a; for such sequences, as he says elsewhere, have no 6:pxiJ sequence has no last element.
or are &vapxo1. 6 Now it is indeed true that the sequence (R1, R2, , Rn, . . . ) has
• • •
6 See PH rn 68, 76; M Vil 312; VIII 78; x 76, 139, 256; M 1 180, 242/3. 7 I read O:<p oi'.i TrpWToV KIVi]OETC11 {the MSS have npWTov, not
The thought is not peculiar to Sextus: see e.g. Aristotle, Phys 256a16. rrpWTov).
44 45
no last member. But is it true that in order to establish P I must for our infinite epistemological journey, we shall hardly find a
begin from a last member of the sequence? The question raises a finishing-point. For how can we possibly go through or survey
number of tricky issues concerning the notion of establishing or every member of an infinite series?
justifying a claim, and also concerning the relation of priority This objection to infinite epistemological sequences is no new
which this notion seems somehow to involve. These issues will invention. It was known to Aristotle, and he held it to be conclu
exercise me in my next chapter. But I think that the immediate sive. In a celebrated chapter of the Posterior Analytics, to which I
question can be answered fairly easily, without entering into these shall return, Aristotle reports briefly an argument used - we do not
deeper issues. Let us consider a normal argument, consisting of a know by whom - to a sceptical end.
finite sequence of considerations. Suppose we have a proof of one
These people claim that we are led back ad infinitum, since
of Euclid's theorems (call it theorem T). The proof can be represen
we cannot know the posterior items by way of the prior items
ted as a finite sequence of arguments
if there are no primitive items. And here they are right; for it is
(A1 , Ai , . . . , An) impossible to go through infinitely many items.
(APst 72b8-n)
where A1 yields T as its conclusion, A2 yields the premisses of A1 ,
and so on. The premisses of An are axioms or first principles of We cannot properly sustain the claim that P by an infinite sequence
Euclid's geometry. (In real proofs matters are neither so linear nor of reasons (R1, R2, • , Rn, . . . ); for no one can 'go through' or
• •
so neat; but the additional complexities of real proofs do not survey or produce such a sequence.
matter here.) It is plain, I think, that in offering you a proof of T I The objection, in its most general form, is a commonplace in
may go through the arguments in this sequence in any order I like ancient (and modern) thought: 6:8VvaTov TCx &rrttpa 8 ttA 6tTv, it is
(or in any order you may find perspicuous). In particular, I may impossible to survey an infinite set. Sextus himself applies it in
reason as follows: 'T is true, as you can see from Ai; and Ai I more contexts than one. Suppose, for example, that when faced
support by A,, and that by A,, . . . and that by Am the premisses of with a 8tacpwvia, you suggest that we should accept the opinion of
which are axioms of Euclidean geometry.' As far as I can see, there the majority of men.
is no reason to hold that in arguing for T I must start from An, the Now there is an infinite number of individual men, and we
last item in the sequence. But if that is so with a finite sequence, why cannot survey (ETIEA6Elv) the judgements of all of them, and so
may it not also hold of the infinite sequence (Ri. R2, •, Rn, . . . ) ?
• • assert what it is that the majority of all men assert and what
Why may I not happily start the infinite sequence from R,? Thus the minority. (PH II 45)8
Sextus' objection to infinite epistemological sequences does not
The case is curious (does Sextus really suppose that men are infinite
work: as far as this objection goes, Ocrov Erri ToVTtp, there is nothing
in number?); but the objection and the general principle are clear. If
wrong with epistemological regressions; for we can in fact always
we are to ascertain which view is the majority view, we must survey
find a starting-point for the argument.
the whole class of men and count the number of votes in favour of
each view. But there is an infinitely large class to survey - and we
Consideration of the schematic Euclidean proof will quickly sug
cannot survey an infinitely large class.
gest a second and quite different objection to infinite epistemologi
Sextus does once apply this objection to the matter with which I
cal regressions. In the Euclidean case I supposed, of course, that I
am primarily concerned. In the course of his sceptical attack on the
went through all the items in the sequence, that I produced each Ai
notion of truth, he brings up the mode of regression in the follow
in the sequence (A 1 , Ai. . . . , An). And here our Dogmatic
ing way:
philosopher's infinite sequence does seem to pose a problem. For
although we may, as I have argued, easily discover a starting-point 8 See also PH n 89; M 1 66, 224.
47
If (you try to show that something is true) by way of a proof, something else which I can do and which anyone who can count
we shall enquire again how you know that this is true, and so can do: we can each of us name the successor of any number. Give
ad infinitum. Now since, in order to learn that something is me any number whatsoever in the sequence, however remote, and I
true, you must first grasp infinitely many things, and it is
can at once name its successor. In this sense I can master the infinite
impossible for infinitely many things to be grasped, it turns
sequence - nor is my mastery in the least bit idiosyncratic or
out to be impossible to know firmly that anything is true.
remarkable.
(M Vlll r6)
The point is worth stating rigorously. There is something which
The standard Sextan objection to infinite regression - that no I cannot do, viz. name the successor of every number, and some
starting-point is available - is not entirely lost sight of here; for the thing which I can do, viz. name the successor of any number. Thus
claim that 'you must first grasp (rrpoAllq>6f\vm) infinitely many we must distinguish between
things' at least hints at the alleged impossibility of securing a first
point or apx!\. But the main weight in this passage evidently falls (r) It is possible that for every number n I name the successor
to n
elsewhere. The thought is that you cannot survey an infinitude of
propositions. and
Now it may s.eem disappointing, or even perplexing, that Sextus
produces this objection only once, in M VIII 16, whereas the other, (2) for every number n, it is possible that I name the successor
bad, objection is, so it appears, his standard or official objection to to n.
infinite epistemological regressions. For the objection of M VIII r6
The crucial difference between (r) and (2) is made by the difference
may actually seem to be a good objection, an objection which really
in the relative order and scope of the modal operator ('it is possible
does show that such sequences are unacceptable.
that . . .') and the quantifying phrase ('for every number n'). In the
But is it a good objection? We shall surely grant that no one
most general and abstract case, there is a difference between
can survey or produce each member in an infinite sequence. If I pur
port to rest my claim that P on the sequence of reasons <Ri, R2, ,
• • •
(a) It is possible that for every x, x is F
R n , . . . ), then I surely cannot produce or state or formulate every
R;. But how exactly does this impossibility validate the mode of re and
gression? The objection plainly supposes that I must be able to pro ([3) For every x, it is possible that x is F
duce or state or formulate every reason or argument in a sequence if
I am to base my claim on the sequence. But why should this be so? These two formulae are evidently different in their syntactic struc
Perhaps it just seems obvious that an appeal to a sequence of ture. They are also different - and this is the important point - in
arguments can be nothing other than an appeal to each and every what they say. For they are not equivalent to one another. In
member of the sequence? Yet it is not clear to me that it is obvious. particular, although if a formula of the form (13) is false, the
And in fact the objection, as we have it so far, glosses over an corresponding formula of the form (a) must also be false, the
important distinction. An analogy will help to show what I mean. converse is not the case: a formula of the form (a) may be false
The sequence of positive integers <r, 2, 3, . . . ) is an infinite when the corresponding formula of the form (13) is true. In other
sequence. Each member of the sequence has a unique (immediate) words, formulae of the form (a) entail corresponding formula of
successor (2 succeeds r, 3 succeeds 2, . . . , 737 succeeds 736, . . . - and the form (13), but not vice versa. A simple example should make the
so on, for ever). Neither I nor anyone else can name the successor of point clear. In a fair lottery with only one winning number it is false
every number in the sequence; neither I nor anyone else can in this that
\Vay survey the sequence or produce all its members. But there is (r�") It is possible that every number wins
49
-

but true that must not only reject (ia) but also reject (2a). And we have as yet
been given no reason to reject (2a) in all its generality.
(2 ") For every number, it is possible that that number wins.
Now sentence (2") is of the form (j3), whereas sentence (1*) is of the I believe that the preceding considerations are sound, and that they
form (a) . And that is how (2'' ) may be true while (1*) is false. demonstrate the insufficiency of the good-seeming objection to
Similarly, sentence (2) is of the form (j3), whereas sentence (1) is of infinite epistemological regressions. But how much is the demon
the form (a), and so (2) may be true while (1) is false. stration worth? To what extent are infinite regresses thereby
I have laboured this point because it is easily and often over seriously defended? Can there be epistemological sequences for
looked. Once it is stated clearly, its truth will hardly be disputed. which (2a) is in fact true?
But unless it is stated it is liable to be ignored. Consider again proposition (2 ) . How is it possible that I can
How does the distinction between (1 ) and (2), or between (a) and name the successor to any one of an infinite list of numbers? The
(j3), apply to infinite epistemological sequences? The objection we answer is plain: I know a general way of constructing the successor
are considering rests on the thesis, which is surely true, that we to any number; I have-as the logicians say -an algorithm by which
cannot survey every member of the whole infinite sequence of I can construct the successor to any number. For in order to
reasons or arguments. That is to say, it rests on denying a proposi construct the successor to n, all I need is the concept of adding one;
tion of the form (a), a proposition formally analogous to sentence since I have an entirely general grasp of this concept - of the
(1), viz. the proposition: operation denoted by · � + l' - I can thereby name the successor to
(1a) It is possible that, for every reason Ri, I produce the next any number. It seems evident that my mastery of the infinite
reason in the sequence, R;+,, in support of Ri. sequence depends upon - or consists in - my grasp of this algo
rithm; and it seems evident that in the absence of any algorithm of
Now there is also a corresponding proposition of the form (j3), a
this sort I could not (except by magic) come to master an infinite
proposition formally analogous to sentence (2), viz. the
sequence in the fashion of (2).
proposition:
Thus infinite sequences of reasons will satisfy (2a) only if they
(2a) For every reason Ri, it is possible that I produce the next are associated with some algorithm. If I have no general method for
reason, Ri + 1, in support of Ri. constructing R i + 1 from R i (no algorithm however complex), then
Just as we may consistently reject proposition ( 1 ) while accepting magic apart - I cannot hope to produce an R i + 1 for every R;.
proposition (2), so we may consistently reject (ia) while accepting At this point one matter - and a not unimportant matter -
(2a). becomes plain. For it is plain that epistemological regressions are
Then my suggestion is this: although (ia) is false, its falsity is not not epistemologically serious things. As a matter of fact, no philo
sufficient to show that infinite epistemological sequences are sopher has ever produced such a sequence, and no epistemologist
unacceptable. For the acceptability of such sequences may (for all has ever attempted. to ground knowledge on infinite regressions.
that has been thus far said) depend only on the truth of (2a). In When Sextus refers, as he frequently does, to the threat of an
other words, it may be enough if I can produce any reason in the infinite regression, the regressions are never given a detailed
sequence, even though I cannot produce every reason in the se description - rather, Sextus merely gestures, in abstract terms, at
quence. And, for all that has thus far been said, proposition (2a) the possibility, the theoretical possibility, of such a regression. And
may well, in some cases at least, be true. In other words, it may, in it is clear why this is so: in most of the cases which Sextus has in
some cases, be possible for me to produce any reason in the infinite mind, there is no readily imaginable way of actually constructing
sequence. The objection to infinite regression rests on the denial of the regression, because no appropriate algorithm can be conjured
(1a). But this foundation is insecure. For the objection to succeed, it up.
50
Thus even if the Dogmatist is correct, as I have thus far argued, in (R2) (i) If it is true that if if P2, then P 1, and P2, then P 1, then the
holding that proposition (2a) is sufficient to ground the possibility argument 'If P2 , then P1; and P2: hence Pi' is acceptable. (ii)
of infinite epistemological regressions, he has gained very little. For But it is true that if if P2, then P 1, and P2, then P 1• Hence (iii) the
unless he can also construct or find some appropriate algorithm, he argument 'If P2, then P1; and P2: hence P1' is acceptable.
cannot actually produce any infinite sequence which will do him
epistemological service. The theoretical possibility of an accept Now if the sceptic asks his opponent to warrant this argument, it is
able regression is just that - an empty theoretical possibility. evident that he can do so by reapplying the same procedure to
None the less, regressions can in fact be constructed (at the price produce a yet more complex argument, R3• And it is equally evident
of some artificiality); and even if they are unlikely to be of any that the procedure will generate a sequence of infinitely many
serious epistemological importance, they may still be interesting as arguments. Each argument will be more complex than its prede
theoretical possibilities. I offer two such sequences - indefinitely cessor. The second argument in the sequence - the one I have
many others can readily be constructed on their model. formulated - is not altogether easy to comprehend; and it would in
The first example is this. A mathematician claims that the practice be difficult and tedious to formulate even the third argu
number two is even. He supports this by arguing as follows: ment in the sequence. But the matter is in principle perfectly easy,
Four is even; and if four is even, two is even: hence two is even. and we can readily 'see' how the sequence is constructed.9
In each of these cases we have an infinite sequence of reasons or
He supports the first premiss of this argument thus: arguments; and the sequences can, in the defined sense, be mas
Six is even, and if six is even, four is even: hence four is even. tered. No doubt the particular sequences I have constructed would
not in fact be offered as proofs; and no doubt there are few if any
And so on. Evidently, an infinite sequence of arguments is thus such sequences which ever would be offered as proofs. But never
generated. Evidently, proposition (2a) holds of the sequence. For theless, given that such sequences can, sometimes, be constructed,
there is a simple algorithm for constructing any argument from its may they not in principle be offered as proofs? Have we not secured
predecessor. {Suppose that one of the arguments in the sequence is the possibility of an infinite epistemological sequence against
'n is even; and if n is even, then n - 2 is even; hence n - 2 is even'. which no objection can be made?
Then the next argument in the sequence will of course be: 'n + 2 is
even; and if n + 2 is even, then n is even; hence n is even'.) Thus for 9 Each argument is in modus ponens form. Suppose that argument Ri
any argument R i in the infinite sequence, we can produce the is:
supporting argument R i + 1• If A, then B; and A: hence B.
The second example is slightly different. It is an attempt to meet Then argument R;+, will be:
the challenge issued by Sextus at PH I 122, where the mode of If if if A, then B, and A, then B, then Ri is acceptable; but if if A,
infinite regression is first thrown down. A philosopher claims that then B, and A, then B: hence R; is acceptable.
P, (the content of the claim is immaterial). He supports the claim by The sequences of ifs make this hard to grasp. A semi-symbolized
version may make it a little easier. (The arrow stands for 'if . . . ,
a modus ponens argument: then . . . ', and the inverted V for ' . . . and . . .' Thus R;+1 will be:
(((A-.B) /\ A)-.B)--t(R; is acceptable)
(((A�B) /\ A)�B)
Hence Ri is acceptable.
And the sceptic challenges him to show that this argument is
trustworthy. He replies by producing the following more complex (The example is based on a similar infinite sequence generated by
Lewis Carroll in his classic paper, 'What the Tortoise said to
argument: Achilles', Mind 4, 1895, 278-80.)
52 53
r
1
The example can be generalized. For any infinite sequence L
You do not need to be a Pyrrhonist to be sceptical of all this: you which we can construct, we may generate a rival sequence Lrc with
may well feel there is something very fishy about the claim I have the following two properties: first, L•:· is precisely analogous in
made for these infinite sequences. Even if it is admitted that such form and structure to L; secondly, the claim purportedly warranted
sequences will not, in practice, make any serious contribution to by :i:"' is inconsistent with the claim purportedly warranted by L. By
epistemology, you may still wish to urge that there is something virtue of the first property, either L and P are both acceptable or
more to be said- that they cannot, in principle, make any contribu they are both unacceptable. By virtue of the second property, they
tion at all. Can we show that this is right? And if so, can we explain are not both acceptable. Hence neither is acceptable. Hence L is not
why? I end with some tentative thoughts on these questions. acceptable. 1 0
Look again at the first infinite sequence, to the conclusion that Does this show that no infinite regressions can have any
the number two is even. Call this conclusion P. The first argument epistemological power? lt is tempting to think so. But the argument
was: is certainly less than probative. I mention two difficulties. First, is it
If four is even, then P; and four is even: hence P. true that an appropriate L'' can be constructed for any offered L? I
do not know: the claim seems fairly plausible (at least, it is hard to
Call this argument R1• We know how to construct R, and R, and R, see how one could construct a L for which one could not find a L':· );
. . . Thus we can marshal! the following infinite sequence of but I cannot prove that it is true. It is at best a tempting conjecture.
arguments in support of P: Secondly, even if an appropriate Lr, is found, is the formal
similarity between L and L'' enough to warrant the epistemological
(Ri, R2, • • • , Rn, . . . )
conclusion? Granted that L and :i:» do not differ formally, are we
Now imagine, in Pyrrhonian vein, a rival mathematician, locked in then justified in holding that the one is epistemologically accept
51o:q>wvia with the first. The rival strenuously maintains that the able if and only if the other is? Perhaps the champion of infinite
number two is odd, not even. Call this claim P*. The rival's first regressions will insist that L and L* will, or must, differ in some
argument for P'' is, of course, this: non-formal way, and that this difference warrants his accepting L
If four is odd, then P''; and four is odd: hence P�'. and rejecting L"'. And of course there are non-formal differences
between L and L''. In the particular case we are examining, there is
Call this argument R\. Plainly, the rival will be able to produce R *, an obvious and obviously pertinent non-formal difference between
and R* 3 and Rrc 4 Thus he can generate an infinite sequence of
• • •
the two sequences. The elements in the P�sequence are all true, and
arguments in support of P'', viz.: the elements in the p•:--sequence are some of them false. It is true
that if four is even, then two is even; and it is true that four is even. It
is (let us grant) true that if four is odd, then two is odd; but it is not
His sequence, the P* -sequence, has exactly the same formal proper true that four is odd. Hence despite their formal symmetry, the P
ties as the P-sequence; and it is constructed and mastered in exactly sequence and the P* -sequence are not on an epistemological par
the same way. with one another.
Since the two sequences are exactly analogous to one another, 'Ah', the sceptic will quickly reply, 'but we are trying to prove
then by 'parity of reason' the one will be epistemologically the truth of the elements in L and L':-. You cannot prefer one
acceptable if and only if the other is. But P and p•> are incompatible sequence to the other on the grounds that its elements are all true -
with one another. Hence both sequences cannot be acceptable.
Hence neither sequence is acceptable. Hence the P-sequence is not 10 The general idea behind this line of argument, which the ancients
acceptable. would have called an 6:VTlTrapaf3oi'-.fj, was given to me by Stephen
Everson.
54 55
that is to beg the question.' This is a neat rejoinder. But it is not item in the sequence takes us out of the realm of belief and into the
plain that it is a just rejoinder. For after all, the proponent of the P realm of fact.)
sequence is right and the proponent of the P*-sequence is wrong. I have often heard thoughts of this sort expressed. Like others, I
And even if that does not - and cannot - give the proponent of the find in them a Siren seductiveness. I am prepared to countenance
P-sequence a reason to prefer his sequence, it surely shows that he is the possibility that they voice the song which Sextus himself was
right to prefer his sequence. trying to sing. I am even inclined to believe that there may be a truth
It is worth seeing precisely what is at issue here. The sceptic in towards which they are enticing us. But they do not succeed in
effect urges that voicing that truth themselves. For it is in fact easy to see how a
You may properly claim to know P on the basis of an infinite
Dogmatic proponent of infinite regressions will reply. 'Links with
sequence of arguments L only if you have good reason to the world?' he will say, 'Of course beliefs must be linked with the
prefer L to any rival Lr, world: who�ver could deny that? But why on earth do you think
that my infinite sequences have no worldly ties? Nothing could be
-and given the formal identity between L and L'�, you can never have further from the truth: P is itself linked to the world, i.e. it is true.
good reason for a preference. The Dogmatist suggests rather that Moreover, its link to the world is revealed and ensured by R,. As for
You may properly claim to know P on the basis of an infinite R1, that too is linked to the world - for its premisses are true. And
sequence of arguments L provided that the arguments in L its worldly links are revealed and ensured by R,. And so on. My
are in fact good arguments; you do not need in addition to sequence has no links with reality, do you say? Nonsense, it has
know that they are good arguments or to know that L is infinitely many worldly links.'
superior to any rival L'� The reply is just. To be 'linked to the world' is simply to be true
- and hence the formal identity between the two sequences does not of the world, i.e. to be true. Being linked to the world is not an
matter. Who is right here, the sceptic or the Dogmatist? The alternative to being linked to other beliefs, and our Dogmatist is
question is perplexing. It is similar to a question which arose at the not obliged to choose between two bollards round which to tie his
end of the previous chapter and which I there postponed. The stern-rope. He may use both bollards. He may link his beliefs both
present question I also postpone to the final chapter. to the world and to other beliefs.
There is room for a few more dialectical manoeuvres, room for a
little more squirming. But I cannot see that in the end any remarks
Some of you, I suspect, have been losing patience with the last part
about 'links with the world' are likely to embarrass the infinite
of my discussion. Surely, you may be thinking, it is utterly plain
regressionist.
both that and why infinite sequences are epistemologically absurd
And on that note I leave infinite regression.
- they are absurd because they do not link our beliefs to reality.
They embark on an endless succession of claims or beliefs, each of
which ties down an earlier claim or belief, and each of which is tied
down by a later claim or belief. But in order to prove or justify or
warrant a belief or claim, it is not enough to tie it to other beliefs or
claims. You must anchor it to the world, to reality. Beliefs must be
linked or nailed or glued to what is in fact the case. (And then one
might wonder if this is not really the very point Sextus is trying to
make when he objects that infinite sequences give you no Cxpxri or
starting-point. There is no O:pxf} to the sequence inasmuch as no
56 57
:'j
Reciprocity
3 convertibility; and a little later, showing the convertibility of
I-propositions, he uses the convertibility of E-propositions.
And this mode of proof is agreed to be unsound (µoxSf)pos).
Reciprocity (in APr 3 1 .27-32.3; cf. Philoponus, in APr 49.11-14)
Thus Aristotle, too, allegedly argued from P, to P, and then back

again from P, to P,; and he allegedly thought that he had thereby
proved both P, and P,.
In the two texts I have cited, we see Dogmatists snapping at
Dogmatists. The charge of reciprocal reasoning is meant to show
that the rival Dogmatist is mistaken, that his views are at best
The Stoics maintained that everything is fated, and they main
ungrounded. The sceptics, too, bring the charge of reciprocity
tained that the art of divination exists. According to Diogenianus
against the Dogmatists. But they use the charge for specifically
(an Epicurean critic of Stoicism),
sceptical ends: reciprocity induces ETroxft, suspension of judge
Chrysippus has offered us a proof in which he tries to ment. Sextus refers explicitly some thirty-five times to the
establish these contentions reciprocally (51' 6:i\i\i]Awv). For he 'reciprocal mode of suspension (6 51aMriflos Tponos Ti']s hroxfi s) ' ,
wants to show that everything happens in accordance with the method of inducing scepticism by pointing to the impotence of
fate from the fact that divination exists; and he could not reciprocal argument. Like Diogenianus, he characterizes recipro
prove that divination exists unless he presupposed that
cal proof as 'most unsound (µox6rip6ToTos)' (M m 99); and the
everything occurs in accordance with fate. And what mode of
reciprocal mode is correspondingly described as 'most perplexing
proof could be more unsound {µoxerip6Tepos) than this?
(cmop&:nOTos)' (M VIII 445; IX 47; cf. Vlll 379).
(frag. 4 Gercke Eusebius, PE IV iii 2-3)
=
The reciprocal mode is the fifth of the Five Modes of Agrippa.

Thus Chrysippus allegedly argued from P, to P,, and then back Here is Sextus' formal account of it.
again from P, to P,; and he allegedly offered these arguments as
The reciprocal mode comes about when what ought to be
proofs of P 1 and P2• But his reasoning, according to the Epicurean
confirmatory of the matter in question requires warranty
opponent, is reciprocal and therefore unsound.
from the matter in question. Thus being unable to assume
Some critics accused Aristotle of committing a similar error in a either matter for the establishment of the other, we suspend
passage in the Prior Analytics. At APr 25ar4-r7 he offers a concise judgement about both. (PH I I69)
proof of the convertibility of £-propositions, a proof, in other
words, that fro1n 'No A is B' you may infer 'No B is A'. Then, at Diogenes Laertius gives the same description of the mode, and he
25a20-2, he gives an argument for the convertibility of ! adds an illustrative example:
propositions (from 'Some A is B' you may infer 'Some B is A')
E.g. if someoll'e wants to confirm that there are pores by the
which invokes the convertibility of £-propositions. Alexander of fact that there are effluences, and assumes this very fact to
Aphrodisias reports the following criticism. confirm that there are effluences. (ix 89)
Some people think that in the proof <of E-convertibility) he A Sextan example: in the course of his attack on Aristotelian logic,
assumes that I-propositions are convertible, and they accuse
Sextus considers the following sample syllogism:
him of using reciprocal proof (T1 Ot' 6:i\Afii\wv Oei�1s). For,
wanting to show that E··propositions are convertible, he uses Socrates is a man
for this proof - so they say - I-propositions and their No man is a quadruped
Therefore, Socrates is not a quadruped
59
Reciprocity Reciprocity
He comments on it as follows. Thus far I have talked only of pairs of arguments; and I shall in
They want to confirm the premiss 'No man is a quadruped'
future restrict the term 'reciprocal' to cases in which exactly two
by induction from the particulars; and they wish to infer each arguments are involved. But it is plain that reciprocal arguments
of the particular propositions from 'No man is a quadruped'. are a special case of a more general type of argumentation. For
Thus they fall into the perplexity of the reciprocal mode. there is no need to restrict attention to pairs. We might think of
(PH II 197) three arguments, standing in the following relation to one another:
'P3: so P2'; 'P2: so Pi'; 'P1: so P3'. Here there is what we might call
Thus syllogisms - of this particular sort - cannot provide proofs. circularity -you start with P3 and end up at P3 again. And the circles
For they argue from a universal to a particular; and yet they must may be as large as you like. They may have four or five or six
also invoke an argument from particular to universal. (This elements - or n elements for any n you like.
passage in Sextus is the origin of the notorious charge that Since the arguments in the circle may each have several
syllogisms 'beg the question' if they are employed as proofs. A premisses, our account of circularity must accommodate the fact.
syllogism cannot establish that Socrates is not a quadruped; for we We may have, say, 'A: so P2', 'B: so P 1 ' 'C: so P3', where P2 is
must already know that Socrates is not a quadruped in order to included in B, and P 1 is included in C, and P, is included in A. More
avail ourselves of the premiss that no man is a quadruped. I shall rigorously: Let TI1, n2, • • , Tin, be sets of propositions. Then the
•
not comment on the merits of this argument.) sequence of arguments 'TI 1 : so P1', 'n2: so P/, . . . , 'Tin: so Pn' is a
Sextus' example shows that we must interpret the notion of circular argumentation just in case every P1 is included in ni + 1 (and
reciprocal proof a little more broadly than the other texts I have Pn is included in TI1 ) . More informally: you argue in a circle when
cited may have suggested. From those texts we might have inferred you produce a sequence of arguments such that the conclusion of
that reciprocity occurs when we are offered a pair of arguments one argument is a premiss for the next argument, and the
('P 2: so P1 '; 'P 1 : so P /) where the premiss of the first argument is the conclusion of the last argument is a premiss for the first argument.
conclusion of the second argument, and vice versa. But in the It is plain that reciprocity is a special case of circularity:
Sextan example both arguments have several premisses. The first reciprocal arguments are those circular arguments which have just
argument is a syllogism with two premisses ('No men are two members. All reciprocal arguments are circular arguments,
quadrupeds' and 'Socrates is a man'); the second argument is an but not all circular arguments are reciprocal arguments.
induction with numerous premisses (of the form 'Socrates is a man I shall suppose that the 'reciprocity mode' is essentially an
and not a quadruped', 'Plato is a man and not a quadruped', attack, not just upon reciprocal argument, but more generally
'Aristotle is a man and not a quadruped' . . . ). So we cannot talk of upon circular argument. And I shall eventually discuss the
the premiss of the argument. credentials of the mode in the light of this supposition. But the
Rather, we should explain reciprocity as follows. Reciprocity supposition, though philosophically necessary (as I shall explain),
occurs when we are offered a pair of arguments ' A: so P 1' and 'B: so is historically false 7 or over-simplified. So before I examine the
P2 ', where P2 is one of the premisses collected in A and P 1 is one of mode as a general sceptical device, I shall say something about the
the premisses collected in B. More rigorously: let n1 and TI2 be sets way in which it actually functions in the writings of Sextus.
of propositions. Then the arguments ·n 1: so P1' and 'TI2 : so P2 ' form
a reciprocal pair just in case P2 is a member ofTI1 and P1 is a member
of n,. Roughly: you argue reciprocally when you use the Sextus deploys the reciprocal mode, as I said, some thirty�five
conclusion of one argument as a premiss in a second argument times. Several of these examples are- or seem to be - similar to the
which itself is supposed to establish one of the premisses of the first syllogistic example at PH II 1 97; and thus they are genuine
argument. illustrations of the type of argumentation which I have defined as
60
reciprocal argument. 1 But more than half of Sextus' arguments are to which this example belongs might run like this: we have generic
not of this type at all. Among these deviant cases, two groups may reciprocity when we have an infinite sequence of claims, the even
be distinguished. members of which fall into one epistemologically suspect class, and
An example of the first group is provided by the very first the odd members of which fall into another epistemologically
reference to the reciprocal mode in PH (the context is immaterial): suspect class. At PH I 116, the even members are of the form: 'Ai is a
If he judges the matter, he will of course say that he has judged proof that K; is reliable' - and all proofs are suspect and in need of
it by a criterion - and for this criterion we shall seek a proof, support. The odd members are of the form 'K; is a criterion which
and for the proof a criterion. For the proof will always require ensures the soundness of Ai - 1' - and all criteria are suspect and in
a criterion in order to be confirmed, and the criterion a proof need of support.
in order to be shown to be true . . . And thus criterion and In many places where Sextus appeals to the reciprocal mode, it is
proof fall into the reciprocal mode. (PH 1 116-17). an argument of this sort which he has in mind.' I shall say nothing
I am concerned - as usual - not with the soundness of Sextus' here about the credentials of generic reciprocity; for, as far as I can
argument but with its form. And it is plain that it does not have the see, it is properly construed as a special case of infinite regression,
same form as those arguments which I have defined as reciprocal. and I have already discussed infinite regressions.
' As an illustration of the second group of deviant cases, take a
For it does not consist of a pair of arguments, 'n1: so P1 , ' n2: so P2',
passage from Sextus' discussion of the concept of cause:
with P1 in n1 and P2 in n1•
Rather, we seem to be faced with an infinite regression of a Now in order to conceive of a cause we must first recognize
special kind. The Dogmatist under attack is supposed to claim that the effect, and in order to recognize the effect, as I said, we
K1 is his criterion for judging that P1• He then produces a putative must first know the cause. Then the reciprocal mode of
proof, A1, for the reliability of K1; and next, a criterion, K2, for the perplexity shows that both are inconceivable -we can form a
adequacy of A , ; and so on. Thus: conception neither of the cause as cause nor of the effect as
effect. {PH 111 22)
K1: P1 is true
Ai: K1 is reliable Here, too, there is no reciprocal argumentation of the sort
K2: Ai is sound defined. Indeed, there is no argument at all. Rather, Sextus is
A2: K2 is reliable concerned with reciprocal conceptions - or, better, with reciprocal
K3: A2 is sound definitions. For the Dogmatist is supposed to explain the concept
of cause by saying
{C) A cause is what produces an effect,
There is no genuine reciprocity involved, since no proposition is and to explain the concept of effect by saying
ever repeated: the Kis are all different criteria, the Ais all different .
{E) An effect is what is produced by a cause.
proofs.
We can, of course, see why Sextus invoked the notion of Now, so Sextus argues, (C) presupposes an understanding of the
reciprocity here; for we do in fact oscillate from criteria to proofs concept of effect, i.e. it presupposes (£). But (E) presupposes an
and back again. There is, if you like, a generic reciprocity. But there understanding of the concept of cause, i.e. it presupposes (C). 3
is no individual reciprocity. A rough characterization of the group
2 See PH I 172, 176, 186; II 68, 92/3, 183 (bis); m 35, 53; M VII 341; VIII
r See PH 11 9, 36(?), 114(?), 196, 199, 202; M v111 261(?), 342(?), 379/80; :Z.2, 29, l:Z.:Z., 181, 445(?).
Xl 183. (A question-mark indicates that it is not dear whether the On this see Jonathan Barnes, 'Ancient Skepticism and Causation', in
case is of this type or not.) M.F. Burnyear (ed.), The Skeptical Tradition (Berkeley, 1983).
There are several examples of this sort of thing in Sextus.' I shall Aristotle discusses circular proof in the chapter of the Posterior
comment briefly on their substance later. Here I want only to insist Analytics to which I referred in the previous chapter. He shows
that such examples are not pieces of reciprocal argument - for they himself explicitly aware of the fact that circular reasoning may
are not pieces of argument at all. Again, we can see why Sextus contain any number of component arguments. But he chooses to
should invoke the notion of reciprocity. But he cannot be invoking discuss only the simplest case, i.e. the reciprocal case.
the particular notion of reciprocal argument.
For it makes no difference whether they say that it bends back
Sextus thus calls upon the reciprocal mode in three distinct
through many items or through few, nor whether through
contexts: first, to deal with genuinely reciprocal arguments; sec
few or through two. (APst 72b36-,)
ondly, to deal with generically reciprocating regressions; thirdly, to
deal with reciprocal definitions. (There are some examples which it Large circles or small, many items or few - it is all the same. The
is hard to categorize with any confidence; and a few texts seem to objection to circularity does not depend on the magnitude of the
convey no more than the vague thought that a Dogmatist is circle. I suppose that this Aristotelian thought lies behind Sextus'
somehow obliged by his theory both to do A before he does B and procedure. (Later I shall suggest that the Aristotelian chapter may
also to do B before he does A.) Sextus never explicitly marks any actually have inspired Agrippa to collect his modes.) In Sextus' text
difference among the three sorts of case; and in a loose sense he can we find only reciprocal arguments. But the explanation is, as it
reasonably say that all of them involve reciprocity - in each type of were, pedagogical: reciprocal arguments are the simplest sort of
case, the Dogmatist goes from A to B and back to A again. But the circular reasoning, and since no extra issues are raised by larger
cases are distinct, and they need to be distinguished. Indeed, they circles, the exposition may properly restrict itself to reciprocity.
need to be distinguished by the Pyrrhonist; for they raise quite Whether or not this is true - whether or not large circles are no
different problems. different, in any epistemologically important features, from small
There is another curious fact about Sextus' use of the reciprocal circles - is an issue to which I shall return.
mode. All the cases he considers are cases of reciprocity. All, that is
to say, involve pairs of arguments (or of generic types or of In the last chapter, I asked why infinite regression should be
definitions). He nowhere examines or criticizes a circular argu thought to lead to Eiroxrl. The same question arises in the case of
ment with more than two elements, a circular argument which is reciprocity. Since the answer in this case is parallel to the answer in
not also reciprocal. How is this fact to be explained? We shall the regressive case, I shall be brief.
hardly entertain the thought that Sextus might have had no objec Clearly the fact - if it is a fact - that Chrysippus argued
tion to large circles, holding that reciprocal argument was the only reciprocally for the existence of fate is no reason at all for us to
unsound form of circular argument. Indeed, we must surely sup suspend judgement over the existence of fate. For even if reciprocal
pose that his objections to reciprocity were intended to hold arguments are bad arguments, the fact that Chrysippus argued
generally against all circular reasoning, for otherwise he would badly for the existence of fate does not imply that there are no good
have left a loophole for Dogmatic belief. (I mean that the Dogma reasons for admitting (or for denying) the existence of fate.
tists could produce circles containing three or more elements, and Again, we must understand a qualifying phrase: 'as far as this
thus escape Sextus' objections to reciprocity. And that would be argument goes', Ocrov Eir\ TOVT(}l. As far as Chrysippus' argument
ridiculous.) goes, we must suspend judgement. If the only reason we have for
accepting or rejecting the existence of fate is supplied by
Chrysippus' reasoning, then we should suspend judgement. And
4 See PH Ill 242; M VII 4i.6; VIII 86; IX 47; M Ill 99. In the first three of
these passages Sextus' example is the same - an alleged reciprocity in why? Because reciprocal arguments are bad arguments; and if
the Stoic definition of qiavTacr!a KCTTai\tyrrT1Kt']. the only reason we have for accepting or rejecting P is a bad
argument, then we should neither accept nor reject P but suspend the genus, we referred to the species, and in describing the
judgement. species we referred to the genus. We reply that reciprocal
It might be said, and truly, that the Oaov E.rri 10VT� qualification proof is not always to be avoided. For if the objects with
leaves the reciprocal mode with little to do; for surely it will hardly which the accounts (A6yo1) are concerned do not depend for
ever be the case that the only reason we have for accepting or their subsistence upon one another (like cows and horses and
rejecting P is a reciprocal argument. (The same, as I observed, holds so on), then we should avoid reciprocal proof with all our
of the regressive mode: it will hardly ever be true that the only power. But if the objects do depend for their existence upon
reason we have for accepting or rejecting P is an infinite sequence of one another, then it is not possible to give an account of them
except by reciprocal proof. For just as it is impossible for a
arguments.) This is surely true; but it does not mean that the
shadow to be still while the object which casts it is not still, so
reciprocal mode lacks real interest. For if the mode will rarely, if
it is impossible for the accounts not to require one another
ever, be useful by itself, it may still have power as a member of a set when the objects with which the accounts are concerned do
or team of modes. And in my last chapter I shall try to show that as require one another. (in Porph isag 62.9-19)5
a team-member it is powerful, even if as an individual performer it
lacks strength. Since correlatives are indissolubly linked to one another, their
definitions or A6yo1 may - indeed must - also be interlinked or
Now the chief question is this: What, if anything, is actually wrong reciprocal.
with circular reasoning? But before I turn to this question, it may be What exactly is Elias defending? Although he talks of 'reciprocal
diverting to record that some ancient thinkers actually defended proof', he is plainly not defending what I have called genuine
circular argument. Three texts, or groups of texts, are relevant. reciprocity. Elsewhere, indeed, he insists that reciprocal argument
The first group clusters about Porphyry's lsagoge or Introduc must be avoided in all arro5Ei�EIS or proofs (proleg 9.23-5) . Nor is
tion to Philosophy. In a notorious passage, Porphyry explains that he defending generic reciprocity; for the text of Porphyry on which
a genus is something which stands over species and that a species is he is commenting does not contain a generically reciprocal or
something which is included in a genus (isag 4·4-9); and 'in saying regressive argument - indeed, it does not contain any argument at
this', as his commentator Ammonius remarks, 'Porphyry is rightly all. Despite his reference to 'reciprocal proof', it is clear that Elias is
thought to have used reciprocal proof' (in Porph isag 74.8-rn). actually concerned with reciprocal definition. (In fact, he distin
After a brief discussion, Ammonius concludes that Porphyry's guishes, in a thoroughly traditional fashion, between Optcrµoi or
procedure is in fact unobjectionable; for genus and species are 'real' definitions and VTioypaq>ai or outline descriptions; and he
correlatives, and holds that reciprocity is never legitimate for 6p1crµoi but only for
vrroypmpai (in Porph isag 58.20-8) . But the distinction need not
relatives (TO: Tip6s TI) subsist at the same time and are thought
trouble us here, and I shall continue to speak simply of definitions.)
of at the same time as one another . . . Thus it is absolutely
In sum, Elias holds that when you are defining relative terms,
necessary that anyone who is expounding one of them should
then your definitions may - or rather, must - be reciprocal. This
refer to the other; for if you are ignorant of one relative, you
will not know the other either. (in Porph isag 76.5-Io) may seem a modest enough position. Even so, if Elias is right, some
of Sextus' applications of the reciprocal mode will be in error. In
No doubt Ammonius is right; but his remarks seem a little ad hoe. particular, cause and effect are correlatives, in Elias' sense; hence,
A later commentator, Elias, is more satisfactory; and I quote an according to Elias, reciprocal definition is permissible - indeed
extract fro1n his discussion of the same passage in Porphyry: obligatory - in their case.
But, they say, we have fallen unawares into reciprocal proof,
See also Olympiodorus, in Cat 109.21-8; David, in Porph isag
a thing which philosophers should avoid. For in describing 132.6-25, 14+26-31.
66
But is Elias' modest suggestion right? I do not think so. First, it is of one another; rather, they define one and the same notion, the
perfectly possible to define causes without referring to effects, and notion of the causal relation.
vice versa. Sextus offered us the formulae: Thus when Elias says that, in the case of correlatives, reciprocal
(C) A cause is what produces an effect, definitions are unavoidable, he is wrong - or at best he is speaking
in a highly misleading way. And the best defence of Porphyry's
and controversial procedure over genus and species will not invoke the
(E) An effect is what is produced by a cause. sort of reciprocity to which Sextus objects.
The Porphyrian case apart, we might still wonder if reciprocal
We need only make a trivial modification to these formulae to get: definitions are ever permissible. The answer to this question de
pends on what we expect from a definition. If the function of a
(C*) A thing is a cause if it brings something about,
definition is to introduce or explain a new term by way of terms
and already understood, then reciprocity must in all cases be avoided.
For suppose that (C) - 'A cause is what produces an effect' - defines
(E�') A thing is an effect if it is brought about by something.
the 'new' term 'cause' by means of words already understood.
Here (C*) does not use the word 'effect' and (E') does not use the Then (E) - 'An effect is what is produced by a cause' - is at best
word 'cause', and so there is no reciprocity. But-and secondly - no superfluous and at worst misconceived. For 'effect', the term which
doubt (C*) and (E') seem far too thin and weak to function as (E) purportedly defines or introduces for the first time, is - by
definitions; and no doubt, too, what we really want is a definition hypothesis - a term already known and understood.
or an explanation of the causal relation, i.e. we want an analytical Now though definitions sometimes serve to introduce and estab
definition of the form lish new terms, they do-not always function in this fashion. And so
the argument I have just sketched does not show that reciprocal
(CR) x causes y �
df P
definitions are always improper. But l shall not pursue this issue
Now if we produce something of the form (CR), must we thereby any further, since it is tangential to the main sceptical circle.
refer to effects? Clearly not. Various thinkers have made various
attempts to produce satisfactory definitions of causation in the So much for Porphyry and his commentators. My second group of
form of (CR). None of them does, and none of them need, make reciprocal defenders clusters about the rhetorician Hermogenes,
explicit reference to effects. who was a contemporary of Galen. There is a connexion between
Yet there is a little more to be said. Plainly, x causes y if and only this group and the Porphyrian group. In one of his works, Hermo
if y is caused by x, i.e. if and only if y is an effect of x. Thus, given genes appears to indulge in something like reciprocal definition; for
(CR), we can at once produce he says that 'pretty well all stylistic features are discovered and
become plain reciprocally' (id I iii [ Rhetores Graeci n 279
(ER) y is effected by x� df P
=
Spengel]) . Later commentators defended him: they insisted that

So (ER) is, in a sense, indissolubly linked with (CR). You cannot reciprocal definition is permissible in the case of relatives - and one
define causes without implicitly defining effects, and you cannot of them expressly appealed to the example of Porphyry on genus
define effects without implicitly defining causes. If Elias was grop and species. 6
ing for this point, then it was a truth he was groping for. But the But there is a second and more interesting reference to recip-
truth has nothing to do with reciprocal definition, as I have
explained the notion, and we should not talk of reciprocity here at 6 John Doxapatres refers to Porphyry: sec his scholia to Hermogenes,
in Rhetores Graeci [Rhet Gr] VI 160---1 Walz; cf. Maximus Planudes,
all. For (ER) and (CR) do not define two different notions in terms in Rhet Gr v 470 Walz; anon., in Rhet Gr v11.2 928n69 Walz.
68
rocity in Hermogenes. In the account of 'conjectural inference' of what is unclear and in need of judgement depends on some
or crToxacrµ6s in his book on Issues, Hermogenes distinguishes evident sign; and the evident signs here are the suspicion of
a form which he calls 'co-established inference (cnoxcxcrµos adultery - not the adultery itself - and the freeing of the
crvyKaTacrKeva�6µevos) ' . He explains it thus: servant. It is through these that it is established that the
master was killed by the servant, and by way of the murder we
A co-established inference is made when the signs of the establish not the suspicion of adultery (there is no need to
matter are established reciprocally (Cit' Cx'AAT)i\wv). E.g. the establish what is evident) but the adultery itself. Hence there
prisoners had to be freed at the festival of Thesmophoria. A is no genuine case of reciprocal proof.
man suspected his servant of adultery with his wife. He (Maxim us Planudes, in Rhet Gr v 297 Walz) 7
imprisoned the servant and left town. His wife freed the man
at the Thesmophoria while her husband was still away. The Thus Planudes maintains that reciprocal proof - genuine recipro
servant ran off. The husband returned and was found mur cal proo f - is contrary to reason; and he supposes that Hermogenes
dered. Then the wife was accused of conspiracy. Now the fact has given a faulty analysis of the imaginary court case which he
that the husband was killed by the servant is established by describes.
way of the adultery, and the fact of the adultery is established Planudes does not explain why reciprocal proof is wrong any
by way of the killing. And in general, they are established more than Hermogenes indicates why he sees no objection to it. But
reciprocally- - but together with other facts, of course (the he does give an illustration which invites a brief digression.
man's suspicion, the wife's freeing the servant, and so on).
(stas 111 [ Rhet Gr II r 52 Spengel])
= Suppose that you do not know where Dio or Theo lives but
that you do know where Plato lives. Someone says to you:
This is evidently an example of genuine reciprocity. We are invited 'Dia lives where Theo lives, and Theo lives where Dia lives.'
to conclude, first, from the evidence of the adultery that the servant Now if he goes no further and stops there, the proof is
killed his master, and secondly, from the evidence of the murder reciprocal. But if he adds that Dio or Theo lives where Plato
that the slave committed adultery. (Thus we have two signs or lives, then there is a clear proof of the statement.
pieces of evidence that the wife is guilty of conspiracy, and each (Maximus Planudes, in Rhet Gr v 295--6 Walz)
sign is established through the other.) Hermogenes expressly notes,
The example is found elsewhere.' The dummy names 'Theo' and
in the last sentence of the extract, that other premisses are involved.
'Dia' suggest a Stoic origin; and an anonymous text apparently
In short, we have, according to Hermogenes, a pair of arguments of
confirms the suggestion.
the form
Ai, Al, . . ., An, P2: so P1 How will an enquiry of this sort come to an end if the
Bi. Bi, . . . , Bm, P1: so Pi elements of each proof are reciprocally generated? For such
things will remain controverted. Thus there is an argument
and that is the very form of a genuinely reciprocal argumentation. which the Stoic� call reciprocal, and which is 6:vaTI65eJKTO).
Hermogenes appears to assume that there is nothing wrong with E.g. Dio lives where Theo lives; and Theo lives where Dia
these 'co-established inferences'. His commentators were embar lives. Such an argument, which, as I say, offers a reciprocal
rassed: the passage caused much discussion and controversy, and
in the end Hermogenes found no defenders. I cite one late text. 7 See also Sopater, in Rhet Gr 1v 460-1 Walz; Marce!linus, ibid. 461-3;
anon., ibid. 463; Epiphanius, ibid. 465; anon., Rhet Gr v11.1 381--9
Here Hermogenes says that unclear items - the murder and Walz. (The last text is the most elaborate; but it adds little of
the adultery - are proved reciprocally. He is wrong. For this substance to the slimmer commentaries.)
sort of proof is contrary to reason. In every inference, the test 8 See anon., in Rhet Gr vn.2 928n69 Walz; Elias, proleg 9.12-21.
70
proof,? is 6.va-rr65e1KTOS and never comes to a conclusion. And which Stoics were interested in the thing. We do not know what or
the same occurs, they say, with co-established inferences. how much they said about it. We cannot even be sure exactly what
(anon., in Rhet Gr VII.I 383 Walz) sort of argument they took to be reciprocal. (For our texts present
This passage undeniably shows that some Stoics had spoken of the example of Dio and Theo in telegrammatic form, and several
reciprocal argument. But the text as a whole has been different interpretations can be ventured.)
misunderstood.
It has been inferred that the Stoics reflected on the rhetoricians' After this journey through relatively remote texts, I return for my
idea of co-established inference, that they defended a variety of final group of defenders of reciprocity to a familiar author and a
reciprocal proof (holding it to be 6:vaTI65etKTOS or indemon familiar work. In the Posterior Analytics Aristotle observes that
strable), and that the sceptical mode of reciprocity was directed some hold that knowledge can only come from proof. But
against - or influenced by - this Stoic view. Were these inferences they maintain that nothing prevents there from being a proof
warranted, we should have in the anonymous scholiast on Her of everything - for it is possible for proof to be circular and
mogenes an important and intriguing witness to the history of reciprocal. (APst 72bI5-18)
reciprocal argument. Aristotle adverts to defenders of genuinely reciprocal argumenta
But the inferences are not warranted. First, there is no evidence tion and we must believe that in Aristotle's day some thinkers had
that the Stoics knew the rhetoricians' co-established inferences: the exp :essly allowed the legitimacy of circular proof. But although
'they' in the last sentence of my extract refers not to the Stoics but to Aristotle argues against their view at some length (APst 72b25-
the earlier critics of Hermogenes whom our anonymous scholiast is 73a20), he does not tell us who his opponents were; he does not
reviewing. The scholiast does not state or imply that any Stoics explain under what conditions they accepted circularity; and he
actually referred to co-established inferences.
does not indicate how - if at all - they argued in defence of the sorts
Secondly, there is no evidence that the Stoics defended what they of proofs they accepted. Aristotle's reticence is frustrating. Schol
called reciprocal arguments. In Stoic terminology, the word
arly conjectures on the matter have taught us little. 10
6:vaTI65e1KTOS does indeed mean 'indemonstrable', and it is applied
to fundamental - and fundamentally valid - argument-forms. But Thus little, alas, emerges from the meagre texts in defence of
our scholiast uses the word in a different sense: he regularly uses
reciprocity. A few ancient thinkers did indeed accept reciprocal or
6:vaTI65etKTOS in the sense of 'non-probative'. Thus the text does circular argumentation. But we do not know exactly what kind of
not say that reciprocal argument is indemonstrable, and therefore
argument they sanctioned, and we do not know what (if anything}
fundamentally valid. It says that it is non-probative. Moreover, the
they alleged in its defence. What, next, can be said against it? Why
phrase 'and which is non-probative' expresses the scholiast's own
is circular reasoning a bad thing, 'to be avoided by the
view and not a view which he ascribes to the Stoics. Perhaps the
philosophers' ?
Stoics did regard reciprocal argument J.s non-probative. The text .
Once or twice Sextus suggests that with circular reasoning, as
does not say that they did.
with infinite regressions, we have no starting-point (PH III 22; cf. II
The text is interesting and frustrating. It allows us to say that
9). But, again, the remark is implausible: why not start wher� ver
some Stoics had discussed - or at least had named - r�ciprocal
you like? Any point on a circle - as Heraclitus observed - is a
argument. It tempts us to guess that they, like other ancient
starting-point.
logicians, had rejected it as non-probative. But we cannot tell
9 Reading 6 016:AAT]A6v f5i6:i\i\11A6s Walz) q>t)µl Tfiv 6rr65E11;1v Exwv, 1 0 Nevertheless, I may be allowed to refer to Jonathan Barnes,
with von Arnim: see vn.1 384.7-8 Walz. (But note that von Arnim's 'Aristotle, Menaechmus, and Circular Proof', Classical Quarterly 26,
text at SVF II 273 is horribly mutilated.) 1976, :z.78-512.
72 73
r
I
Once Sextus connects reciprocal argumentation with trying to all was fated before he came to believe that divination was possible.
prove 'the unknown by way of the unknown' (M vm 86), and hence And that is absurd.
with a class of mistakes - or supposed mistakes - which are Sextus' argument has a shine to it. An ingenious caviller might, it
generally collected under the title of obscurum per obscurius, is true, seek to defend Chrysippus' putative procedure along the
proving the obscure by way of the more obscure. 11 The connexion following lines. Suppose that Chrysippus already believes P,. Then
between reciprocity and obscurum per obscurius points to an at noon, say, he proves P1 from P2, thereby acquiring the belief that
objection which Sextus often brings - at least implicitly - against P1. But he at, once forgets P1• Why may he not now, at 12.1 5 ,
reciprocity. The objection involves the notion of priority. Here is a proceed to prove P2 from P1 ? In general, given sufficiently bizarre
simple case: conditions on his memory, Chrysippus may happily, and without
contradiction, go on forever proving first P1 from P2 and then P2
Thus the reciprocal mode turns up, according to which the
from P 1 . But whatever we make of such fantasies, they are not
mind must first have been judged ( rrpOKEKpfcr6a1) if the senses
are to be decided, and the senses must first be determined
worth pursuing here; for evidently they do not touch the nerve of
(rrpoS1aKpivecrBa1) if the mind is to be assessed. (PH II 68) Sextus' argument.
The real difficulty with the argument is this. It supposes that
You must judge the mind before you judge the senses, and also proofs are essentially ways of forming or generating beliefs. It
judge the senses before you judge the mind." In general, reciprocal works against the Dogmatists only insofar as they appeal to proofs
proof requires you to do A before B and also to do B before A. And in order to give an account of how our beliefs may be formed or
that is evidently absurd. generated. But the Dogmatists were not-or should not have been-
Now Sextus is surely right to imply that you cannot both do A concerned solely or primarily with this. Rather, they were - or
before B and do B before A. But why does he think that this is an should have been - concerned with the warranty or justification of
objection to reciprocal proof? He imagines, I suppose, that if you beliefs. Now it is plain that I may be justified in believing that P1
prove that P 1 by way of an argument 'P2: so P1', then you come to because I believe that P, even if I originally acquired the belief that
believe P1 on the basis of an antecedent belief in P2• Hence you must P, after I had acquired the belief that P1• Something like this is true, I
first believe P, (without yet believing P1}, and then acquire the belief suppose, of very many of our actual beliefs. I am pretty sure that I
that P1• Thus Sextus construes proofs as psychological events, myself believed the Magna Moralia to be spurious some years
which occupy a period of time and which end with the production before I acquired the beliefs which now warrant my confidence
of a new belief (or perhaps of a new item of knowledge}. Suppose that it is spurious. The beliefs which once led to or generated my
that Chrysippus simultaneously comes up with two proofs: 'Divi belief in the spuriousness of the Magna Moralia must have ante
nation is possible: so all is fated'; 'All is fated: so divination is dated that belief; the beliefs which now warrant or justify it did not
possible.' Then in virtue of the first proof, he must have believed - and that is no objection to then1. For justification is not a matter
that divination was possible before he came to believe that all was of producing new beliefs but of warranting old ones. The fact that
fated; and in virtue of the second proof, he must have believed that reciprocal proofs cannot be used to generate fresh beliefs does not
show that they cannot be used to warrant existing beliefs.
ll Sec further Karel Jan3.Cek, Sextus Empiricus' Sceptical Methods Thus Sextus' standard - if implicit - objection to reciprocal
(Prague, 1972), PP· 35-'7·
12 For the priority obicction see PH II 9, 199, i.02; 111 22; M VII 426; VIII proof is an ignoratio elenchi. Or at least, it is an ignoratio elenchi if,
122, 181, 261; M Ill 99 (priority expressed by way of verbs with the as I have supposed, the priority which it invokes is a temporal or
prefix TTpo-); PH II 9; M VIII 342, 380 {the preposition TTp6); PH II 20 chronological priority. But tetnporal priority is not the only variety
(the adverb rrp6TEpov); PH I l 17; M VII 341, 426; VIII 261 (the verb
1TEp1µEvE1v). of priority. Aristotle's main objection to circular proof also turns
74 75
on the notion of priority but not on the notion of temporal priority. proof and on which the Aristotelian objection to circular proof
According to Aristotle, turns.)
Let us sum this up by saying that if •n: so P' is a proof, then each
it is clear that it is impossible to prove anything circularly if a member of n must be epistemically prior to P. Now Aristotle
proof must depend on what is prior and more familiar. For it plainly takes the relation of epistemic priority to have two salient
is impossible for the same things at the same time to be prior
features: he takes it to be asymmetrical, and he takes it to be
and posterior to the same things. (APst 72b25�8)
transitive. In other words, if X is prior to Y, then it follows that Y is
It is quite clear from the context that 'prior' and 'posterior' do not not prior to X (this is asymmetry). And if X is prior to Y and Y is
here refer to temporal relations: Aristotle's objection to circular prior to Z, then X is prior to Z (this is transitivity).
reasoning does not turn on any ideas about the order in which our Suppose, then, that we are offered a reciprocal argumentation,
beliefs are acquired. say the pair of arguments: 'n1: so P 1 '; 'n2: so P2'· Now if the first
The Aristotelian thought is echoed by numerous later writers. argument is a proof, then each member of n1 is prior to P1 • And
Thus Ammonius: since P2 is in n1, P2 is prior to P1• But then, by asymmetry, P1 is not
prior to P2. Hence not all the members ofn2 are prior to P2. Hence
in proofs of this sort the same things are assumed to be both
'P,: so P,' cannot be a proof. It follows that any reciprocal argu
prior and posterior to, both more clear and more unclear
than, the same things. (in Porph isag 75.13-15) mentation contains at most one argument which is a proof. Hence
reciprocal argumentation is not probative, i.e. the pair of argu
And Elias: ments cannot prove both P, and P,. Asymmetry alone is thus
Reciprocal proof should be avoided by philosophers; for it enough to kill off reciprocal proof.
makes the same things prior and posterior to, more clear and Circular proof is tougher, and must be attacked by a combina
more unclear than, explanatory of and explained by, the same tion of asymmetry and transitivity. Consider the very simplest sort
things. For when we prove by way of Dio where Theo lives, of circle, consisting of three single-premissed arguments: 'P3 : so Pi ';
Dio is clearer, prior and explanatory. And when, on the 'P2: so P1'; 'P1 : so P/. Each constituent of this circle is supposed to
contrary, the proof is by way of Theo, Theo is clearer, prior be a proof; so P, will be prior to P,, P, to P, and P, to P,. Then by
and explanatory. transitivity, P3 is prior to P 1• But by asymmetry, P3 is not prior to P 1•
(proleg 9.15-20; cf. David, proleg 25 .4-15)
Hence at least one item in the circle is not a proof. Hence the circle
is non-probative.
Priority here is not a temporal but an 'epistemic' property - it is a
This argument must be generalized in two ways: it must be
property of the same sort as clarity and explanatoriness.
extended to cover circles of any size, and it must be extended to
The general thrust of the Aristotelian objection to circularity
embrace arguments with more than one premiss. But it should be
and reciprocity is this. If ·n: so P' is a proof of P, then there must be
easy to see how these generalizations can be made. And it should be
certain epistemic relations holding between the members of n and
clear that asymmetry and transitivity will together scupper circular
P: the members of n must together ground or support P; they must
argumentation.
explain P or make P intelligible or rational to believe; they must
The question, then, is this: Is the relation of epistemic priority
illuminate P; and so on. (These relations are different from and
asymmettiCal ahd transitive? if it is, then the sceptic has made out
additional to any logical relations which link n and P. In Aristotle's
his case against circular and reciprocal proof.
view, n must entail P if 'n: so P' is to be a proof or 6:TI65E1�15. For a
proof is a sort of syllogism: it is a deductively valid argument. But
the logical relations, in virtue of which ·n: so P' is a valid argunJent, It might be thought that it is simply obvious that epistemic priority
are distinct from the epistemic relations which make ·n:��g·p· a is both asymmetrical and transitive. After all, if X is prior to Y, then
77
surely Y can not be prior to X? And i f X is prior to Y and Y is prior

to Z, then surely it follows at once that X is prior to Z? But we need Amphinomus, a contemporary of Aristotle, must evidently have
to remember that our talk of epistemic 'priority ' is metaphorical - thought that in geometrical proofs the premisses are not explana
or at any rate, that it does not invoke the normal temporal notion of tory, even though geometrical proofs are surely paradigmatic
priority. Although in the literal sense the word 'prior' picks out a proofs, proofs in the strictest sense of the term.
relation which is asymmetrical and transitive, we cannot infer that Secondly, in everyday arguments the premisses are not usually
the word picks out a similar relation in its transferred use. More explanatory of the conclusion. Rather, they stand as evidence or
generally, we use a variety of metaphors in speaking of epistemic signs of its truth. Hermogenes' fictitious orator concluded that the
relations ('grounding' and 'supporting' are two familiar exam servant had committed adultery with the wife, and he concluded
ples); and the terms we use often seem 'evidently' to pick out a this from the fact that the servant had killed the husband. The
relation of a particular sort. But this may be a specious evidence - orator did not, of course, suppose that the murder explained the
we may find ourselves seduced to our intellectual ruin by the paint adultery. The murder was cited as evidence - perhaps even as
and powder of a caddied metaphor. conclusive evidence - for the adultery. It was, in the ancient jargon,
It might be thought, secondly, that at least one of the terms used a sign and not a cause. (And Hermogenes explicitly talks of signs,
in the Aristotelian account of epistemic priority is employed in a not of causes.)
non-transferred sense, and that here asymmetry (surely) and transi Thus even if Aristotle's appeal to explanatoriness will suffice to
tivity (perhaps) may quickly be established. The term I mean is exclude circular 6:TI0Sei�e1s, it will not exclude all circular proofs,
'explanatory'. If X is explanatory of Y, then surely Y is not let alone all circular arguments.
explanatory of X? And, perhaps less evidently, if X explains Y and It is plain that we need a clearer understanding of epistemic
Y explains Z, then surely X ('at one remove') explains Z? Since priority before we can proceed any further. What exactly is episte
Aristotle invokes literal and not metaphorical explanatoriness, mic priority? Unless we can answer this question, we cannot
there is no danger of specious evidence here. sensibly ask what properties the relation has. So what should we
True - but then it is far from clear that the notion of say? Well, although the notion of epistemic priority was introduced
explanatoriness must be a constituent of the notion of epistemic in connexion with Aristotelian proofs, we are not here specifically
priority. In the Posterior Analytics Aristotle is concerned with a interested in 6:TI08ei�e1s. For reciprocity and circularity may infect
very special type of proof or argument, a type which he calls any sort of proof or argument and are not proper to scientific
c:m65et�ll and which he regards as the paradigm of scientific demonstrations; and, in the context of scepticism, we must be
argumentation. In a scientific proof or 6:n6Se1�1s, Aristotle main interested in any argument in which the premisses are - or are
tains, the premisses must be explanatory of the conclusion. But alleged to be - sufficient to warrant belief in the conclusion. In
Aristotle does not hold that the same condition governs the short, we are centrally concerned with the notion of warranty or
premisses of any justificatory argument; and we should not our justification. The relation which matters in the present context is
selves try to extend the scope of Aristotle's thesis. the relation whicfi holds between two propositions when - very
For it seems clear, first, that the premisses of a proofneed not in roughly speaking - the one justifies you in holding the other.
all cases explain the conclusion. Thus according to Proclus, This relation is in a further sense relative - it is concerned not
with what warrants beliefs simpliciter but with what warrants
many people have held that geometry does not consider your beliefs or my beliefs. The notion is thus distinct from what
explanations and the question 'Why?'. Amphinomus held might be called an 'objective' notion of justification, a notion in
this view. (in Eucl 202._9-1 I)
which epistemologists have also been interested. I mention this fact
79
in order to be able to push aside a tempting but irrelevant argu X, or rather my belief in X, provides me with my warrant for
ment. The modern literature has stock examples which are de believing in Y. (It might be different in your belief system: your
signed to show that 'objective' justification is not a transitive
belief in Y might be what justified you in believing X. For example,
relation. Here is one of them." Consider the following three
the wife may justifiably believe that the servant killed her husband
propositions. because she knows that he has committed adultery with her. But a
P3: There were two consuls at Rome suspicion of adultery may not be what warrants the husband's
P2: There was a prime number of consuls at Rome belief that it is the servant who poisoned his food. Rather, he might
P1: There was an odd number of consuls at Rome finally believe in his wife's infidelity when he hears the servant
chucklingly confess that the mushrooms he dished up for dinner
Now - the suggestion is - P, plainly warrants P,; for P, actually
were death-caps.)
follows from P,, and if there were two consuls then there certainly
The first attempt will not quite do. For X may be one of a whole
was a prime number of consuls. Equally, but in a different way, P2
set of reasons which together warrant my belief in Y. (I may argue
warrants P 1 ; for of the infinity of prime numbers, all but one are
to Y not from a single premiss, the premiss that X holds, but from a
odd, so that if there is a prime number of anything, then it is
set of propositions one of which is X.) Hence we need to accommo
overwhelmingly probable that there is an odd number of it. But of
date the case in which X is just one of the things on which my belief
course P3 cannot Warrant P1 - that there are two things is hardly a
in Y depends. We might try this:
warrant that there is an odd number of things. On the contrary, P3
warrants the contradictory of P 1; for P1 entails not-P1 • Thus the sort X is epistemically prior to Y in my belief system just in case I
of justification or warranty illustrated by this example is not am justified in believing Y because of a set of beliefs, one of
transitive: P1 warrants P:z. and P2 warrants P 1; but P3 does not which is my belief in X.
warrant P 1 .
And we may as well put this a little more rigorously, thus:
This is ingenious and no doubt important. But I do not think that
it is directly relevant to my present concerns. I do not think it proves P2 is epistemically prior to P1 for believer x just in case P2 is a
that justification, in the sense which matters here, is not transitive. member of a set of propositions n1, and x is justified in
believing that P1 because x believes the members of n1•
For even if it is true that, in the example, P2 in some objective sense
warrants P1, it does not seem to me to be true that if I believe P2 I am I doubt if this formulation is exactly right. But I am fairly sure that
thereby justified in believing P,. Whether or not my belief in P, it is not wildly wrong - and I hope it will be good enough for the
justifies a belief in P, depends, inter alia, on my other beliefs. In moment.
particular, if I also believe P, - and especially if my belief in P, is Then is this relation of epistemic priority asymmetrical and
justified by a belief in P, - then belief in P, does not justify belief in transitive? If P, is one of a set of beliefs which justifies my belief that
P,. P,, does it follow that P, cannot be a member of a set of beliefs
.
So the epistemic priority we are looking for must be 'subjective' which justifies my belief that P,? I hope you may agree that it is at
or relative to a given system of beliefs. A first, rough attempt to least not obvious that the answer to this question is Yes- that it is at
explain it, then, might look like this: least not obvious that epistemic priority is asymmetrical. And the
X is epistemically prior to Y in my belief system just in case I same holds, and more so, for transitivity. If P3 belongs to a set of
am justified in believing Y because I believe X. beliefs which justifies P, and P, belongs to a set of beliefs which
justifies P,, does it follow that P, belongs to a set of beliefs which
justifies P 1? Again, it is not obvious - it is not obvious to me - that
13 John Watkins drew my attention to the example. See further his
Science and Scepticism (Princcton, t984), pp.61-4. this is so.
80 Sr
I think there are two reasons why we are tempted to take it as The flags are all numbered. The ropes have a 'direction' in the
obvious that epistemic priority is asymmetrical and transitive. sense that their ends are differently coloured: one end of every rope
First, we are seduced - as I have said before - by the metaphor of is green, the other is red. There are a few rules which determine
priority, and we tacitly ascribe to epistemic priority those proper how you may rope the poles together. The first rule is that every
ties which evidently hold of temporal priority. Secondly, and no pole must have at least one green end attached to it. A second rule is
less importantly, we tend to think in terms of the simplest exam permissive: a pole may have any number of ropes tied to it. A third
ples. If we look at the simple Chrysippean reciprocity, it is difficult rule says that at most one rope may directly link two poles - you
not to be sure that there is something badly wrong with reciprocal may not tie one pole to another by several strands of rope. A fourth
argument. But it does not follow, and we should resist the tempta rule requires - trivially - that no flag-pole ever be roped to itself. It
tion to infer, that the Chrysippean reciprocity is unacceptable is also true that no flag-pole is directly roped to any pole which is
precisely because epistemic priority is asymmetrical and transitive; more than a few hundred miles away from it; this is not, however, a
and hence we should not conclude from this example that there is rule - it is contingently determined by the fact that no rope is more
something wrong with every reciprocity and with every circle. than a few hundred miles long.
Most cases are complex cases, and complex cases are unlikely to be When your task is finished, the surface of the earth will be criss
clear cases. crossed by ropes. The criss-crossing will construct a large net or
But if the properties of epistemic priority are not obvious, how web. At the nodes or knots of the web there are the numbered flags,
can we establish them by argument? I am not sure. I shall end this waving in the winds.
chapter with two inconclusive lines of thought. The first proposes The rules do not determine any particular configuration or type
an analogy or model which is intended to make non-transitive of configuration for the web. But the web will probably be intricate
justification seem somehow attractive and plausible. The second and irregular and messy. It will probably form a sort of hollow
offers a popular argument which purports to show that the analogy sphere about the earth. In particular, it will probably contain
or model is misleading and that circular argumentation cannot ever within itself a vast number of irregular rope 'circles' - that is to say,
be satisfactory. if you start from some particular flag and track along the ropes
from pole to pole, always in a green-red direction, then you will
First, the analogy. You have a vast collection of flag-poles and often be able to get back to your starting-point again. Start from
flags. There are thousands - indeed millions - of them. You plant Flag One, in Naples, and proceed in a northerly direction, taking
the poles all over the surface of the earth. They need not be planted more or less a great circle route. Track along the ropes, some short
at regular intervals: they may come in clumps and clusters, there and some long, moving from green to red, from green to red.
may be forests of flag-poles and there may be relatively desert Eventually, perhaps, you will recognize that you are in Bari, and a
places. But no flag may be more than a few miles from its nearest dozen more ropes will get you back to Flag One in Naples.
neighbour, and most flags will have several flags within a mile of The point of the story is this. The numbered flags represent your
them. In addition to the flags and the flag-poles, you have a stock of beliefs. The planting of the flag-poles represents the acquisition of
ropes, an extraordinarily large stock. The ropes come in various beliefs. Different geographical areas with different densities of
lengths. The shortest rope is about a mile long, the longest rope is flags represent different areas or topics, on some of which you will
several hundred miles long. You must tie the flag-poles together. hold many beliefs and on others few. More importantly, the ropes
The tying and the planting may be carried out concurrently; but the are lines of justification. The direction of the ropes corresponds to
order in which you plant the poles need not at all determine the way the direction of justification. Thus if there is a rope with its green
in which you rope them together, and the way the poles are roped end tied to Flag One and its red end tied to Flag Seventeen, that
together will not reveal the order in which they were originally models the thought that in your belief system proposition P,, is
planted. epistemically prior to proposition P 1 - that P17 is one of the set of
beliefs which justifies your belief in P1 • If we follow out all the ropes in more or less the fashion I have indicated, as a network of
tied by their green ends to Flag One and collect the flags at the red interconnected propositions, each proposition tied to several
ends of these ropes, then we shall have collected the set n 1 on which others, and the whole set forming an intricately interwoven mass.
your belief that P, depends for its justification. The rope-web as a They have supposed that any particular belief is justified immedi
whole represents the tangle of justification which holds together ately by the other beliefs it is roped to, and that the belief system as
the flags of your beliefs. a whole is justified in virtue of the whole intricate texture of the
No flag-pole is tied to itself. That is to say, no belief is web - by the fact that each belief is roped to various other beliefs
epistemically prior to itself. Again, if there is a green-red rope from which are themselves roped to further beliefs, and so on.
one flag to another, then there is no other rope - and hence no red This interweaving is, I take it, one way of understanding what
green rope - from the second to the first. That is to say, the rules of have been called 'coherence' theories of knowledge or justification:
roping presuppose that epistemic priority is asymmetrical. But what justifies my holding a particular belief is the fact that it is
epistemic priority will not be transitive. Let us see why this is so. connected to other beliefs and thus forms part of a coherent belief
Take Flag One - F1 - again, tied in green-red direction to F1 7, system - the fact that it is connected, by various ropes, directly and
which is (let us suppose) five miles away from it. Now F1 1 will, in indirectly, to all or most or many of my other beliefs. (The term
turn, be tied in green-red direction to several further flags, includ 'coherence theory' has not been used to pick out any one determi
ing (say) F,,. Will F, also be tied toF,,? And if it is tied, will it be tied nate theory of knowledge- and often, or so I think, it has been used
in green-red or in red-green direction? The answer is that the rules in a culpably vague fashion. I do not mean to say that my flag
do not determine anything here. F1 may or may not be tied, in either analogy pictures the coherence theory - only that it pictures one
direction, to F7s. As a matter of fact, if F75 is more than a few reasonably intelligible version of that 'theory' or cluster of
hundred miles from F,, which it perfectly well may be, then it will theories.)
not be tied to F1• But as far as the rules go, F1 may or may not be tied, Coherence theorists have recognized that their views
in either direction, to F75 • Any of three states is possible, states countenance, or seem likely to countenance, something rather like
which can be pictured like this: circular argumentation. Some theorists welcome this, saying that
(1) G - - - - - - - - - - - R (17) G- - -- - - - - - - - R (75) there is nothing wrong with a circular argument provided that the
G- - - - - - - - - - - - - - - - - - - - - - - - - - - - - R circle is 'large' enough (provided that any green-red journey from
Naples back to Naples takes you past a reasonably large number of
(1) G- - - - - - - - - - - R (17) G- - - - ------- R (75) different flags). Other coherence theorists think that there must be
G- - - - - - - - - - - - - - - - - - - - - - - - - - - - - R
objections to circular reasoning, and so they maintain that there
(1 ) G- - - - - - - - - - - R G- - - - - - - - - - - R
(17) (75) are important differences between ordinary circles and the sort of
circles which form parts of a web or net of beliefs. I cannot myself
It should be clear what these pictures represent when flags are
see that this is a significant disagreement. The interesting question
translated into beliefs and ropes into justificatory connexions.
is not whether the 'circularities' within a web are rightly called
The metaphor of the 'web of belief' has been used by some
1 circles, but whether the circularities are vicious or virtuous. And
modern epistemologists. 4 They have pictured our belief systems,
that question turns, as I have said, on the nature of epistemic
14 Numerous models or analogies of this sort can be found in the priority: in particular, on the transitivity or non-transitivity of the
licerature. I take the phrase 'the web of belief' from W.V.O. Quine
and J.S. Ullian, The Web of Belief (New York, 1970). But Quine's relation.
webs are like spiders' webs, and there can be a significant difference The picture of the web assumes (as I think virtually all
between their centres and their peripheries. My webs may have any epistemologists have assumed} that epistemic priority is asym
configuration - they may be hollow spheres - and there need be
nothing in them corresponding to the distinction between centre and metrical. It also assumes that it is not transitive. Does it show that
periphery. epistemic priority is asymmetrical but not transitive? Does it
thereby show that at least some circular arguments may be accept of certain beliefs about the way in which cameras work. And so on
able? Of course it does not. A picture is not a proof. The rules of - the story could continue ad libitum, but enough of it has probably
roping which I proposed have no special status. Other rules can been told.
quickly be invented, other pictures painted. But a picture or a The question is this: Is my belief that Galen wrote a commentary
model may have a persuasive, or even an illuminating, force. The on Aristotle's Categories justified, in part, by my belief that photo
model I have described purports to have such a force. graphic film is light-sensitive? It seems to me that the answer is
For our belief systems are, as a matter of fact, web-like in their pretty clearly No. But if epistemic priority is transitive, then the
structure. You will soon come to see this if you start reflecting on answer to this and countless similar questions must be Yes. I do not
the beliefs you actually have. Kick the common philosophical habit insist on the conclusion that epistemic priority is not transitive. But
of picking on beliefs one at a time and asking of each one why it is I do insist that some strange and unwelcome consequences result
justified. Look instead at clumps of beliefs or, if you can, at the from the thesis that it is transitive.
whole set of your beliefs: you will find (or so the model invites you
to expect) that they do form a complex web, tied one to another in a Where do we now stand? It is at best unclear whether epistemic
messy and disorderly fashion by the cords of justification. And priority is asymmetrical and transitive. Hence it is at best unclear
since this is what your belief system is actually like, and since it whether or not the objection to circularity which we are consider
seems on the whole a reasonable and justifiable system, you will ing is enough to show that epistemological circularities are intoler
find that you must reject the supposed transitivity of epistemic able. Of course, even if the objection fails, it does not follow that all
priority. For your belief system is more or less reasonable; and it is circles - or even that any circles - are acceptable. Circularity is
incompatible with a transitive notion of epistemic priority. acceptable only if epistemic priority is not transitive. But if epi
Hence the story of the roped flag-poles gives us some reason for stemic priority is not transitive, it does not follow that circularity is
thinking - or suspecting - that epistemic priority is not transitive. acceptable - there may be other objections to circles.
At the very least, it gives us a reason for hoping that priority is not And so I shall end by briefly rehearsing a different objection, one
transitive. For if it is, then our actual beliefs which have no doubt which is frequently raised against coherence theories - and a
been assembled in a splendidly rational manner will turn out to be fortiori against circular argumentation. The objection is closely
an unreasonable muddle. parallel to an objection I considered in the previous chapter in
Consider a very simple example. It is a real example, not an connexion with infinite sequences of arguments or reasons. It runs
artificially invented case. I believe that Galen wrote a commentary like this.
on Aristotle's Categories. My main reason for this belief is simply Take any circle or web of beliefs, W. Let it be as complex as you
that Galen himself says that he wrote such a commentary. Why do I like. Now construct a different web, W'', in the following fashion.
think that Galen says he wrote a commentary on the Categories? Replace the constituent beliefs, all or some of them, by different
Because this statement is made in the lnstitutio Logica, which I and incompatible beliefs. (Replace 'Honey is sweet' and 'Naples is
believe to have been written by Galen. I believe that the /nstitutio in Italy' by, say, 'Honey is bitter' and 'Naples is in Greece'.) Rope
Logica was written by Galen for several reasons: one of them, and up the new beliefs in the same sort of way as you roped up the old
the most obvious, is that the sole surviving manuscript of the work the ropings follow the same rules and exhibit the same complexity
ascribes it to Galen. Now I believe this in part because I have a as before.
photograph of the manuscript on which I can read the letters Now compare W with W". In W we shall find a belief, P, which
Ai\HNOY, which I take to be the remains of IAi\HNOY ('Galen' s'), does not appear in wrc . Instead, W'' will contain some belief P*,
and I take the photo to represent the manuscript reasonably which is incompatible with P. Now P was supposedly justified by
accurately. I suppose that the photo is reasonably accurate because its position in W - we were allegedly justified in believing that P by
86
the fact that P was tied to or supported by numerous other the world; for every flag is tied to a pole, and every pole is planted
propositions which were themselves tied and supported in turn. firmly in the earth. It is simply wrong to say that the beliefs which
But W'� has just the same structure, in all relevant respects, as W. are roped to one another have no connexion with the world. The
Hence if P was justified by its position in W, P* will be justified by dialectical toing and froing at which I hinted at the end of the
its position in W"', But we know that P and P* are incompatible previous chapter may be practised in the same sort of way here too.
with one another. Hence we cannot be justified in holding both of But in the end I cannot see that the need to establish 'links with the
them. Hence we are justified in holding neither of them. And so, in world' will embarrass circular reasonf)rs.
sceptical fashion, we must suspend judgement. Although I find myself pulled by the objection from similar
In general, any appeal to W can be matched by an appeal to W*·. webs, and hence disinclined to give in to the seductive powers of the
One appeal succeeds if and only if the other appeal succeeds. But flagging game, I do not think I can yet produce a good argument
both appeals cannot succeed. Therefore �either appeal succeeds. which will either establish or refute the objection. Thus I end this
Webs, or circles, are no better in this respect than infinite lecture, as I ended the last, with a sort of temporizing scepticism: I
regressions. cannot yet tell whether circular arguments should be shunned by
Now a Dogmatist may object to this argument in exactly the philosophers.
same ways as he objected to the parallel argument about infinite
sequences. The particular objection I want to set out is this. The
sceptic in effect suggests that
You may properly base your belief in P on its position in a
web W only if you have reason to prefer W to any rival web
W*.
The Dogmatist replies that, on the contrary,

You may properly base your belief in P on its position in W,
provided that W is in fact a true and coherent web; you do
not need in addition to know that W is superior to W''.
The sceptical suggestion and the Dogmatic reply run parallel to the
suggestions and replies which I rehearsed at the ends of the two
previous chapters. Again, I postpone discussion.
But I may perhaps append a final observation, once again
parallel to an observation at the end of the previous chapter. The
objection to coherence theories which I have just sketched seems to
have an almost irresistible pull. It is tempting to explain the pull by
observing that webs and circles cannot justify beliefs because they
appeal only to other beliefs - they do not link beliefs to the world.
The web is, so to speak, detachable from the world; and for just
that reason it cannot serve any warranting function.
As I said before, this line of thought must be resisted. Webs, like
regressions, are (or so the Dogmatist may properly claim) linked to
88
Hypotheses
sources do not say much about this kind of utterance, they do give
4 some helpful illustrations. Thus according to Ammonius,
the Stoics say that there is an hypothetical kind of utterance,
Hypotheses e.g. 'Let it be laid down (VTToKe!o-6w) that the earth is the
centre of the solar sphere'. (in lnt 2.31-2)
And the Stoic Epictetus provides the following example:
We should behave in life as we do in the case of hypothetical
utterances. 'Let it be night'. Let it be (EcrTw). 'Well, then, is i t
day?' No; for I assumed the hypothesis o f its being night. 'Let
it be that you think that it is night'. Let it be. 'Then take it that
it is night'. That does not follow from the hypothesis.
At the start of M III, his essay Against the Geometers, Sextus
(diss I xxv II-13)
decides to deal with the so-called hypothetical method of geo
metrical proof. He observes that 'for the sake of orderly progress, Thus it appears that the characteristic mark of an hypothetical
we should realize at the beginning that things are called hypotheses utterance is an initial third person imperative: 'Let it be supposed',
in many different senses' (M III 3). Sextus is prudent; and since the 'Let it be', VrroKe{cr6w, EcrTw. This is confirmed by Sextus, who
subject of this chapter is the hypothetical mode of scepticism, I speaks of 'the mathematicians who assume by hypothesis the first
shall follow his example and first say something about the different principles of their proofs and theorems, saying "Let it be granted
sorts of things which the Greeks called hypotheses or about the (6E56u6w)"' (M III r 7 ) . Sextus plainly takes 'Let it be that . . .' as the
different uses of the Greek word VTT66ecr1s. mark of an hypothetical utterance; and he plainly takes hypotheti
Like Sextus, I shall leave aside the special senses which the word cal utterances to be the canonical way of expressing hypotheses.
bears in rhetoric and in dramatic theory and focus on the senses Then an hypothetical utterance has the form: 'Let it be the case
which it has in logical and epistemological contexts. Even within that P'. The hypothesis itself is the content of the utterance: in
logic, the word l11r66ecr1s was used in many ways. The different uses saying 'Let it be the case that P', you hypothesize - you lay it down
all have something in common -roughly speaking, an hypothesis is as an hypothesis - that P; and what you hypothesize, viz. that P, is
something which is set down or laid down or supposed. Thus the hypothesis. Hypotheses are not in any normal sense a class of
Philoponus can use the word Vrr66ecr1s to denote the antecedent of a propositions; for we cannot intelligibly ask, in the abstract,
conditional proposition (e.g. in APr 243.15-24) - in a conditional, whether or not a given proposition is an hypothesis. A proposition
you 'lay down' the antecedent and see what follows. Similarly, is an hypothesis when, and in the context in which, it is hypo
Proclus can use the word Vrr66ecr1s to denote the subject-term of a thesized; and it is thus an hypothesis not absolutely and without
universal affirmative proposition (e.g. in Eucl 252.5-23) - in saying qualification, but relatively and within a determinate context of
'All isosceles triangles have their base-angles equal', you 'suppose' discourse. Is the proposition that it is night an hypothesis? The
you have an isosceles triangle and 'conclude' to the equality of its question is misconceived. Has the proposition that it is night ever
base-angles. been advanced as an hypothesis? Yes - in the little discussion which
Neither of these sorts of 'laying down' or 'supposing' concerns Epictetus reports or invents.
me here. I am interested in cases where an hypothesis is a complete It may seem to follow that anything at all may be hypothesized;
utterance, not part of an utterance. A good starting-point is the for surely I may sensibly say 'Let it be that P' for any proposition P
Stoic classification of utterances. For the Stoics distinguished what whatsoever? (By 'sensibly' I mean 'without syntactic or semantic
they called an 'hypothetical' kind of utterance. Although our impropriety': what we may hypothesize sensibly in the sense of
9r
Hypotheses Hypotheses
'advisedly' is another matter.) In a way this is true, and it is a truth Can virtue be taught?, it will evidently be fruitless to hypothesize
which Sextus will exploit to the full in the hypothetical mode. (It is that the price of wheat will rise next year.) Secondly, in
also a truth important to and familiar from many modern systems hypothesizing that P, you do not commit yourself to the truth of P,
of formal logic. For 'natural deduction' systems standardly contain nor do you assert that P. You may, of course, hope- or even expect
a 'rule of assumption', and such a rule, in the words of one - that it will turn out that your hypothesis is true. But in the nature
textbook, 'permits us to introduce at any stage of an argument any of things, many hypotheses will turn out to be false, and you will
proposition we choose as an assumption of the argument'. Sextus abandon them: in doing so, you are not going back on anything you
will not be showing himself captiously or idiosyncratically said or changing your mind. (In the special case of reductio ad
Pyrrhonian if he insists that we may hypothesize anything whatso absurdum proofs, you may positively hope that the hypothesis will
ever.) But if you may hypothesize absolutely anything so far as the turn out to be false; for you hypothesize that P, and then try to
form of the hypothesis goes, the purpose and function of derive a contradiction or an absurdity, which will allow you to
hypothesizing may yet put constraints on the content of permiss deny that P. Here you are evidently not asserting that P or commit
ible hypotheses. ting yourself to the truth of P when you say 'Let it be the case that
What is - what was - the function of hypotheses? Here the P'.) Thirdly, in hypothesizing that P, you do not argue for P or
Greeks had two distinct traditions. The first tradition may be produce any sort of reason in favour of it. Later, of course, you may
called Platonic, since it is best known from certain famous passages look for reasons in support of the proposition you had previously
in Plato's dialogues! In the Platonic tradition, hypothesizing has hypothesized; but insofar as you produce reasons and argue that P,
primarily an heuristic function. You are interested in some pro to that extent you do not offer P as an hypothesis. Indeed, one of the
blem or question, say the question: Can virtue be taught? In order points of saying 'Let it be the case that P' (rather than simply 'P') is
to make progress towards an answer to the question, you venture precisely to indicate that P is being put forward - at least pro
some hypothesis. Perhaps you hypothesize or suppose ('for the te1!1pore - without argument or grounds.
sake of argument') that virtue is a kind of knowledge. You next see The second tradition of hypothesizing I shall call Aristotelian.
whether, granted the hypothesis, you can answer the original Aristotle himself used the word Vrr66ecns in a variety of ways, but
question. If you can, you proceed to worry about the hypothesis one use had a particular importance. For Aristotle held that among
itself, asking whether virtue is in fact a kind of knowledge. And you the first principles or 6:pxai on which any science is based there will
will typically tackle this question by advancing a second be hypotheses (APst 72a20-4). Hypotheses are, in this sense, a
hypothesis. species of what we tend to call axioms, the first or primary or
The details of the Platonic tradition of hypothesizing are contro primitive principles from which the remaining truths or theorems
versial. But its general features are, for present purposes, clear of a science are derived. This Aristotelian usage had an afterlife.
enough. Plato's hypothetical method is, as I said, an heuristic Proclus, for example, uses the word VTI66ecr1s in what he takes to be
device; and that fact determines the nature of the hypothesizing the Aristotelian way in order to characterize a subdivision of the
which it involves. Thus, first, although you may in principle axioms of Euclid's geometry (in Eucl 76.24). But the precise delin
hypothesize anything, in practice the context of the business and eation of this special Aristotelian sense need not detain us; for the
the form of the question in hand will determine that some hypo term VTI66ecrts came to be used more generously, so that all first
theses are better and more useful than others. (If the question is, principles, and not merely a subgroup of them, could be called
hypotheses. I shall call this use the broad Aristotelian use of the
1 I mean the reflections on the 'hypothetical method' at Meno 860- term (though it is not clear that the use is to be found in Aristotle
87s; Phaedo 100A-101E; Republic 510B-511E. (Plato did not himself himself).
invent the method: he adapted it from contemporary geometrical
practice.) Proclus records the broad Aristotelian usage:
93
Axioms, postulates and hypotheses are distinguished accord� - for your Aristotelian hypotheses. I stress this point. It will become
ing to Aristotle's own exposition. Often, however, all these important.
things [i.e. all first principles] are called hypotheses. It is clear, too, that in making an Aristotelian hypothesis you are
(in Eucl 76.24-77.3; cl. I78.I-I4) committing yourself to a truth. In laying down something as a first
A similar remark is found some centuries earlier from the pen of the principle, you are thereby supposing that it is true; and if it emerges
Peripatetic scholarch, Alexander of Aphrodisias: that the hypothesis is false - that the propbsition you took to be an
hypothesis is not even true, let alone a first truth - then you were
Hypotheses are first principles of proofs, because there is no I
mistaken and your hypothesizing was an error. Aristotelian
proof of such propositions, i.e. of first principles, but they are hypothesizing is a sort of asserting.
posited as evident and known in themselves (aVT60ev) , and
Finally, you will not - of course - be prepared to hypothesize
. . .
what is assumed without proof they call an hypothesis (or

anything. In hypothesizing that P, you are taking P to be a first
even, more generally, a thesis) and say that it is hypothesized.
principle. Not every proposition, not even every true proposition,
(in APr 13.7-II)
can function as the first principle of a science.
This broad Aristotelian use of Vrr60ecrts is not common in early
writers; but Proclus and Alexander were recording a usage, not Platonic hypotheses and Aristotelian hypotheses are thus two
inventing one, and the word is found not infrequently in later very different birds. And it is Aristotelian hypotheses, broad
Greek texts in just this sense.'" Aristotelian hypotheses, with which the Pyrrhonists were con
The Aristotelian form of hypothesizing has one important fea cerned. In M III, Sextus remarks that
ture in common with the Platonic form: in hypothesizing that P you
in a third sense, we call hypotheses the first principles (O::pxai)
do not argue that P or produce any reason in favour of holding that of proofs; for an hypothesis is the postulating (a\TT\crts) of a
P. (As I said, this feature is characteristic of hypothetical utterances fact for the establishing of something. (M nI 4)
as such and is implicit in their canonical mode of expression.) But
the explanation of this feature of hypothesizing is quite different in This, he says, is the sort of hypothesis which he is going to
the Aristotelian case. In the Platonic case, you hypothesize that p as investigate (M m 6).
part of an heuristic strategy and you may well proceed later to M III is directed specifically against the geometers. We might find
argue that P. In the Aristotelian case, hypothesizing is not an it surprising that the geometers should be taken to use Aristotelian
heuristic but a demonstrative activity: in hypothesizing that P you hypotheses. For Plato associated his hypothetical method with
lay it down, as a first principle, that P; and your hypothesis is not a geometrical practice, and we ourselves might be inclined to think
mid-point in the search for truth but a starting-point in the demon that Platonic rather than Aristotelian hypotheses were the stuff of
stration of truth. You do not argue that P, because, if P is indeed a geometry. But when Sextus talks of the hypothetical method of the
first principle, you cannot argue that P. Aristotelian hypotheses are geometers, he has in mind a method of proof which begins by
things for which no argument is possible. They are not temporarily laying down, or . hypothesizing, certain propositions as first
unsupported (as they are for Plato) but necessarily unsupportable. principles:
They are 'evident and known in themselves', as Alexander puts it. the geometers) seeing the mass of problems which dog them,
They cannot, trivially and by definition, be known on the basis of retreat into what they think to be a matter safe and free from
anything else; and it is for this reason that you will not argue - ever danger, namely the postulating of their geometrical first
principles by hypothesis. (M III r)
2 A few examples from Sextus: PH 1 183; M IX 2, 419; M v1 5 (TO:s As far as Sextus is concerned - and hence as far as we are concerned
&px1KO:s VTI06Ecre1s}; from Galen: sect ingred 1 93 K; us part 111 45-6 K;
foe aff Vlll 324, 316 K; lib prop XIX 43 K. - the geometers' hypotheses are &pxai, Aristotelian hypotheses.
94 95
Nor is Sextus being perverse or unfair; for the definition of hypo those who assume something by hypothesis and without
theses which he gives at M III 4 was lifted verbatim from a geometri proof are satisfied by a bare assertion (1..fJ1Ai] <p6:cr1s) alone.
cal textbook. (We happen to know this because a learned bishop of (M m 7)
Laodicea by chance copied the very same text into his notebook.') A 'bare' assertion is not a type of assertion; to make a bare assertion
Although in M III Sextus is gunning for geometry, he does not is to make an assertion barely - it is merely to assert. 5 Sextus
suppose that the hypothetical method is peculiar to the geometers. supposes that hypothesizers assert, and do nothing more than
Indeed, he characterizes hypotheses at M III 4 in entirely general assert. This supposition is - or seems at first blush to be - justified.
terms, and he illustrates the type of hypothesis he has in mind by For, as I explained, Aristotelian hypothesizers do indeed commit
rehearsing the three principles on which the doctor Asclepiades themselves to the truth of what they hypothesize (they produce a
hoped to ground medical science (M III 5). Elsewhere Sextus q>6:cr1s); but they do not offer any argument or reason for the
observes that hypothesis (they produce a \fllA1\ q>c'xcn5).
Now the fact that hypotheses are bare assertions is, for the
it is not only proofs but pretty well the whole of philosophy
sceptics, the central fact about them. Hence Sextus will often use
which the Dogmatists claim to advance by way of hypothesis.
the phrase 'flIAfi q> c'xa15 when he wishes to refer to an l'.rrr66.a15- and
(M Vlll j69)
usually where the phrase \fllA1\ q>c'xa15 occurs the hypothetical mode
And he is right. The Dogmatists did indeed suppose that all oHnoxfi is being invoked.' Moreover, since a \fllA1\ q>c'xms is simply a
q>•Aocroq>fa, all knowledge, depends ultimately on c'xpxai or cp6:cr1s- since a bare assertion is simply an assertion - Sextus can use
(mo6foei5. And in attacking the hypothetical method and the the term cp6:cr1s, without qualification, to advert to an hypothesis: the
claims of hypothesizers, Sextus is attacking the foundations of all context will make it clear that when he says 'They use a cp6:cr1s', he
Dogmatic knowledge. means 'They use a bare cp6:cr1s', i.e. 'They use an lrrr6Becr1s' . 7
The attack employs what Sextus calls the hypothetical mode of Making a bare assertion contrasts with offering an argument or
scepticism or the mode 'from an hypothesis' (PH r r64, 168). This reason for your assertion. In Sextus, the contrast is often explicitly
mode is adduced more often than any other form of Agrippan drawn; for Sextus often uses the hypothetical mode in conjunction
argumentation, Otcx<pwv{cx excepted: it is discussed and justified with some other sceptical mode in order to produce a dilemma for
thrice, and it is employed on some forty or fifty occasions in Sextus' the Dogmatist: 'Either he makes the bare assertion that P, or else he
different writings. The prevalence of the hypothetical mode might argues that P.' (The contrast is usually between making a bare
be doubted; for in fact Sextus refers to (mo6foei5 by name on very assertion and offering a proof or 61r6Se1�1s;8 but other contrasts
few occasions indeed. 4 But the doubt would be mistaken - there are also occur.9) This provides Sextus with another way of referring to
numerous implicit references to hypotheses in his writings. It is For comparable uses of l.VIA6s in Sextus see e.g. M vn 376 {a 1.V!Afi
worth dwelling on the matter for a moment. 6Ai\o[wo1s, a change and nothing more); VIII 476 (a l.VIAT] eE01<;:,
Sextus begins his discussion of hypotheses in M III by remarking merely entertaining a thought without assenting to it); M 1 49 {a
l.VLAfi yvWa1s, merely knowing your letters without being able to
that explain their discovery or their nature).
6 For 1.V1Afi qi6:01s see PH u 1z.1; M vn 315; v111 15, 76, 368, 435/6; M 111
7. Note the synonymous phrase, 1.VLAT] Un6axEa1s, at M 111 179, i.59.
3 The bishop was Anatolius; the text in question is most easily found 7 For qi6:a1s sec PH II 107, 153i M VII 315, 337, 339; VIII 61, i.81, 360,
in volume IV of J.L. Heiberg's Teubner edition of Hero (Leipzig, 444; M I 157, 188, i.79. Note also 6noqia(VET01 µ6vov at M VIII 15.
1912): Definitions 138.8 (=p.166.4-16). See Karel Jan3.Cck, "O ES 8 'An-65Et�!S is the contrast at PH 1 60, 114/5, 12.2., 173; 11 107, 121, 153;
Uno6EcrEwS Tp6nos', Eirene i.5, 1987, 55--65. M VII 315, 339; VIII 15, 61, 76, 78, i.lh , 343, 374, 463/4; M I 157, 188,
4 Outside the discussions of the hypothetical method and mode (PH 1 i.79; II 109; Ill 7, 8, 12., 13, 34-
168, 173-4; M VIII 369-18; M Ill 1-17) see PH I 177, 186; II 20; M VIII 9 Contrast with Kpnfip1ov: PH 88; M v11 337, 440; v111 i.6; with i\6yos:
343, 367. (In addition, Sextus sometimes uses the word Un-66Ecr1s in M VIII 120, 463; with Un-6µvricr1s: M VIII 444; with Eqio5os or µE0o5os:
other contexts where the hypothetical mode is not at issue.) M VIII 436.
97
hypotheses. For instead of writing 'Either a bare assertion or a The remark about when to use the hypothetical mode is absent
proof', he will often write 'Either no proof or a proof' to make from the account in Diogenes Laertius (IX 89), and perhaps it was
exactly the same point. So that in some contexts the phrase 1no an ill-judged addition by Sextus himself." (Otherwise the descrip
proof' or the like must count as an implicit reference to tion in Diogenes is in substance the same: there are some minor
hypothesizing and an implicit invocation of the hypothetical differences in expression, which would repay examination; but I do
mode.'° Thus Sextus has a variety of ways of adverting to the not think that they have any effect on the philosophical under
notion of an hypothesis. And the fact - a mildly curious fact - that standing of the mode.)
he rarely adverts to the notion by using the term VTI66ecr1s is Sextus nowhere explains why the hypothetical mode should be
perfectly consistent with the fact that he frequently refers to thought to further the ends of the sceptic, why it is a mode of
Dogmatic hypothesizing. irroxiJ. But part of the answer should be apparent from my discus
sion of the two previous modes. If the only thing that can be said for
Sextus' references to Dogmatic hypothesizing are all of them or against P is that some Dogmatist has hypothesized it, and if
hostile; and it is the hypothetical mode which he employs in his hypothesizing that P does not establish or warrant belief in P, then
attacks upon this hypothesizing. The mode is named at PH 1 r64 we should suspend judgement over P. The doctor Asclepiades
and formally described at I r68: hypothesized his three apxcxi, and thought thereby to establish his
medical science. He did nothing more. But hypothesizing, as the
The mode from an hypothesis occurs when the Dogmatists,
hypothetical mode allegedly shows, is a valueless procedure.
being thrown back ad infinitum, begin from something which
Hence, as far as Asclepiades' hypothesizing goes, Oaov ETii TOVTC}>,
they do not establish but claim to assume simply and without
proof by virtue of an agreement.
we should suspend judgement about his apxcxl. There is thus a
parallel between the hypothetical mode and the modes of regres
The description is less than illuminating. He says when the hypo sion and reciprocity: each of them leads to scepticism by way of a
thetical mode will be adduced, viz. whenever the Dogmatists qualifying oaov irri TOVT<p. But I shall show later how the hypo
attempt to escape from an infinite regression. But he does not say thetical mode is also tied in a more intimate way to suspension of
what the mode is - he does not here explain why the sceptic objects judgement.
to Aristotelian hypothesizing or how the mode is supposed to Sextus does give us, at least implicitly, an explanation of what
reduce hypothesizers to irroxiJ. the hypothetical mode consists in. For at PH I r73-4, in the middle
Moreover, even the account of when the hypothetical mode is of his illustrative example of how the Five Modes of Agrippa may
used seems to be false: Sextus implies that the Dogmatists only work together to achieve suspension of judgement, Sextus some
hypothesize when they are being threatened with an infinite regres what inappropriately offers three arguments against the reason
sion, so that the dilemma they confront should always take the ableness of using Aristotelian hypotheses. The three arguments
form 'Either you barely assert or else an infinite regression follows'. reappear at M VIII 369-78, where they are joined by a fourth
But the Dogmatists are not invariably confronted by that particular argument, which is interpolated between the second and the third
dilemma - it is one of several dilemmas which Sextus sets for them. arguments of the PH text, and also by a rebuttal of a Dogmatic
counter-argument. And at M III 7-17 the four arguments of M VIII,
lo See Pff I 60, 122; II 34, 54, 85, 113, 121; Ill 34; M VIII 436 (all usmg together with the rebuttal, reappear. The arguments which support
&va:rr65etKTOS, O:vev 6:rro5ef!;ews, or the like). Also Pl-l 1 114; M vn
440; VIII 120 (O:KpiTWS cf, PH 11 88, &verr1Kphws (if the text is sound)); the hypothetical mode implicitly explain what it is.
and PH 111 23; M IX 204 (xwpls a:iTLa:s). Note also the common use of
a:UT66ev in this context to point a contrast with µeT' 6:rro5ei�ews: M
VIII 26, 28, 75, 120, 343, 371; M 11 109; 111 9; and cf. E� ETofµov at M 1 I So Karel Jan3.Cek, 'Skeptische Zweitropenlehrc und Sextus
VIII 26, 78. Empiricus', Eirene 8, 1970, 47-55.
99
The three texts to which I have referred are very close to one offering us is designed precisely to show that hypothesizing is a
another, in language as well as in content. There can be no doubt suspect business - and it is odd to argue that hypotheses are suspect
that they are all drawn from the same source, which Sextus copied by assuming that they are suspect. You might perhaps imagine that
three times with greater or less fidelity. I myself suspect that the Sextus is reporting some widespread suspicion of the hypothetical
version in M 111 is closest in content to the source, and hence that method: 'People as a rule look askance at hypotheses - so don't
Sextus took his attack on hypotheses from an earlier sceptical make them if you can avoid them.' But it seems to me unlikely that
discussion of geometrical method." It is at least possible that this Sextus means to offer this sociological observation; and even if he
source was an essay by Agrippa himself. But this is speculation, and does, it will carry little weight. No decent Dogmatist will worry
speculation of a pretty pointless kind. much if laymen regard his procedures with suspicion. What mat
Better turn to the arguments. I shall first outline the four argu ters is that they are scientifically legitimate, not that they get a
ments of M VIII and M 111, and then comment upon them. The favourable press.
rebuttal I leave until later. The second curiosity about the argument is yet more puzzling. If
(1) If it is acceptable for a Dogmatist to hypothesize that P, i.e. to lay what you hypothesize is true, Sextus urges, then 'you should
down P, by a bare assertion, as a first principle or O:pxr), then it must assume it directly as true (ai'.JT66Ev Aaµ136vE1v ws 6A116is)' (M vm
be equally acceptable for a sceptic - or another Dogmatist - to 37I; M 111 9 ); for 'no one hypothesizes what is true and is the case'
hypothesize that P'', where P" is the 'opposite' of P. But if P'i- is no (M m 9 ) . But whatever can be the contrast between hypothesizing
less acceptable than P, we cannot accept P as a first principle just something on the one hand and 'directly' assuming it as true on the
because the Dogmatist hypothesizes it. (See PH I 173; M VIII 370; M other? Sextus himself characterizes hypothesizing in terms of as
Ill 8.) suming in his initial description at PH 1 168; and he frequently uses
(2) What the Dogmatists hypothesize is either true or false. If it is true, the adverb 'directly (ai'.n66Ev)' in allusions to hypotheses." If a
they should not hypothesize it (for hypothesis is 'a matter full of
Dogmatist hypothesizes that P, then he thereby assumes directly,
suspicion') but rather assume it straight off. If it is false, it can do
i.e. without argument, that P is true; and if a Dogmatist assumes
them no good - for a false starting-point cannot ground a science or
directly that P is true, then he thereby hypothesizes that P.
a branch of knowledge. (See PH I 173; M VIII 371; M III 9-10.)
(3) If the Dogmatists hold that the consequences of an}' hypothesis are Hypothesizing and directly assuming are one and the same thing.
acceptable, then all enquiry is subverted. For, given any absurd Sextus' contrast is false - and perplexing. (It is true that some
proposition, we can find some hypothesis from which it follows; Dogmatists made terminological distinctions between hypotheses
hence any proposition whatsoever will be acceptable. And this is and assumptions; but these technical distinctions are - so far as I
evidently si!Iy. (See M VHI 372-3; M III lI-12.) can see - of no relevance at all to Sextus' argument.)
(4) If in order to establish that P2 you first hypothesize that P1 and then I cannot find any way to make the second of Sextus' arguments
derive P2 from P1, why not establish P2 directly, by hypothesizing it, even mildly plausible. Nor can I discover any persuasive explana
and thus save yourself the labour of looking for arguments? (See PH tion of how Sextus might have come to advance this bewildering
I 174; M VIII 374; M III 13.) line of thought.
The second of these four arguments is curious, for two reasons. The remaining three arguments rely each on a similar supposi
No doubt any hypothesis is either true or false; and no doubt any tion. The first argument supposes that if you may hypothesize that
false hypothesis is a 'rotten foundation' (M 111 rn). But what is P, then you may equally hypothesize that P'' (where P* is an
wrong with a true hypothesis? Sextus alleges that hypothesis is 'a 'opposite' of P). The fourth argument supposes that if you may
thing full of suspicion' (M VIII 371; M Ill 9 ) . But the argument he is hypothesize that P, then you may equally hypothesize anything
12 For a contrary view, see Jan:iCek, "O E� lnroeEaEW) Tp6nos'. lJ Sec the references in n.10.
JOO IOI
which follows from P. And the third argument supposes - or seems (where it is under investigation or in dispute whether or not P2). He
to suppose - that if you may hypothesize that P, then you may adduces the hypothesis that P, from which he will proceed to show
equally hypothesize anything at all: that if you may hypothesize that P,. Sextus retorts that if you are ready to hypothesize that P,
one thing, then you may hypothesize any thing. If that is so, then we you may with equal justice hypothesize that P,.
might consider the third argument to be, in effect, a general If what is hypothesized is firm and reliable qua hypothesized,
argument of which the first and the fourth are specific instances. then let the Dogmatic philoJophers hypothesize not the
It might be doubted that the third argument does make the points from which they infer the unclear item but rather the
wholly general supposition that if you may hypothesize one thing unclear item itself, i.e. not the premisses of the proof but the
you may hypothesize any thing. What Sextus actually affirms is conclusion. But let them hypothesize this ten thousand times
that 'if someone claims that whatever follows from what is as over, it will not be warranted, because it is unclear and
sumed by hypothesis is firm, then he confounds all philosophical because it is under investigation. So it is evident too that if
enquiry' (M vm 372). He offers two particular illustrations: let us they postulate the premisses of the proof without proof, they
achieve nothing towards warranting them, because they too
hypothesize that three is equal to four (M vm 372; M m r r ) , or let us
are matters of controversy. (M vrn 374)
hypothesize that what is in motion is at rest (M m 12). And he
points out that in each case we shall get lunatic results. Both The argument here is compressed. Sextus takes a Dogmatist to
examples are cases of inconsistent or self-contradictory hypo argue for P, by offering the following proof: 'P,; so P,'. And the
theses; so perhaps the third argument supposes that if you may Dogmatist is supposed to advance P1 as an hypothesis. The Sextan
hypothesize that P, then you may hypothesize (not anything at all, retort is this: 'Then why not hypothesize P, directly?' 'Plainly',
but) some inconsistent proposition. Yet it is hard to see how this comes the reply, 'this will not do; for P:i. is unclear and under
supposition could be made plausible. How could anyone think that investigation - it needs argument, not bare assertion.' 'Well then,'
if he accepted any one piece of hypothesizing then he would have to concludes the sceptic, 'you may not hypothesize P1 either - for P1
accept some inconsistent hypothesis but would not have to accept will also be under investigation and a matter of controversy.'
every hypothesis whatever? Surely, if something is going to oblige There are two oddities about this argument. First, Sextus does
you to allow an inconsistent hypothesis, it will oblige you to allow not explain why he thinks that P, will be under dispute. He
any hypothesis at all? So I incline to think after all that the third supposes that any and every hypothesized proposition will be a
argument does tacitly make the utterly general supposition that if matter of controversy, but he does not say why he makes this strong
one hypothesis is acceptable then any hypothesis is acceptable. supposition. I can only imagine that he is tacitly invoking the mode
However that may be, it is plain how the Dogmatist will attempt of disagreement, or rather the claim made in the presentation of
to rebut Sextus' third argument. For he will certainly deny the that mode that everything is subject to disagreement or Otaq>wvia.
general supposition that if you may hypothesize one proposition If everything is under dispute, then everything hypothesized will be
you m"y hypothesize any proposition. And he will certainly deny under dispute. Now since whatever is subject to disagreement is
the specific supposition that if any hypothesis is acceptable then &811Aov or unclear, and since a Dogmatist will not hypothesize
some inconsistent hypothesis must be acceptable. Indeed, he will what is unclear, nothing can be hypothesized. (If this is what Sextus
surely maintain that no inconsistent hypothesis is acceptable - and has in mind, then the hypothetical mode is here made to depend for
for an obvious reason. Any inconsistent proposition is thereby its efficacy on the mode of disagreement.)
false; and Aristotelian hypotheses are advanced as truths. There Secondly, it is not clear that the Dogmatist can only justify his
fore no inconsistent proposition may be acceptably hypothesized. refusal to hypothesize P, by appealing to the fact that P, is unclear
And a Dogmatist will react in a similar way to the fourth of and a matter of investigation. Surely he might instead - and better
Sextus' arguments. The Dogmatist wishes to establish that P, offer a formal reason for refusing to hypothesize P,. He might hold
I02 I03
that if a proposition is provable, then it may not be hypothesized. making the historical claim that 'the Dogmatists' in fact adduced
For Aristotelian hypotheses are intended to be offered as first and accepted Sextus' first argument; and they may observe that
principles; hence they cannot be proved. So if there is a good Sextus, like any other philosophical controversialist, will often say
argument from P, (or from any other proposition) to P,, you should in a dialectical context 'They state that so-and-so', without intend
not hypothesize that P,; for P, cannot then be a first principle. ing his remarks in any rigorously historical sense. This is true.
(Matters are in fact a little more complicated than this; but it is None the less, the two texts do give us some reason for thinking
clear that a purely formal criterion will rule out at any rate some of that Sextus' first argument was also used by some Dogmatists. And
the hypothesizings which the Pyrrhonist wishes upon the it is, of course, a thoroughly Pyrrhonian practice to turn Dogmatic
Dogmatist.) weapons against the Dogmatists.
In both the fourth and the third arguments, then, Sextus makes a Sextus' account of the first argument runs like this:
supposition of the form 'If it is legitimate to hypothesize P, then it is If what they say they assume by hypothesis is warranted
legitimate to hypothesize X'. The Dogmatist will reject both sup because it is assumed by hypothesis, then its opposite
positions; and - which is the important point here - he may do so (EvavT{ov), when assumed by hypothesis, will also appear
on purely formal or logical grounds. For he may say, against the warranted - and in this way we shall hypothesize conflicting
fourth argument, that if P2 can be proved from P 1 1 then P2 may not things. And if in the latter case - I mean, in the case of the
be hypothesized. And he may say, against the third argument, that opposites - the hypothesis is weak in warranty, then it will be
if a proposition is self-contradictory (if it entails something of the weak too in the former case, so that this time we shall
hypothesize neither. (M VIH 370; cf. M HI 8)
form 'P and not-P'), then it may not be hypothesized.
Up to this point, Oaov Errl TOUT(}>, the Dogmatist surely has a Here Sextus uses the word Evav1iov, 'opposite'. Elsewhere, and
case. I do not wish to suggest that Sextus has no possible answer to more often, he employs 6:v11Keiµevov ('contradictory') or one of its
the case. But I think that any answer will in effect lead into the first cognates. The two words are to be taken as synonyms, and they are
of Sextus' four arguments against hypothesizing, the argument to to be construed in a generous sense: P* is 'opposite' or 'contradic
which I now turn. tory' to P just in case it conflicts (µc'xXETat) with P (cf. PH I rn, 190).
The essential thought is this: P and P* are opposites provided that
It is plain from our texts that the first argument is the chief and the truth of the one excludes the truth of the other." (Usually,
most important of the four. Outside the three passages in which Sextus writes as though any claim must have exactly one 'oppo
Sextus discusses the faults of hypothesizing, he never alludes to the site'. But he should not be held to this thought, and his argument
second or to the third of his arguments, and he alludes only once to will be unaffected if we allow that a claim may have several
the fourth (M vm 343) . But he frequently refers to the first argu 'opposites'.)
ment, thereby suggesting that he thinks it the best or most persua Thus the argument is this. Suppose a Dogmatist hypothesizes
sive argument - perhaps, indeed, that it is in itself constitutive of that P. Then there .will always be some proposition P*, an 'oppo
the hypothetical mode." site' of P or incompatible with it, which we may hypothesize (or
Moreover, he twice says that this argument was actually used, which someone else may hypothesize or indeed actually has hypo
and therefore accepted, by the Dogmatists themselves (M vm 360, thesized). Suppose, then, that hypothesizing that P warrants the
463). In both passages Sextus refers generically to oi 5oyµaTIKoi: he Dogmatist in affirming that P and warrants our believing his
neither names actual Dogmatists nor cites actual texts. Some affirmation. Then hypothesizing that P* must warrant the affirma
scholars may be led to wonder if the passages really warrant our tion and legitimate the belief that P*. Hence we shall hypothesize -
14 See PH n 107, 153; 111 23; M vu 315, 337; VIII 15, 26, 28, 76, 78, 120,
281, 360, 436, 463; M I 157, 188, 15 For the notion of conflict see note 12 to Chapter I .
and be warranted in believing- both P and P*. But P and P* conflict should one accept every offered hypothesis? And if not every
with one another; and since we know them to conflict, we cannot one, then which? (diss l vii 22-3)
be justified in believing both of them to be true. Suppose, then, that
Epictetus clearly means that we should not accept 'every offered
we therefore reject P\ holding that hypothesizing it gave us after all
hypothesis'; and he implies that studying logic will enable us to
no grounds for belief. Then we must equally reject P, for exactly the
answer the crucial question, 'And if not every hypothesis, then
same reason.
which?'. Thus Epictetus is in effect saying that one of the main
The argument is clear, and its conclusion irresistible, provided
benefits of Stoic logic is that it can provide an answer to the
that the general principle implicit in it is true. The general principle
hypothetical mode of the Pyrrhonists. (And I like to think - though
is this: for any proposition P there is a proposition P" such that (i) P
it may be deemed rash to do so - that Epictetus was actually aware
and P" are mutually incompatible, and (ii) hypothesizing that P will
of the problem posed for Stoic philosophy by the hypothetical
give warranty to P just in case hypothesizing that P" will give
mode of the Pyrrhonists.)
warranty to P*.
Now why should Sextus suppose that it would be difficult to find
Now it is evident that any Dogmatist will deny, and hope to
an answer to the crucial question? Why should he think that the
refute, this general principle. He will of course allow that for any P
selection of hypotheses is a difficult- even an impossible - matter?
there is an incompatible p•:-; but he will maintain that if it is
We might expect him to have noted that his first argument
legitimate to hypothesize P, then it is not legitimate to hypothesize
against the hypothetical method is more threatening than the third
P*. He will claim, in short, that he is entitled and able to choose
or fourth arguments in the following way. All three arguments rely,
among putative hypotheses.
as I have noted, on some general principle to the effect that if you
Sextus is perfectly aware of this. In connexion with his third
may properly hypothesize P then you may equally hypothesize X.
argument, after he has himself pretended to hypothesize that three
In the case of the third and the fourth arguments, the Dogmatists
is equal to four and that what moves is at rest, he notes:
could (or so it seemed) offer a purely formal criterion on the basis of
which the relevant principles could be rejected and the Sextan
And just as the geometers will say that these hypotheses are
out of place (for the foundation must be firm if what follows argument evaded. But in the case of the first argument no formal
is to be agreed upon), so we ourselves will refuse to admit criterion offers itself. True, we may properly take it as a formal
without proof anything which they assume hypothetically. principle that if P and P* conflict, then both may not be hypo
(M III r2; cf. VIII 373) thesized. But no fortnal criterion will enable us to determine which
of several conflicting hypotheses (if any) we may legitimately ad
The Dogmatists - here, the geometers - are made to say explicitly vance. Now you may say that there is nothing especially desirable
that not any putative hypotheses may legitimately be hypothesized: about finding a formal criterion for rejecting putative hypotheses:
some hypotheses are 'out of place', and the wise man will choose any criterion will do, formal or informal. I agree. But it is in fact far
among potential hypotheses. The same point is made in a passage easier to hit upon formal criteria, and far easier to commend them,
of Epictetus. Epictetus is discussing the question: What is the point than it is to formulate and make persuasive an informal criterion.
of studying logic? He answers that errors in reasoning are immoral However that may be, Sextus himself does not advance this
and 'against one's duty (napix TO Ka6ijKov)', and that they can only consideration. He has, I think, a simpler and more general point in
be avoided by studying logic. mind. It emerges from the following text - one of the many
passages in which the hypothetical mode is applied.
And the same is true of hypotheses and hypothetical utter
ances. Sometimes it is necessary to postulate (aiTijcra1) an The Dogmatist will not simply assert. For then one of his
hypothesis as a sort of basis for the ensuing argument. Now opponents will utter the assertion which claims the opposite,
I06
and in this way the former will be no more warranted than the and its horsepower. For it is because hypothesizing is hypo
latter - for one bare assertion is worth the same as another. thesizing - and not because it is a sort of bad argument that it
(M VII 315) induces or helps to induce suspension of judgement. The regressive
The crucial sentence is the last sentence: 'one bare assertion is mode and the reciprocal mode help to produce scepticism insofar
as regressive arguments and reciprocal arguments are bad arguR
worth the same as another'. Since P and P* are each bare assertions
they have the same worth; they are, as Sextus says elsewhere : ments. The hypothetical mode serves the sceptic in virtue of the
special nature of hypothesizing.)
equipollent or laoa8evTjs (M vm 436). But in that case, as Sextus
Now if the Dogmatists try to choose among putative and rival
asks at M VIII 120, 'why should we assent to one rather than to the
hypotheses, they must surely offer some reasons for rejecting their
other?' We have no reason to prefer P to P'' or vice versa; and as for
opponents' P* and for preferring their own P. But in that case, as M
the Dogmatist, he 'will have nothing to say against anyone who
asserts the opposite' (M vm 26).
!II 12 implicitly remarks, they will be abandoning the hypothetical
method and seeking instead for proofs or arguments. For insofar as
The idea is simple and effective. If you advance P as an hypo
they produce reasons for P and against P*, to that extent they are no
thesis, then by definition you neither argue for P nor allege any
longer treating P as an hypothesis, and they cannot claim that it is a
thing at all in its favou r - your utterance of P is a bare assertion. But
first principle. The hypothetical mode of the Pyrrhonists has forced
then someone else hypothesizes that P*. There is nothing to be said
them to abandon the hypothetical method.
for P which cannot be said for P'', or vice versa. And for a
In all this the Pyrrhonist has, or so I think, a very plausible case;
delightfully straightforward reason: there is nothing at all to be
and in it lies the strength of the hypothetical mode and of Sextus'
said for P and nothing at all to be said for P*. If there were, they
first argument against the hypothetical method. But the Dogma
would not be being hypothesized. Again, the Dogmatist cannot say
tists are not yet dead, nor has their hypothetical method been
anything against P*. For anything said against P* would be some
finally exploded. In the next chapter I shall consider this matter
thing said in favour of P. But nothing is said in favour of P, since P is
further. Here I conclude by casting an eye over the remaining part
hypothesized.
of Sextus' description of the hypothetical mode, viz. his reply to the
One general Pyrrhonian strategy consists in the setting up of
Dogmatists' attempt to save their method.
equipollent oppositions: the sceptic takes two incompatible views
and shows that they are laoa8EVij, equally strong, in that the
The Dogmatists had, Sextus says, a customary reply to the hypo
arguments for the one exactly match the arguments for the other.
thetical mode:
Hence suspension of judgement. Now the hypothetical mode i s - as
the mathematicians put it - a 'limiting case' of this strategy. The In reply, they habitually say that a warrant that the hypo
two hypotheses, P and P», are laoaeevij, equipollent or equally thesis is strong (EppWcr6a1) can be found in the fact that what
strong, in that they are supported by exactly matching arguments. is inferred from the hypothetical assumptions is found to be
And they are supported by exactly matching arguments because true - for if what follows from them is sound, then the
assumptions fiom which the conclusions follow are also true
they are not - and cannot be - supported by any arguments at all.
and indisputable. (M VIII 375; cf. M Ill 14).
(Here- to revert briefly to an earlier matter - we can see how the
hypothetical mode is more closely connected to E-rroxl\ than is An hypothesis is thus tested by its consequences: we may prefer P to
either the regressive mode or the reciprocal mode. Like them, the P'' provided that the consequences of P are true and the conse
hypothetical mode will usually (but not always) work in collabora quences of P'' false.
tion with other modes; unlike them, there is something about its This idea (which finds its origins in Plato's Phaedo) is not
own specific character which determines the type of work it does obviously silly. Modern epistemologists may see in it something
ro8
akin to the procedure known as 'inference to the best explanation'. to be thereby confirmed or given strength. But the Dogmatists may
We may - it is said - rationally choose among competing hypo not sunnily assume that we can discover the truth or falsity of the
theses by selecting the hypothesis which best explains the pheno consequences of the hypothesis.
mena. For example, we may prefer a Copernican to a Ptolemaic Sextus is not merely wheeling in the old Pyrrhonian machine; he
theory of the solar system just because the Copernican theory is not merely calling on general sceptical conclusions ('We can
offers a better explanation of the observed celestial phenomena. never discover the truth') and then applying them specifically to the
The two theories may be regarded as conflicting hypotheses. Each consequences of the hypothesis. Rather, his argument is directed
has certain consequences or implications for the phenomena; and specifically at the status of P, as the consequence of an hypothesis.
we shall rationally choose the theory or hypothesis with the truer The Aristotelian hypotheses are apxai. They are epistemologically
implications. Thus a theory - an hypothesis - is properly tested by prior to their consequences. They are epistemological primitives.
its consequences. 16 The Dogmatists adduce them in order to explain how we can
Sextus raises two objections (M VIII 376-8; M III r 4-r 7) . His properly warrant beliefs or even claim knowledge. We may claim
second objection accuses the Dogmatists of bad logic; for he (they say) to know that P2 in so far as we can derive it from some
represents them as arguing in favour of the hypothesis that P, by known apxl\ or (m6ern15; and we can claim (they add) to know the
inferring it from two premisses, viz. P2 and the conditional 'If P1, apxl\ or (m66rn15 in virtue of itself or directly - we may simply
then P,'. (They consider the consequences of P., and if the conse hypothesize it.
quences are true, they infer that the hypothesis is true. Hence they Now if the Dogmatists go on to make the further claim that the
commit the fallacy of 'affirming the consequent'.) But this is hypotheses may be chosen for their consequences, the whole epi
ignoratio elenchi on Sextus' part. For the Dogmatists do not - or stemological order of things is overturned. Instead of resting our
need not-hold that the truth of the hypothesis can be inferred from claim that P2 on our claim that P1, we go the other way about: P1 is
the truth of its consequences. Rather they hold - or they should justifiably hypothesized because it yields P,, which we may, prop
hold - that an hypothesis may be confirmed or given strength erly and independently, claim to know. There is a crucial difference
(eppwcrea1) by the truth of its consequences. It is one thing to hold here between the old hypothetical method and the modern method
that P, follows deductively from P, (together with 'If P., then P,'), of inference to the best explanation. The modern n1ethod is a
another to hold that P 1 is confirmed or corroborated by P,. The method for producing explanations. It takes the phenomena as
matter is no doubt less simple than Sextus' Dogmatists make it given and it seeks to justify a theory or hypothesis which will then
appear, and there may be serious logical objections to their pro explain the phenomena. The ancient method is a method for
cedure; but they are surely not guilty of the logical howler which justifying or establishing the consequences of an hypothesis. It
Sextus ascribes to them. takes the axioms - the hypotheses - as given and then seeks to
Sextus' first objection is epistemological rather than logical. The justify the theorems. Thus if the Dogmatists do attempt to select
objection is simply this: How can we discover whether or not the hypotheses by way of their consequences, they suffer an epistemo
consequences of a given hypothesis are true? If we could discover logical bouleversement. And one thing is pretty plain: by this
that P,, say, was true, then perhaps we might allow the hypothesis bouleversement they destroy the hypothetical character of their
hypotheses, and they change totally the nature - and hence the
16 For a critical account of 'inference to the best explanation' see Nancy epistemological appeal - of the hypothetical method.
Cartwright, How the Laws of Physics Lie (Oxford, 1983), pp.4-18. I Should we conclude that Sextus successfully resists the Dogma
do not of course mean that the ancient Dogmatists discovered this tists' reply to the hypothetical mode? Yes and no. Yes: for he does, I
method of inference: their response to the hypothetical mode is
similar to, but not identical with, the method. believe, show that their reply involves a substantial change to the
IIO III
Hypotheses
hypothetical method - so great a change that the whole character
of hypotheses and their epistemological role is altered. No: for 5
Sextus does not show that a new form of hypothesis in a new role
may not do epistemological duty. Sextus may be right in what he
says, but more remains to be said. And that, too, will be a subject of
The sceptic's net
the next, and final, chapter.
In the four preceding chapters I have discussed four Pyrrhonian

'modes of scepticism', four general argument-forms by which the
Pyrrhonists thought that suspension of judgement might be in
duced. I have discussed the modes as though they were independent
argument-forms, each by itself capable - in certain contexts - of
introducing scepticism. And these modes were often so used by
Sextus.
My discussion has left many loose ends, some of which I shall tie
up in this last chapter. In addition, the previous chapters have
deliberately ignored one important fact about the Agrippan modes;
for, as I have said, the Four Modes I have discussed were part of a
set or group of modes: they are four of the Five Modes of Agrippa.
The fifth Agrippan mode is the mode of relativity, the mode O:rro
ToO Trp6s Tt. It is a strange beast, and it poses numerous and
interesting problems; but it belongs - or so I think - to a different
species from the other Four Modes, and I shall say nothing about it
here. 1 Rather, I want to consider some ways in which the Agrippan
P1odes may form a sceptical system; in particular, I want to
consider what I shall call the Pyrrhonian net, a net in which the
sceptical gladiators thought they could entangle their Dogmatic
opponents.
Although Sextus often uses the Agrippan modes independently
of one another, he more often uses them in groups - in pairs or in
triplets. And it is plain that they were considered to have some sort
of systematic interrelationship. This is hinted at in Sextus' initial
characterization of the modes; for he says that the hypothetical
I For some discussion sec Julia Annas and Jonathan Barnes, The
Modes of ScefJticism (Cambridge, 1985), Chapter 1 1 .
I I2 II}
The sceptic's net The sceptic's net
mode is used 'when the Dogmatists are thrown back ad infinitum', move on to R3 • IfR3 is not a new idea (if it is identical with P or with
i.e. when they are threatened by the mode of regression (PH I I68). R,), then the reciprocal mode is brought up. If it is new, then the
And after this initial characterization he remarks that 'every matter hypothetical mode obliges us to produce a further reason, R,. And
of enquiry can be brought under these modes' (1 I69) and proceeds in this way, by repeated application of Oia<pwvia, hypothesis and
to construct a system involving all Five Modes (1 170-7). reciprocity, we are led into an infinite regression, R4 being sup
The System of Five Modes - the system which Sextus presents, ported by R,, R, by R,; and so on without end. But this endless path
and which is perhaps his own invention2 - is a curious thing: it is is forbidden by the mode of regression.
positively rococo in its complexity, yet it possesses neither aesthetic (Note that when the reciprocity mode is used in this or in any
elegance nor philosophical cohesion. Nor are the Agrippan modes other systematic fashion, it must be construed so as to outlaw any
ever again used in this systematic fashion in Sextus' writings. form of circular argumentation. Suppose that R4 is not a new idea:
I shall not trouble to expound the Sextan system. In any case, it is if it is identical with Ri, then we have a simple case of reciprocity in
easy enough to construct a modified version of the system which the narrow sense of the word; but if R4 is identical with R1 or with
involves only the four Agrippan argument-forms I have discussed. P, then we have a circularity rather than a reciprocity. I shall
The System of Four Modes works in the following fashion: take continue to talk in this chapter of the reciprocal mode - it would be
any problem ?Q. Suppose that there are (at least) two incompatible confusing to change the established nomenclature. But the reader
solutions to it, P and P*. Now, by the 01a<pwvia mode we shall be should bear in mind that the arguments which the 'reciprocal'
aware that there is disagreement over ?Q, some opting for P and mode is intended to exclude may well be circularities rather than
others for PY-« Hence if we are to answer the problem ?Q we must true reciprocities.)
decide or resolve the 01a<pwvia. Suppose we think that P is in fact The System of Pour Modes is my own invention in the sense that
the correct answer to ?Q. Can we warrant or justify this thought? no ancient text explicitly describes it. But it has some claim to
At the outset we seem to have two possible procedures: we might historical reality; for it is implicitly employed by Sextus in several
simply affirm P without more ado; or we might offer some reason passages. The briefest and clearest is this:
in support of P. If we follow the first procedure and simply affirm P, If the disagreement about the criterion is to be decided, then
then the sceptic will adduce the hypothetical mode - to our bare we must have an agreed criterion by which to judge it; and if
assertion of P he will oppose the bare assertion of P*, and we shall we are to have an agreed criterion, we must first decide the
be stymied. Hence we must follow the second procedure. Let us disagreement about the criterion. Thus the argument falls
then advance reason R 1 in support of P. Now, by Staq>wv{a R 1 will into the reciprocal mode, and the search for a criterion
be contested, and we must somehow decide in its favour. We reaches an impasse. For we shall not allow them to assume a
cannot - by the hypothetical mode - merely assert it. Therefore we criterion by hypothesis; and if they want to judge a criterion
by a criterion, we shall throw them back ad infinitum.
must produce some reason in favour of R 1 • Let that be R 2 • Well,
(PH II 20)
either R 2 is identical with P or it is a new idea. If it is identical with P,
then we are brought to scepticism by the reciprocal mode. (For we Sextus does not exactly follow the map of argument I sketched out,
are supporting P by R, and R, by P.) If R, is new, then it will be and his invocation of the reciprocal mode is flawed. (It is a case of
subject to dispute. We cannot resolve the dispute by merely assert what I have called generic reciprocity, not of genuine reciprocity.}
ing R, (the hypothetical mode forbids this easy option). Hence we But he does intend to use the four modes which the System in
vokes; and it is plain that his brief argument can plausibly be
:z. Note the first person (UrroBEi�oµEv) at the start of the system (1 169) expanded and organized by way of the System of Four Modes.
as opposed to the third person (rrapa8186ao-1} at the start of the
description (1 164). But this point is less than probative. There is an oddity about the Four Modes: one of them,
II 4 rr5
S1acpwvla, seems to perform a different function from the other inclined to think that this system comes from Agrippa himself.' But
three. Whenever the hypothetical mode or the mode of regression the System of Two Modes introduces neither new modes nor two
or the reciprocity mode is deployed, the result is '1rox1\ - a line of modes. Rather, at PH I 178-9, Sextus outlines a way in which three
thought is blocked or shown to lead only to suspension of judge of the familiar Five Modes may be deployed in concert. The three
ment. But 01aq>wvfa does not block any line of thought - it does not modes are 8ta(}lwv\a, regression and reciprocity; and by their joint
itself induce Eiroxi). Then what is its function? application we are supposed to see that any issue whatsoever must
In my first chapter I mentioned the possibility that the existence lead to scepticism and Eiroxf\.
ofOtaq>wvia should be regarded as a necessary condition for broxl\:
if an opinion is not disputed, we shall not suspend judgement over Since everything which is known is thought to be known
either from itself or from something else, they point out that
it. I urged both that there was no good textual evidence for
things are known neither from themselves nor from anything
ascribing this idea to the Pyrrhonists, and that the idea is philo
else, and so think to introduce perplexity about everything.
sophically untenable - for it is simply false that suspension of That nothing is known from itself is clear, they say, from the
judgement requires disagreement. Hence in the Four Modes we disagreements among the natural scientists, which have, I
should not regard 01aq>wvia as supplying an epistemologically suppose, concerned every matter, whether an object of per
necessary condition for the generation of scepticism. Rather, it ception or an object of thought, and which are undecided; for
should be thought of as a psychologically useful aid to the sceptic. If we cannot use either an object of perception or an object of
there is no disagreement at all on some issue, then you might well - thought as a criterion inasmuch as anything we might take
if erroneously - imagine that there was no room or reason for will be a matter of disagreement and hence without warranty.
doubt, that you were justified in assenting to the opinion insofar as And for this reason they do not allow that anything can be
there was no dissentient voice. Hence the observation of disagree known from anything else. For if that from which something
is known must always itself be known from something else,
ment is pertinent to Pyrrhonism: it draws attention to the fact that
they fall either into the reciprocal or into the regressive mode.
assent should not be given without ado - doubts might be raised
And should you wish to assume something as known from
because doubts have been raised. This psychological use of dis
itself, so that something else can be known from it, then you
agreement is, I think, an important feature of Sextus' writings. are blocked by the fact that nothing is known from itself (for
After all, he is concerned (or he pretends to be concerned) to lead us the reasons I have given). (PH I I?8-9)
to scepticism; and anything which will entice us, or show us reason
to be enticed, is useful to him. The System of 'Two' Modes may be shown by means of a
That said, we might think to dismiss 5taq>wvia from further diagram. Start with the Dogmatist's claim - with any claim - at the
discussion. Its role as an independent purveyor of scepticism I have top of the diagram. The possible paths which the Dogmatist may
already investigated: in the System of Four Modes it has no take are marked by the downward diagonal lines. When an item
philosophically interesting role to play. Sextus says 'They also hand down (TTapa5156ao-l)' the Two Modes
(PH 1 178). The-subject of Tiapa5!56ao-1 is surely 'the more recent
sceptics', who are referred to in I 177 as the begetters of the Five
And so we might think to develop a system of three modes. Now Modes. Strictly speaking, it does not follow that the same more
there is in fact a system of three modes in Sextus. But it has a recent sceptics produced the Two Modes; but it seems to me that the
misleading nomenclature - and it does not contain the particular natural and obvious way of reading Sextus' text has him ascribing
the same paternity to the Two as to the Five Modes. Since Agrippa is
three modes we want to be left with. Having dealt with the Five father of the Five, he is also father of the Two. (This is an
Modes, Sextus turns briefly to what he calls 'two other modes of unorthodox, though not a novel, view. Most scholars, for no good
reason, assume that the Two Modes must derive from an author
suspension' (I 178). The Two Modes form a system, and I am later than Agrippa.)
r r6 r r7
appears in the diagram for the first time, it is called 'new'; other Here 5ia<pwvia has, after all, a function in a systematic presenta
wise it is old. When a diagonal is blocked by horizontal bars tion of the Agrippan modes. (As he says, Sextus has undecided
(= = = =), the meaning is that, so far as this line of argument
=
disagreement in mind.) The thought is this: any proposition which
goes, the Dogmatist must suspend judgement. Since all possible you may claim to know 'from itself' - any claim which is allegedly
paths in the diagram are eventually blocked by horizontal bars, the self-supporting- will be subject to disagreement; and the disagree
Dogmatist must suspend judgement in all cases (see diagram r). ment will be undecided, since we have no agreed decision pro
p
cedure by which to decide it.
Thus the place of 81aq>wvia in the System of 'Two' Modes
Diagram r depends on the argument which I discussed at the end of the first
chapter. Although that argument is interesting, I do not want to
!If-supporting
discuss it again; and although the 'Two' Modes form an interesting
system, they ignore the hypothetical mode - which, as I have said, is
a mode of the first importance to the Pyrrhonists. I therefore
propose to replace the 'Two' Modes by what I shall call the System
supported by R1
of Three Modes. The new system is in fact isomorphic with the
/ System of 'Two' Modes: all that happens is that the position of

81aq>wvia in the 'Two' Modes is taken in the Three by the hypo
thetical mode. Thus the mode of disagreement finally departs from
(reciprocal) these pages. I do not deny its interest or its sceptical importance.
But I have no more to say about it; and it will - I hope -be plain that
the System of Three Modes is superior to the System of 'Two'.
The Three Modes work like this. Suppose you are considering
the claim that P. Then either (r) the claim is merely asserted, or else
self-supporting
(2) it is supported. If (r), then the hypothetical mode applies. If (2),
then P rests on some reason or set of reasons, R1 • Either (2a) R1 is an
'old' item, i.e. (in this case) it is the same as P, or else (2b) it is a new
supported by R2
item. If (2a), then the reciprocal mode applies. If (2b), then either
(2bi) R, is merely asserted or (2bii) R, is supported. If (2bi), then the
hypothetical mode applies. And so on . . . until the regressive mode
is invoked.
Again, a diagra !11 will make things plainer. The new diagram is
(reciprocal)
constructed on the lines of the old (see diagram 2). This diagram,
Rz new
and the System of Three Modes, represents - or so I think - the
/'\
self-supporting
\
\
philosophical core of Agrippan scepticism; and it conveys what is
epistemologically most important and most challenging about this
aspect of ancient Pyrrhonism.
======== ==
' ,
\
\
The Three Modes, like the Four, are in a sense my own inven
tion. But they too have an historical reality of sorts. And I shall
(regressive)
rr8 r r9
p that any known proposition must be based on some other known
Diagram i.
proposition; that if I know that P, then I must base my knowledge
/\ on knowledge of some R;. In that case, of two things one. Either
/\
every item will be supported by a new item, and an infinite chain of
�s�� _ supported by R1 knowledge will be generated. But, according to Aristotle, such
y��th;,.�;;) infinite epistemological sequences are impossible. Or else at some
point in the process an old item will be reintroduced, and we shall
have a circular proof. But, according to Aristotle, there are
_ _R� o�d_ _ R1 new insurmountable objections to circular proofs.
-
------- /'/\\
(reciprocal) Hence, Aristotle concludes, we must reject the suggestion that
every known proposition must be based on some other known
proposition - if we are to have any knowledge at all. In some cases
_ �� _
(hy-p�th;,�;;) /\ supported by R1 we may properly claim knowledge even though we do not base the
claim on any further claim.
Thus Aristotle rejects infinite epistemological regressions and
implicitly accepts the Pyrrhonist's regressive mode. He rejects
Rz old R1 new circular reasoning and implicitly accepts the Pyrrhonist's recipro
cal mode. But he rejects scepticism, and explicitly supposes that at
~
(rectprocal)
some point in the argument a 'new' Ri may, with reason and
justification, be simply asserted. Thus he implicitly rejects the
�:e��d-
_ supported by R1 Pyrrhonist's hypothetical mode or else implicitly requires that
\
(hypothetical) \
there must be some via media between hypothesizing and support
\
\
ing, between producing a bare assertion and offering supporting
\
\ reasons.
The connexion between APst A 3 and the Agrippan modes has
(regressive) often been noticed. 5 There is the closest thematic similarity - the
diagram by which I illustrated the System of Three Modes (or the
System of 'Two' Modes) will equally serve in an elucidation of
indulge in a brief historical comment before I turn to the funda Aristotle's argument. Moreover, there are linguistic parallels be
mental philosophical issues which the Three Modes raise. tween Aristotle's text and Sextus' exposition of Agrippa's modes.
The similarities are striking, and they can hardly be accidental.
I have already referred more than once to the passage in Aristotle's There must be so�e historical link between Aristotle and Agrippa.
Posterior Analytics where he adverts to epistemological regres Now we know that there was a renascence of interest in
sions and to circular proofs. In that passage, APst A 3, Aristotle is Aristotelianism towards the end of the first century Be.' We might
concerned with the general possibility of knowledge (or rather, of
'scientific' knowledge, of imaTi]µT)), and hence he is implicitly 5 Sec e.g. A.A. Long, 'Aristotle and the History of Greek Scepticism',
in D.J. O'Meara (ed.), Studies in Aristotle (Washington DC, 1981).
concerned with scepticism.4 He considers the suggestion 6 See most recently H.B. Gottschalk, 'Aristotelian Philosophy in the
Roman World from the time of Cicero to the end of the second
4 For a detailed analysis of APst A 3 see Jonathan Barnes, Aristotle's century AD', in W. Haase (ed.), Au(-tieg und Niedergang der
Posterior Analytics (Oxford, 1976), pp.106-12. ROmischen Welt 11 36.2 (Berlin, 1987).
120 I2I
guess that Agrippa, who lived in roughly this period, read, or assertion and rational support. (And I am inclined to think that
learned about, the Posterior Analytics; that he was impressed by there is not a genuine disjunction here: there are not two distinct
Aristotle's discussion in A 3; and that he saw how it could be things which different epistemologists have attempted to do;
adapted to Pyrrhonian ends. He had only to agree with Aristotle in rather, their attempts to escape the net may - or so I suggest - be
his treatment of regression and reciprocity while rejecting described indifferently as rejections of the hypothetical mode or as
Aristotle's own way of escaping from scepticism. This rejection he inventions of a via media.)
could have secured in either of two ways: by appealing to the The Aristotelians, then, try to characterize a class of proposi
notion of 81mpwvfa, already dear to the Pyrrhonists, and thereby tions which may with propriety be hypothesized. There are some
constructing the System of 'Two' Modes; or by adducing the propositions, they urge, which we may reasonably accept as hypo
hypothetical mode, familiar from sceptical arguments against the theses - because in their case either bare assertion is admissible or
hypothetical method of the geometers, and thereby constructing else hypothesizing is something other than bare assertion. These
the System of Three Modes. Thus Agrippa's modes - or their propositions will then form the basis of our belief systems; they
central philosophical core - will have derived historically from the wil1, in a standard metaphor, constitute the foundations of our
Posterior Analytics. Aristotle was unknowingly the great-grand knowledge. (The metaphor is an ancient one.' The modems have
father of the sturdiest child the sceptics ever produced. used it in subtly different ways, and I do not pretend that my use of
The story, as I have sketched it, requires - and can be given - it here is the only permissible or fruitful use.) Thus Aristotelians
further elaboration and refinement. But it must remain to some offer a 'foundationalist' epistemology. Such an epistemology di
extent a fantasy; for although there is no evidence against it, and vides our items of knowledge or our justified beliefs into two
although some such history is surely needed to account for the classes: some items of knowledge - some justified beliefs - are
facts, there is no evidence for any particular or detailed version of fundamental or basic; the remainder are dependent or derivative.
the historical tale. Let us hope that new discoveries - perhaps in the An item is dependent or derivative just in case it derives its
excavations at Herculaneum - may eventually produce the telling epistemic status- its status as an item known or justifiably believed
document. Until then, we shall be prudent if, while believing that - from some other item (or set of items) which is known or
Aristotle somehow influenced Agrippa, we refrain from speculat justifiably believed. An item is fundamental or basic just in case it is
ing about the exact way in which the influence was exerted. known or justifiably believed and yet does not derive this epistemic
I turn therefore from historical to philosophical speculation. status from any other item which is known or justifiably believed.
How might a Dogmatist, ancient or modern, evade the Pyrrhonian More formally, we may offer the following schema as an abs
cast? Or how might he wriggle out of the net in which the Three tract and general representation of a foundationalist epistemology.
Modes tangle him? x knows or justifiably believes that P just in case
either (i) there is some set of propositions n such that (a)
As far as I am aware, no philosopher has seriously supposed that every member of n is known or justifiably believed by x,
the mode of regression should be rejected, or that a Dogmatic and (b) P stands in an appropriate epistemic relation to the
epistemology could be founded on infinite sequences of beliefs. members of n;
The reciprocal mode has been questioned, implicitly, by some or else (ii) P belongs to the class f3 of basic beliefs.
philosophers both ancient and modern. But the ancient patrons of As I have said, almost all Dogmatists have been foundationalisrs.
circularity have left us nothing of constructive interest. In point of Hence almost all have subscribed to one instance or another of this
fact, nearly all ancient and most modern Dogmatists have in effect abstract schema.
followed Aristotle. That is to say, they have attempted either to
reject the hypothetical mode or to find a via media between bare 7 See in Sextus PH n 84; M vn 216; 1x 2; M m 12 (above, p. 106); v 50.
I22 I2J
I say 'almost all' rather than 'all'; for 'coherence' theories of However that may be, the foundationalist schema certainly
knowledge or justification are normally taken - by their pro embraces either all or almost all Dogmatic epistemologies. In
ponents and by their opponents - to entail a denial of foun introducing the schema here, I am not imposing a modern concep
dationalism. A coherence theory, it is said, recognizes no basic tion on the ancient texts. On the contrary, the ancients were aware
items of knowledge: every item owes its epistemic status to its of the general structure of the schema. The earliest clear example of
relationship to other items. And in that case, coherence theorists, as it is to be found in Plato's Theaetetus (20rn-202c). More to the
I hinted in an earlier chapter, must implicitly reject the reciprocal present point, it is this schema which determines the general
mode. strategy of Sextus' attack on Dogmatic epistemology in PH II and
But there is, I think, another way of construing the strategy of M VII-VIII. The distinction between the two classes of items in the
coherentism: coherence theories can be interpreted as special cases schema is represented in Sextus by the ancient distinction between
of foundationalism. Nothing much turns on this odd claim; but it on the one hand signs (cr1iµEia) and proofs (cmo5Ei�EIS) and on the
does, I think, permit us to consider versions of coherentism which other criteria (Kpm']p1a). For we may say- with only slight simplifi
would otherwise seem to be ruled out of court. What I have in mind cation -that items are known by signs or proofs just in case they are
is this. A coherence theorist could accept the foundationalist items of dependent or derivative knowledge, and that items are
schema; and he could use the notion of coherence not to explain known by a criterion just in case they are basic or fundamental
non-basic knowledge, but to determine the basic category 13. He items of knowledge.' And in PH II and M VII-VIII Sextus attacks in
might perhaps say something like this: turn the possibility of providing a criterion of truth and the
P belongs to f3 just in case the degree of coherence within � is possibility of providing a theory of signs.
greater if it includes P than if it excludes P.
Since the Dogmatists think that what is evident is known
directly by way of some criterion and that what is unclear is
Thus we know P by virtue of coherence insofar as P satisfies clause
tracked down by way of signs and proofs by inference frotn
(ii) of the particular coherentist version of the foundationalist
what is evident, let us consider in order, first, whether there is
schema. What of clause (i) of the schema? A cohcrentist may hold any criterion for those items which directly impress the senses
and perhaps all actual coherentists have in effect held - that clause or the intellect, and then whether there is any way of produc
(i) determines an empty set: all knowledge is 'basic' knowledge. ing signs or proofs of what is unclear. (M VII 25)
(The schema remains apt. Its disjunctive condition will be verified
only by the verification of the second disjunct - the first disjunct In other words, Sextus proposes to tackle in turn the possibility of
will be idle. But that is no objection.) Yet, so far as I can see, a basic knowledge and the possibility of derivative knowledge.
coherentist need not hold that clause (i) determines an empty set. Here we are concerned with basic or fundamental items, and
For, while maintaining that coherence is what fixes our basic with the criterion of truth. Sextus' attack on the criterion is a
beliefs, he could well allow that some beliefs get their justification characteristic mish-mash of arguments: some of them are painfully
.
in a derivative fashion. (In terms of the model or analogy in
Chapter 3: there may be some flag-poles to which no rope is tied by
8 On the concept of the criterion see especially Gisela Striker,
its red end.) Whether or not this is possible will depend on the Kp1Ti)p1ov Tt'js 'Ai\T]8E(a:s (GOttingen, 1974); and, most recently, the
precise form which the coherence theory takes. But in principle it essays collected in Pamela Huby and Gordon Neal (edd.), The
Criterion of Truth (Liverpool, 1989). Both historically and
seems that one could develop a coherentism which included non philosophically the matter is far more complex than my simple
basic items. If that is true, then my suggestion that coherentism is a remarks here and later may snggest, but in the present context the
special case of foundationalism is necessary. And if it is not true, complexities arc unimportant. (I should perhaps note that the ancient
use of the word 'criterion' is quite different from the Wittgensteinian
my proposal is pointless - but harmlessly so. use with which modern philosophers are more familiar.)
u4 12 5
,
The scepticJs net The sceptic's net
feeble, some seem merely frivolous, a few are profound and chal Dogmatists are not at all prepared to accept this suggestion. And
lenging.' I touched on one of these arguments in the first chapter. we can now see exactly how they will reject it: they will say that
Here, rather than trace the discussion under Sextus' guidance, I they are entitled to hypothesize P (rather than P*) precisely because
shall follow a different path. P (and not P*) belongs to 13. If P belongs to 13, then P* does not
belong to 13. For whatever the composition of 13, it will contain only
The question concerns basic beliefs, members of the class 13. true propositions. Hence if P is in � then neither P'� nor any other
'
Different epistemologists have determined membership of this 'opposite' of P can be in 13. Thus the Dogmatists do have a reason
class in different ways: perhaps it contains self-explanatory propo for advancing P rather than P*. They advance P because P is in 13.
sitions, or self-evident propositions, or propositions which are They do not advance P*, because P* is not in 13. When they
directly given in perception, or records of sense-data, or analytical hypothesize P, their claim may or may not still count as a 'bare
truths, or innate beliefs, or propositions which we are naturally assertion' (they do not really care); but it is certainly not arbitrarily
determined to accept, or propositions which form the structure and made or unwarrantably advanced - its warranty is found in its
framework of our human or social 'form of life'. And so on. membership of 13.
Different epistemologies offer different sorts of basic beliefs, and But to this ready reply by the Dogmatists, the Pyrrhonists once
these differences import different particular questions and differ again have a ready response. For the Dogmatists surely now appear
ent particular problems. But from the Pyrrhonian point of view, to be offering an argument in favour of P. They appear to be
they all face one common and daunting objection: how can it be justifying their claim that P by reference to the fact that P is a
that membership of � whatever may determine it, is enough to member of 13. They seem, in other words, to be producing the
'
generate knowledge or to justify belief? following little piece of reasoning.
For the Dogmatist, of whatever particular persuasion, affirms
that P, and takes his affirmation to be legitimate insofar as P is a (r) P is in �
(2) Therefore, P
member of 13. And the sceptic retorts as follows: 'You affirm that P.
But you offer nothing in support of P - your claim is an hypothesis For example:
or bare assertion. Why then should I accept it? Or come to that,
{r*) It is directly given in perception that honey is sweet
why should you accept it? Since P is putatively basic, you may not
(2'') Therefore, honey is sweet
offer any argument or reason in its support. But if you offer - i f you
can offer - nothing in its support, then it is surely worthless. Or - as an Epicurean might have argued. Now this argument may be a
rather, it is worth exactly as much and as little as any other perfectly decent argument in itself. But - or so the Pyrrhonist will
incompatible claim.' (This line of reflection is of course no more evidently urge - it is not an argument which the Dogmatists may
than a rehearsal of the hypothetical mode.) use at this point in their reasoning. For it derives P from some other
Now it may seem quite obvious that the Dogmatists have an easy proposition, viz. t�e proposition that P is in �, and insofar as it does
reply to this sceptical retort. For, as I said in the previous chapter, so, it no longer takes P itself as a basic proposition.
they claim to be able to choose among putative hypotheses: they do The sceptic is not arguing that the Dogmatists' position is
not offer any proposition as a suitable hypothesis. The sceptics incoherent or paradoxical. He is not arguing, for example, that if P
urge that the Dogmatists might as well offer P* as P. But the is derived from proposition (r), then P cannot itself be a member of
13 because in that case the proposition that P is a memberof 13 will be
9 See A.A, Long, 'Sextus Empiricus on the Criterion of Truth', Bulletin
of the Institute of Classical Studies 25, 1978, 35-49; Jacques a member of 13. That might sound neatly paradoxical- but it would
Brunschwig, 'Sextus Empiricus on the kriterion: the Skcptic as be an unreal paradox. The sceptic's point is simpler and sounder
Conceptual Legatee', in J.M. Dillon and A.A. Long (edd.), The
Question of 'Eclecticism' (Berkeley, 1988). than that: if P is derived from anything at all, and hence in
126 127
particular, if it is derived from the proposition that P is in f3, then it For the Dogmatists averred that they were justified in claiming that
is not a member of p. we sweat because it is a basic truth that we sweat - and the
Of course, if a Dogmatist uses a piece of reasoning of this sort, if, P)rrrhonists took this to amount to the suggestion that the Dogma
say, he offers ( r* ) in support of (2''), then the sceptic can no longer tists argue for their basic belief from the premiss that it is a basic
object, by the hypothetical mode, that his claim - the claim, say, belief.
that (2y') is true - is a bare assertion. But this does not constitute an But since, according to the Pyrrhonists, argument (B) is on a par
answer to the sceptic: it merely displaces the sceptic's challenge. with argument (A), the Dogmatist will now be abandoning any
For - and evidently - he will now turn his attention to (r* ) . Is that claim to basic knowledge: just as argument (A) makes the invisible
proposition supported or not? If it is supported, then reciprocity pores an item of derivative knowledge, so argument (B) in effect
and regression threaten and the Dogmatist has in effect abandoned treats our sweating as a derivative and not a basic item of
his foundationalism, at least pro tempore. But if ( r * ) is not sup knowledge.
ported, then why should we accept it - and why should the Now it is plain that argument (B) is not in all respects like
Dogmatist accept it - rather than an 'opposite' proposition, say the argument (A), and it is reasonable to wonder whether the differ
proposition that P* is in p? In short, if the Dogmatist protects (2 * ) ences may not be to the advantage of the Dogmatist. In particular,
from the hypothetical mode by the shield of (r"), he leaves ( r * ) to in (A) the premiss gives evidence for the conclusion - our sweating
face the sceptic's attack, unarmed and armourless. Nothing what is, as the Dogmatists said, a sign of the existence of pores. In (B), on
soever has been accomplished. the other hand, the premiss is not evidential in this way: that
Is the Pyrrhonist now triumphant? No. There is a further step in sweating is immediately observed is not evidence- it is not a sign -
the dialectical dance. It is relatively subtle. that sweating takes place. Premiss and conclusion are differently
Let us take a standard Dogmatic argument. Dogmatists believed related in (A) and in (B).
that there are invisible pores in our skin, and their reason for The point may emerge more clearly if we adduce a third argu
believing so was that we sweat. 10 According to them, we have ment, namely:
derivative knowledge that there are invisible pores in the skin and
(let us suppose) direct knowledge that we sweat. Thus they claim to (C) The existence of invisible pores can be proved from the
fact that we sweat
warrant their belief that there are pores by way of the following
Therefore, there are invisible pores
argument:
(A) We sweat Argument (C) has the same conclusion as argument (A); but in
Therefore, there are invisible pores in our skin other respects it is parallel to argument (B). For in (C), as in (B), the
premiss does not express evidence or a sign for the conclusion - the
And they claim that we sweat without offering any reason or epistemological relation is a different one. Argument (C) is valid;
argument. and {let us grant) _its premiss is true. But it is a different sort of
The last Pyrrhonist manoeuvre in effect forced the Dogmatists to argument from argument (A). And it is argument (A), not argu
argue for the claim that we sweat, and to argue for it as follows. ment (C), on which the Dogmatist bases his claim that there are
(B) It is immediately observable that we sweat invisible pores.
Therefore, we sweat Argument (B), which the Pyrrhonists foisted upon the Dogma
tists, is parallel to (C) and not to (A). This is clear. And it gives us
IO See PH II 98, r40, 142; M VIII 306, 309; cf. Diogenes Laertius, IX 89 the means to produce a Dogmatic reply to the sceptic's latest
(quoted above, p.59), and note that the first of the three 'hypotheses'
of the physician Asclepiades (above, p.96) affirms the existence of objection.
imperceptible pores (M rn 5). The reply runs like this. If P is a non-basic justified belief, then
128 129
there will be an argument like argument (A) associated with it; and epistemic attitude to the conclusion.) He believes the conclusion;
this argument will be precisely what gives support to the belief or and because the premiss holds he believes the conclusion. But it is
justifies the claim that P. In addition, there will be an argument not the case that he believes that the conclusion holds because he
parallel to argument (C). This argument, too, will be sound; but it believes that the premiss holds.
is an argument quite distinct from argument (A), and it does not Tire crucial difference between (A) and (B) is this. In the case of
provide support for the belief or justify the claim that P. Now (A) we may say:
suppose that P is a basic belief. Here there will be no argument at all (a) Because x believes that Ri. x believes that P.
corresponding to (A): there cannot be, for if there were, P would
not be a basic belief. Since there is no argument for P corresponding In the case of (B), we may say:
to argument (A), the claim is indeed unsupported, and to that ([3) Because Rt. x believes that P.
extent and in that sense it is a bare assertion or a naked hypothesis.
But there is an argument - a sound argument - associated with P, Now the fact that in argument (B) the Dogmatist does not believe
namely an argument parallel to argument (B). Since (B) does not the premiss -is not committed to believing the premiss- shows that
correspond to (A), the existence of (B) does not show that P is after in this case there is nothing corresponding to (a}; and hence that his
all supported and so non-basic. But since (B) exists, and corre belief in P is indeed properly manifested in a bare assertion or an
sponds to (C), its existence does show something about P: in fact, it hypothesis. (For what it is for a belief to be barely asserted is
shows that we are justified or warranted in claiming that P, and precisely for there to be no proposition of the form (a) to back it
that we are justified or warranted precisely because (B) is a good up.) But the fact that in argument (B) proposition ([3) does hold
argument. (Note how argument (B) distinguishes P from any of its shows that the belief in P is grounded and justified. (For a belief is
justified provided that some appropriate truth of the form ([3) is
'opposites'. No argument similar to (B) will be available for the
associated with it.)
<opposites' of P; for any such argument would be unsound
In offering this account of argument (B), I am ascribing to the
inasmuch as any opposite of P would be false.)
Dogmatists what modern philosophers have called an 'externalist'
Thus the Dogmatists may, after all, welcome argument (B). It
account of basic beliefs and basic knowledge. The sense of the
does indeed enable them to escape from the hypothetical mode. But
nomenclature is this. Beliefs which are justified by virtue of propo
it does not mean that P loses its status as a basic belief.
sition (a) are justified by reference to some further belief (or, more
Now that may sound very clever, but it may also have the
generally, some further attitude) of the believer: x's belief in P is
appearance of a conjuring trick. For is the Dogmatist not claiming
that (B) somehow both supports and warrants and also does not grounded on x's belief in R,. Hence the justification is 'internal' to
support or make derivative his belief that P? How can he explain x. Beliefs which are justified by virtue of proposition ([3) are
the role of argument (B) without actually turning it into a support justified without reference to any further attitude of the believer:
ing argument of the objectionable sort? x's belief in P is grounded, not on any attitude of x towards Rh but
Well, argument (A) represents the Dogmatist's own reasoning. on the fact of R1.' Hence the justification is <external' to x. Thus
He believes the premiss of (A); and because he believes the premiss basic beliefs are grounded, but they are grounded on something
he believes the conclusion. If the premiss is R1 and the conclusion is 'external' to the believer. They are not grounded internally on
P, then because the Dogmatist believes that R1, he believes that P; other items of his belief system.
and his belief in R, justifies his belief in P. Now this is not so in the
case of argument (B), or in the case of (C). Here the Dogmatist does The last few pages have been abstract and non-historical. I shall
not believe the premiss. (Or rather, he need not believe the premiss end this chapter, and my discussion of the structure of Agrippan
- his epistemic attitude to the premiss does not determine his scepticism, by returning to the ancient sources. First I shall say
130 IJI
The sceptic�s net The sceptic's net
something to indicate that it is historically appropriate to ascribe are Galen's examples. The potatoes are not. The comparison with
an 'externalist' form of foundationalism to the Greek Dogmatists. artificial criteria is a commonplace in the ancient texts.)
Then I shall ask whether the Pyrrhonists had any reply to the What are these natural criteria? Galen's answer is unsurprising.
externalist response.
I start, obliquely, from the notion of nature and natural belief. I say that you all possess natural criteria, and in saying this I
'All men by nature desire to know', said Aristotle in a celebrated
am reminding you, not teaching or proving or asserting on my
own authority. And what are these criteria? Eyes in their
bon mot (Met 98oar). It is echoed by Sextus: 'Man is by nature a
natural state seeing visible things, ears in their natural state
truth-loving animal' (M vn 27). And gentle nature would not have hearing audible things, a tongue tasting tastes, a nose for
given us a desire without also supplying the means to satisfy it. smells, the whole skin for tangible things; and in addition to
'Seek and you will find, for you possess from nature dispositions these, judgement or thought or whatever you like to call it, by
towards the discovery of truth' (Epictetus, diss I iv 5r). Or, as Galen which we discern what follows and what conflicts, and other
put it, 'all men possess by nature certain first principles of reason things of the same sort - division and collection, similarity
(AoyiKcxi apxcxl)' (Thrasyb v 846-7 K). and dissimilarity. (PHP V 723 K)
It is in Galen, in fact, that we find this theory of 'natural
And a little later.
epistemology', as I may call it, most fully expressed. I cite three
passages.11 The first comes from his essay against the Academic How does Hippocrates say that the nature of things is discov
sceptic, Favorinus. ered? By starting from what is greatest and easiest: greatest in
It is plainly apparent to us, even if the sophists have done their utility, easiest for us to know. For nature has given us two
best to make it dubious, that there are natural criteria. A pair things: the criteria themselves, and untaught trust in them.
of compasses describes a circle; a rule determines lengths, just Now the criteria themselves are the sense-organs and the
as a balance determines weights. These criteria men have faculties which use these organs; and an untaught and natural
constructed for themselves, starting from their natural or trust in them is found not only in men but in the other animals
gans and criteria, than which we possess no higher criterion too. (PHP v 725 K)
more noble or more honourable. It is here, then, that we must Place an apple before me, and my eyes will report - by nature - that
start. For our mind tells us that it is possible to trust or to there is an apple before me. If my eyes report that there is an apple
distrust natural criteria, but that it is not possible to judge
before me, then- again by nature - I will trust and believe that there
them by way of anything else. For how could that by which
everything else is judged be itself judged by something else? is an apple before me. My senses may of course err; and as the
(opt doct I 48-9 K) passage from opt doct shows, I may perversely choose to reject the
evidence of my senses. None the less, if my senses are in their
We construct artificial criteria of truth, artificial yardsticks for natural condition and if my mind is in its natural state, then I will
determining the shapes and si2es and weights of things. And we do believe that there.is an apple before me just when- and just because
so on the basis of our 'natural' criteria. For we possess natural - there is an apple before me.
criteria of truth. Just as a balance will determine, artificially, that a This natural epistemology was not an invention of Galen's.
sack of potatoes weighs 56 pounds, so our organs will determine, Indeed, Galen himself claims to find it in Hippocrates (see PHP V
naturally, that this is an apple and that a fig. (The apples and figs 724 K) and in Plato (see PHP v 732 K). The theory may be
discovered in Aristotle, at least in a rudimentary form. 1 2 We may
II For a detailed discussion of Galen's views sec Michael Frede, 'On
Galen's Epistemology', in V. Nutton (ed.), Galen: Problems and 12 On Aristotle's epistemology see Jonathan Barnes, 'An Aristotelian
Prospects (London, 1981), reprinted in Frede's Essays on Ancient Way with Scepticism', in Mohan Matthen (ed.), Aristotle Today
Philosophy (Oxford, 1987), (Edmonton, 1987).
r32
care to see it in Epicurus' notion that 'Human nature was taught just as it is necessary for the balance to tip when weights are
and compelled in many different ways by the objects themselves' placed on the pan, so must the mind yield to what is evident.
(ad Hdt 75) . But it is in Stoic and Stoic-influenced texts that the (Luc xii 38)
Galenian theory is most clearly and most articulately present. I
Thanks to nature, we may get to know the nature of things.
shall cite a few passages from many, 13
Again, our 'apprehensive impressions' - the experiences in vir
'Nature', according to Epictetus, 'is the strongest thing in men'
tue of which we acquire basic beliefs about the world - form a
(diss II xx 15); and mere philosophers cannot eradicate it. The
natural class or kind of experience.
Epicureans failed to remove our natural desires to be citizens and
friends of one another, The Stoics say that if you possess an apprehensive impression
and the lazy Academics were unable to abandon or blind their you hit expertly on the underlying differences among things;
own senses, even though they made every effort to do so. for an impression of this sort has a special character distin
(diss 11 xx 20)
guishing it from other impressions, just as horned snakes
differ from other snakes. (M vn 252)
In general, we naturally trust our senses, and the sceptical argu
ments of the Academics - Epictetus might have added the Just as horned snakes are a special class of snake, marked off from
Pyrrhonians - can do nothing to destroy this natural trust. For the other snakes by a natural and idiosyncratic feature, so apprehen
Stoics, as for Galen, it is sense-perception which ultimately pro sive impressions are a special class of impressions, marked off by
vides the criterion. For, as Sextus reports, nature from other and unreliable impressions, from dreams and
delusions and the like.
nature has given us our perceptual faculty and the impres Moreover, there are some particular things which nature deter
sions which arise through it as, so to speak, a light for the mines us to believe. According to Epictetus,
recognition of truth, and it would be absurd to reject such a
faculty and to deprive ourselves of, as it were, the light. it is the nature of the mind to assent to truths, to dissent from
(M Vil 259) falsities, to suspend judgement with regard to what is unclear.
And a little earlier, describing the views of the 'the more recent'
'What's the evidence for that?' Feel now, if you can, that it is
Stoics, Sextus remarks that
night. 'Impossible,' Reject the feeling that it's day. 'Imposs
ible.' Feel or reject the feeling that the stars are even in
the criterion of truth is apprehensive impression - not with number. 'Impossible.' (diss I xxviii 2-3)
out qualification, but when there is no obstacle. For this, they
say, is evident and striking, and it all but grasps our hair and If you go out in the midday sun, you will, by nature, assent to the
drags us to assent. (M vu 257) thought that it is day. You are naturally incapable either of denying
that it is day or even of suspending judgement on the matter.
Thus we have a natural faculty for discerning the truth, and we
Again, it is by a natural path that we arrive at knowledge of the
naturally yield to the promptings of the faculty. Nature illuminates
first principles of science and at basic or fundamental beliefs. For
the world for us. Nature grabs us and drags us to assent. Using a
example, the basic principle of Stoic ethics, that all animals love
comparison which has already been invoked, Cicero says that
themselves,
IJ For a full discussion sec Michael Frede, 'Stoics and Sceptics on Clear does not admit of doubt; for it is fixed in our very nature and
and Distinct Impressions', in M.F. Burnycat (ed.), The Skeptical
Tradition (Berkeley, 1983), reprinted in Frede's Essays on Ancient
is grasped so firmly by each man's senses that if anyone tries
Philosophy. to speak against it he is not heard. (Cicero, fin v x 27)
134 135
We believe the first principles or O:pxat because nature requires us epistemologies. I do not think that any ancient text advances an
to believe them. Our beliefs are naturally determined - and by a externalist theory expressly and self-consciously - the distinction
benevolent nature. between internalist and externalist theories was not explicitly
There are many difficulties and obscurities in these several texts. marked by the Greeks. But for all that, it is most plausible to
Moreover, it should not be assumed that they all purvey exactly the construe the Dogmatists as offering externalist theories, i.e. as
same epistemological theory. Thus according to Epictetus, we offering theories which in fact are externalist. Thus I take the Stoics
must assent to certain propositions; according to the 'more recent' to be saying that we are justified in believing that every animal
Stoics, certain truths almost drag us to assent; according to Galen, loves itself simply because this belief is impressed on us by nature:
we are naturally disposed to assent but can always choose not to. the belief is justified not because we believe that it is impressed on
These are differences which it would be foolish to play down. But us by nature (though we may well come to believe this), but because
there is, none the less, a solid mass of common thought in all the it is impressed on us by nature. In general, the Stoics hold that x is
texts. And at the centre lies the concept of nature. justified in believing that P (where P is a basic belief) provided that
Reference to nature is reference to natural or causal connexions. (13')) Because P is impressed on x by nature, x believes that P.
In saying that we are naturally inclined or dragged or compelled to
assent, the Dogmatists are adverting to a causal link between the They do not hold that, in addition, it must be true that:
facts and our beliefs - certain states of affairs are such that once (a'' ) Because x
believes that P is impressed on x by nature, x
they come to our attention, we are thereby caused to believ� that believes that P.
they obtain. This is so, in particular, in the case of the first prin
Sentences (a") and (13"), as their labels suggest, are specifications of
ciples of science and of the fundamental items in our belief systems.
the sentences (a) and (13) which appeared a few pages ago. It was
And thus nature, or causal connectedness, provides the founda
characteristic of an externalist epistemology that it maintained a
tions of our knowledge.
(13)-sentence rather than an (a)-sentence. Hence the Stoics, main
On this view, basic beliefs - the members of class 13 - are what
taining (13") rather than (a"), are marked as externalists.
may be termed 'natural' beliefs; and natural beliefs are those beliefs
towards which nature leads us, i.e. those beliefs which we are
Had the Pyrrhonists any answer to externalist epistemologies? It
naturally caused to have by the very facts which the beliefs express.
might well be supposed that they would reject outright any attempt
More precisely, we might offer something like this as an account of
to intrude nature into epistemology. Yet the supposition would be
natural belief:
false. The Pyrrhonists cannot take blanket exception to any Dog
x has a natural belief that P just in case matic appeal to nature; for nature and natural inclinations played a
(i) it is the fact that P which causes it to seem to x that P' leading part in their own Pyrrhonian comedy. Thus in order to
and explain how sceptics may live without belief, Sextus invokes 'the
(ii) it is the fact that it seems to x that P which causes x to fourfold observation of life', the first part of which consists in 'the
believe that P. instruction of nature, whereby we are naturally capable of percep
(It follows trivially from this that all natural beliefs are true: if tion and thought' (PH I 24). Moreover, as Pyrrhonists we are
something is caused by the fact that P, then it is a fact that P.) The invited to 'follow the phenomena';
account needs certain qualifications and refinements; but it is, I
for we are not moved (KtvoVµeOa) in the same way at the
think, fundamentally correct as a description of the basis of Hellen present moment with regard to 'It is day' and 'It is night', or
istic Dogmatism. with regard to Socrates' being alive and being dead.
The theory of natural belief lends itself easily to externalist (M VII 391)
136 137
The passage is strikingly similar to the Epictetan text I quoted a

house, in search of truth. But it is reasonable that the man
who grasps the truth should doubt whether he has been
moment ago.
successful. (M VII 52)
These and many similar passages raise puzzles for the interpreter
of Sextus' Pyrrhonism." Whatever we make of them, they show In the darkened treasury, someone, no doubt, will have found gold;
that Sextus cannot lightly brush aside the Dogmatic reliance on the but no one can properly claim that he has done so. In the obscure
causal powers of nature. If there are Pyrrhonian objections to world, some philosopher, no doubt, has grasped the truth; but
natural epistemology, these objections must turn upon the particu none can properly claim that he has done so. 'For even if he should
lar form in which nature is invoked; they cannot consistently rest actually manage to say what is indeed the case, nevertheless he
upon a general condemnation of appeals to nature. And in point of himself does not know it.'
fact there are Sextan objections - numerous and varied, implicit Sextus cites the same lines of Xenophanes again in a later
and explicit - to the externalist epistemology of the Dogmatists. context, where he offers a different and equally apt simile:
Here I shall rehearse only the chief and most telling line of
The sceptics cleverly compare enquirers into the unclear to
argument.
archers shooting at a target in the dark. It is reasonable that
I start with a celebrated fragment from the Presocratic philo some will hit the target and others miss it; but it will not be
sopher Xenophanes of Colophon. known who has hit and who missed. Similarly, the truth is
hidden in a deep darkness. Many statements are shot at it, but
And the clear truth no man has seen nor will anyone know which of them agree with it and which disagree cannot be
concerning the gods and about all the things of which I speak; known. (M vm J25)
for even if he should actually manage to say what is indeed the
case, Some nocturnal archers hit the target; but none can properly claim
nevertheless he himself does not know it; but belief is found that he has done so. Some statements about the obscure are true;
over all. (frag. 34) but no one can properly claim that his statement is true. 'For even if
he should actually manage to say what is indeed the case, neverthe
The fragment is cited by Sextus, and was plainly a favourite text
less he himself does not know it.'
among ancient sceptics. Sextus knows that there are different ways
In neither passage does Sextus explicitly direct his similes against
of construing the last couplet; and the Pyrrhonian interpretation
an externalist account of basic knowledge. (In M vn he is merely
has not been universally accepted as an accurate reading of
concerned to interpret Xenophanes; in M VIII he is expressly
Xenophanes. But here we are concerned with what the Pyrrhonists
thinking of derivative knowledge, knowledge supposedly gained
made of the lines, not with the correctness of their interpretation.
by proofs.) But it is plain that the similes are readily adapted to that
Sextus himself offers a nice simile or story to explain
end. Sextus might have argued in the following fashion. Suppose a
Xenophanes' meaning:
Dogmatist to adhere to an externalist epistemology of the sort we
Let us imagine that some people are looking for gold in a dark may ascribe to the Stoics. Suppose him to assert that P (and to take
room full of treasures. It wiII happen that each will grasp one himself to be asserting a basic belief). Then his assertion may, of
of the things lying in the room and think he has got hold of the course, quite well be true. Moreover, P may well be a member of the
gold. But none of them will be persuaded that he has hit upon class [3. But ex hypothesi - he will not claim to know that P is a
-
the gold even if he has in fact hit upon it. In the same way, the member of the class [3; for he holds that he need not know that P is a
crowd of philosophers has come into the world, as into a vast member of f3 in order to be justified in claiming that P is true. The
whole point of the externalist theory is to allow the Dogmatist to
14 See Jonathan Barnes, 'The Beliefs of a Pyrrhonist', Elenchos 4, 1983,
5-43, at pp.32-4I. claim that P, when P is in [3, without also claiming that P is in [3. But
139
in that case (so Sextus might have urged), he cannot properly claim No; if you are to base P on W, then you must have good
to know that P. His arrow may hit the target, he may have struck reasons for preferring W to W'f.
gold- that is to say, his belief may be correct. But he may not claim
And here we now have the Dogmatist holding that
to have hit the target or struck gold or grasped a truth. For, the
darkness being thick, he cannot tell whether or not he has hit the You may properly claim to know that P if Pisa natural belief,
target or struck gold; and, nature being obscure, he cannot tell a member of the class 13; you do not also need to know that
whether or not he has hit the truth - for he cannot tell whether or P is a member of 13.
not P is in � whether or not P is in reality a basic and a natural And again the sceptic answers:
'
belief.
Now at this point it may be illuminating to look again at parts of
No; if you are to hold on to P, then you must have reason to
think that P, and not some rival P*, is a member of 13.
the earlier discussions of the four Agrippan modes. At the end of
my account of the mode of disagreement, I had the Dogmatist It is plain that these pairs of Dogmatic claims and sceptical retorts
suggesting that are all of the same type. It is plain, too, that in each case the
Dogmatist is relying on a sort of externalist theory and that the
You may properly use a yardstick Y to decide a problem ?Q,
provided that Y in fact gives the correct procedure for sceptic is pressing the inadequacy of any externalist account of
deciding ?Q; you do not also need to-, know that Y gives the knowledge or justified belief. And in this way the status of episte
correct procedure for ?Q. mological externalism can be seen to be the deep and fundamental
issue raised by Agrippan scepticism.
And the sceptic answered: It is easy to imagine that the sceptical retorts to the externalist
No; if you are to use Y for ?Q, you must have reason to believe claims all rely on one general epistemological principle. For we
that Y gives the correct procedure for ?Q. might well suppose that Sextus is here tacitly assuming that if you
know that P, you must know that you know that P; that if you are
At the end of my account of the regressive mode, the Dogmatist
justified in believing that P, then you must be justified in believing
toyed with the thought that
that you are justified in believing that P. The sceptical retorts and
You may properly claim to know that P on the basis of an the Sextan similes might seem to amount to this: an externalist is
infinite sequence of reasons L, provided that the reasons in not entitled to hold that he knows that P; therefore he is not entitled
L are in fact good reasons; you do not also need to know to hold that P. For the argument implicit in the similes might
that L is in this respect preferable to a rival sequence L''. appear to run as follows: since the Dogmatist does not know that P
is in 13, he does not know that he knows that P. And since he does
And the sceptic answered:
not know that he knows that P, he does not know that P. The
No; if you are to rest P on L, then you must have reason to second step of this argument relies on the thesis that if x knows that
prefer L to L'f. P, then x knows that x knows that P.
At the end of my account of the reciprocal mode, the sophisticated The 1nodern literature contains many sophisticated discussions
Dogmatist imagined that of the thesis that if x knows that P then x knows that x knows that
P. The thesis is known in the trade as the KK-thesis. There is an
You may properly claim to know that P if P is in fact a part of analogous thesis for justified belief: if x is justified in believing that
a coherent and true web W; you do not also need to know P, then x is justified in believing that x is justified in believing that P.
that W is better in this respect than a rival web W''.
I call this the JBJB-thesis.
And the sceptic answered: What will an externalist say to this? Well, on the externalist
140
account, x knows that P just in case because P is in � x believes that Pyrrhonian argument, the Dogmatist may know something - but
'
P. Hence x will know that he knows that P just in case x knows that he must keep mum about it.
because P is in � he believes that P. That being so, it is plain that an The Pyrrhonian point is this. Suppose our Dogmatist continues
externalist will have little time for the KK-thesis (or for the JBJB to claim that P - that this is an apple or that honey is sweet. The
thesis). For it seems perfectly clear that, in general, someone could claim is advanced as an 'hypothesis'. No reasons for it are given. It
believe in P because Q without knowing that because Q he believes is offered as a putative item of basic knowledge. The Pyrrhonist
in P. In other words, it could well be true that wonders if the Dogmatic claim is justified. He runs through the
because Q, x believes that P, arguments for the hypothetical mode, and then asks the Dogmatist:
'Well, do you think you're justified in claiming that P?' What reply
and false that will he get? Now the Dogmatist will actually be justified in claim
x believes that (because Q, x believes that P). ing that P (on the externalist hypothesis) provided that because P is
in � he believes that P. Hence if he claims that he is justified in
For it seems quite plain that I may be ignorant of the causes of my believing that P, he will in effect be claiming that because P is in � he
own beliefs. For this wholly general reason, and quite apart from believes that P. But ex hypothesi he will not make this further
any particular problems which the Pyrrhonist argument may raise, claim. For, as I said, the whole point and purpose of his externalist
an externalist will want - and have good reason - to deny that if x invocation of basic beliefs is that he may justifiably believe that P
knows that P, then x knows that x knows that P. without making the further claim that because P is in � he believes
And in fact it is clear that the KK-thesis is false. There are in P. Thus whether or not he is justified in claiming that P, he will
perfectly ordinary cases in which, without invoking any philo not respond to the sceptic's challenge by claiming that he is so
sophical theory at all, I can intelligibly say: 'Good Lord. I didn't justified.
realize I still knew that.' And unless such remarks are incoherent, What happens next? The Pyrrhonist has philanthropically
the KK-thesis is false. brought the Dogmatist's temerity to his attention (cf. PH m 280).
Thus if the Pyrrhonian attack on externalism, which I have How will-how should - the Dogmatist react? In particular, will he
derived from Sextus' two similes, depends on the KK-thesis, then or should he continue to maintain that P? If he does continue to
the Dogmatist is able to resist it. Now some sceptical arguments do, maintain that P, then he is maintaining something which he does
I think, implicitly rely on the KK-thesis. Or rather, some sceptical not believe he is justified in maintaining. It is not that he confesses,
arguments, when they are expressed informally and without logi under pressure from the Pyrrhonist, that he is not justified in
cal rigour, confuse or conflate the thought that x knows P with the believing P. It is simply that he does not believe that he is justified in
quite different thought that x knows that x knows P. Any argument believing P. He suspends judgement over the question whether he is
which makes this conflation implicitly depends upon the KK justified in believing P. And the friendly Pyrrhonist has made him
thesis; and any argument which depends on the KK-thesis is aware of all this. Now it is, of course, perfectly possible to believe P
unsatisfactory. while not believing that you are justified in believing P. But is this a
But the Pyrrhonian argument which I presented a few pages ago rational state of mind to be in? Can I rationally say: 'I think that
does not- or at any rate, need not- trade on the KK-thesis. For the honey is sweet, but I don't think I'm justified in thinking that honey
Pyrrhonian is not concerned with whether the Dogmatist knows is sweet'? If I make such a report, I am confessing to a curious state
that he knows that P. He is concerned with whether the Dogmatist of mind; I am not presenting the reasonable consequence of a
will claim that P. The archer may in fact hit the target - but he may respectable philosophical thesis.
not claim to have hit it. The gold-seeker may in fact have chanced If this last argument is right, has the sceptic refuted the Dogma
upon a nugget - but he cannot say that he has. And according to the tist? Has he shown that externalism offers no escape from the
143
The sceptic's net
Pyrrhonian net? Not exactly. The Pyrrhonist argument, if it is
sound, shows only and exactly this: if a Dogmatist is rationally to Note on the ancient authors
claim that P (where P is a basic belief), then he must also be entitled
to claim that P is a basic belief. If our Dogmatist is to remain a
foundationalist and an externalist, then he must be able to claim (1)
that P, and also (2) that P is in j3; but he must not offer (2) as his
justification for (1).
Can a Dogmatist do this? Can he, by some ingenious intellectual
twist or turn, struggle out of the Pyrrhonist net? There is a
8ta<pwvia here: the sceptic doubts that there is any escape from the
sceptical net; the Dogmatist supposes that there must be some way
out, if only he could find it. Is the sceptic right? Or is the Dogmatist In the course of this book I have cited, or referred to, various dead
right? I do not know - µexp• vOv I suspend judgement on the issue. thinkers, some of whose names are no longer familiar, The main function
of this Note is to append a short sentence of information to each name. In
And so I end these chapters on a sceptical note.
addition, it has seemed sensible to explain the abbreviations by which I
But this last 01a<pu:ivkx is a dispute between a Dogmatist and a
have referred to the ancient texts.
sceptic - it is what I earlier called a disagreement in attitude. In that Sextus and Galen are taken separately. Then, the other authors whom I
case, as I argued in the first chapter, I can after all resolve the have quoted. And finally, those authors who have been mentioned but not
disagreement. And I must resolve it in favour of the sceptic. cited.
And so I end these chapters on a sceptical note.
(r) S E X T U S E M P I R I CUS:
For Sextus and his works see the Introduction, pp. vii-viii and notes 3-4.
The two abbreviations are:
PH Outlines of Pyrrhonism
M Against the Mathematicians
Here, and throughout this Note, I offer an English version of the full titles.
For the abbreviations I have followed the normal convention of employ
ing a Latin title. Hence the mismatch between the full and the abbreviated
versions.
All translations of Sextus, and of every other ancient text, are my own. I
have tried to use the latest editions of the Greek texts. For Sextus, the
standard edition is the Teubner t�xt: PH, edd. H. Mutschmann and J.
Mau (Leipzig, r958.z); M VII-XI, ed. H. Mutschmann (Leipzig, 1914); M
1-vr, ed. J. Mau (Leipzig, 1954).
(2) G A LEN:
Born in Pergamum in 129 AD, much travelled, a considerable part of his
career spent in Rome and in the highest circles of society. Died early in the
3rd century AD. The most successful doctor and the leading medical
scientist of his day; a trained and sophisticated philosopher; a scholar and
144 145
Note on the ancient authors Note on the ancient authors
polymath; and in addition, a voluminous author, many of whose works And from Ammonius' school:
have survived. (See e.g. Vivian Nutton, From Democedes to Harvey
in APr Commentary on Aristotle's Prior Analytics
(London, 1988), ch.1-3.) I refer to:
adv Jui Against ]ulianus Saccas: 2nd/ 3rd century AD, teacher of Plotinus and founder
A M M o N 1u s
alim fac Nutritive Powers of 'Neoplatonism'.

const art med On the Art of Medicine
A N A T O L I U S: 3rd century AD, bishop of Laodicea, versed in philosophy
diff puts Types of Pulse
in Hipp morb acut Commentary on Hippocrates' On Acute and mathematics.
Diseases
A R I S T OTLE of Stagira: 384-322 BC, the master of those who know.
lib prop My Own Books
loc aff Affected Places APr Prior Analytics
meth med Therapeutic Method APst Posterior Analytics
nat fac Natural Powers Met Metaphysics
opt doct The Best Form of Teaching Phys Physics
ord lib prop The Order of My Books Top Topics
PHP The Doctrines of Hippocrates and Plato
sect ingred c A E L 1 u s Aurelianus: 5th century AD, medical writer who translated or
On the Schools, for beginners
sub( emp Outline of Empiricism paraphrased S O R A N U S .
syn puts Summary on Pulses morb acut O n Acute Diseases
Thrasyb Thrasybulus
us part The Use of Parts CELSUS: 1st century AD, obscure encyclopaedist; his only surviving work:
References are by volume and page number in the edition of C.G. Kiihn
med On Medicine
(Leipzig, 1821-33) -hence the 'K'. (But sub(emp is not in Kiihn, and there
'B' stands for Max Bonnet, whose edition is most conveniently found in 106-43 BC, Roman statesman, orator and philosopher, who had
C I C E R O:
a penchant for Academic scepticism.

Karl Deichgraber, Die griechische Empirikerschule (Berlin, 1965 2).) For
most of the works cited, Kiihn's text is now superseded, and I have fin On Ends
translated from the modern editions. Leg Laws
Luc Lucullus
(3) O T H E R A U T H O RS C I TED:
They are listed in alphabetical order. Their relative importance is not to be C L E O M E D E S: 4th century AD (?), author of a Stoic introduction to
inferred from the number of words given to each. astronomy.
disc cycl Elementary Theory

ALEXANDER of Aphrodisias: 2nd/3rd century AD; the leading Peripatetic
philosopher of the age; influential commentator on Aristotle. (The work is commonly known as:
in APr Commentary on Aristotle's Prior Analytics mot circ On the Circular Motions of the Celestial Bodies)
A M M oN 1 us,son of Hermias: c.440--52.0 AD, pupil of p R o c L u s, eminent DA v 1D the Invincible: 6th/7th century AD, Armenian philosopher of the
Platonist of the Alexandrian school; commentator on Aristotle. school of O L Y.M P I O D O R U S.
in lnt Commentary on Aristotle's On Interpretation in Porph isag Commentary on Porphyry's Introduction

in Porph isag Commentary on Porphyry's Introduction proleg Prolegomena to Philosophy
1 47
Note on the ancient authors Note on the ancient authors
DIOGENES LAERTIUS: 3rd century AD (?); compiler of Lives of the M A R C E LL I N U S : 5th/6th century AD (?), rhetorician, commentator on
Philosophers in ten books. The Life of Pyrrho, in Book IX, includes a H E R M O G E N E S.
summary account of the later Pyrrhonian philosophy.

N E M E S I U S of Emesa: 4th/5th century AD, Christian bishop and Platonist
D I O G E N I A N U S : 2nd century AD (?), Epicurean (?) philosopher. Fragments philosopher.
of a work attacking CHRYSIPPUS survive.
nat horn On the Nature of Man
John D O X A P A T R E S of Sicily: rrth century AD, monk and commentator
O L Y MP I O D O R U S of Alexandria: 6th century AD, Platonist philosopher,
on H E R M O G E N E S .
pupil of A M M O N I U S, commentator on Plato and Aristotle.
E L I A S : 6th century AD,Platonist philosopher, pupil of O L y M P I O D O R u s , in Cat Commentary on Aristotle's Categories

commentator.
P H I L o D E M u s of Gadara:
rst century BC, Epicurean philosopher, many of
in Porph isag Commentary on Porphyry's Introduction
whose works are preserved on the Herculaneum papyri.
proleg Prolegomena to Philosophy
rhet Rhetoric
c.50-:-120 AD, ex-slave, Stoic philosopher; his conversations
E P I C T ET U S:
were transcribed by his follower, the historian Flavius Arrianus. John P H I LO PONUS: c.490-:-570 AD, Christian Platonist, pupil of
A M M ON IUS, theologian and philosopher.
diss Discourses
in APr Commentary on Aristotle's Prior Analytics
EPICURUS of Samos: 341-270 Bc, founder of the philosophical school
which taught atomism and hedonism. Maximus P L A N U D E S : c.1225-1305, eminent Byzantine scholar and
teacher, voluminous author.
ad Hdt Letter to Herodotus
P L A T O of Athens: 427-347 BC, author of the dialogues to which western
E PIPHANIUS: 4th century AD, rhetorician, commentator on philosophy is a sequence of footnotes.
H E R M O G E N E S.
PORPHYRY of Tyre: 233-305 AD, pupil and biographer of Plotinus,

EUSEBIUS of Caesarea: c.260-:-340 AD, bishop, eminent churchman and leading Platonist philosopher.
prolific author.
isag Introduction to Philosophy
PE Preparation for the Gospel
P R o c Lu s: c.410-:-48 5 AD, eminent Platonist philosopher with literary and
of Tarsus: c. 160-:-225 AD, rhetorical theorist, his works

H E RM O G E N E S scientific interests.
became objects of scholarly commentary.
in Eucl Commentary on the First Book of Euclid's Elements
id On Style
stas Issues L. Annaeus S E N ECA: 4-65 AD, statesman, bellettrist and Stoic
philosopher.
H E RO of Alexandria: rst century AD(?), mathematician and engineer.
de( Definitions apocol Apocolocyntosis
L U C I A N of Samosata: 2nd century AD, cultivated author of numerous SOPATER of Athens: 4th century AD, rhetorician, commentator on
satirical sketches and essays. HERMOGENES.
1 49
Note on the ancient authors
Julius S O R A N US of Ephesus: rst/2nd century AD, leading doctor of the
Methodic school. Index of passages
gyn Gynaecology
X E N O P H A N E S of Colophon: c.580-480 BC, peripatetic poet and hedge

philosopher. Only fragments of his work survive.
(4) OTHER AUTHORS MENTIONED:

A E N E S I D E M U S of Aegae: early rst century BC; the second founder of
Pyrrhonian scepticism.
A G R I P P A: end of rst century BC (?), Pyrrhonist. (See pp.viii-ix and 121-2.) The index catalogues those ancient texts which are quoted or paraphrased or
commented upon. Bare footnote references are not listed.
A M P H I N O M U S: 4th century BC, mathematician. Alexander Cleomedes
in APr 13.7-11 94 disc cycl 1 i 8 36-7
A P E L L A S: rst century BC/An, Pyrrhonist. (See p.viii n.5) 3I.27-32.3 58-9
Ammonius Diogenes Laertius
in lnt 2.31-2 9' viii n5, 40
I X 88
ASCLEPIADES of Bithynia: rst century BC, doctor and medical theorist. 3·I9-26 18 88-9 >6
in Porph isag 74.8-10 66 89 59, 99, 128n10
C H R Y s I P P U S of Soloi: c.280-207 BC, third head of the Stoic school and its 75.13-15 76 rn6 viii n5
76.5-10 66 Diogenianus
most influential figure. [Ammonius) frag 4 Gercke 58
in APr 67.11-15 J8
EUCLID of Alexandria: 4th/3rd century BC, leading mathematician and Anon
author of Elements. Rhet Gr vn.1 383 Walz 71-2, 72n9 Elias
Aristotle in Porph isag 58.20--8 67
APr 25a14-17 58 62.9-19 66-7
F A V O R I N U S of Arles: 1st/2nd century AD, colourful scholar and writer, 25a20--2 58 proleg 9.15-20 76
with an interest in Academic scepticism. APst A 3 120----I 9.23-5 67
72a20--4 93 Epictetus
72b8-11 47 diss 1 iv 51 >J2
HIP POCRATES of Cos: 5th century BC, doyen of Greek doctors and 72b15-18 73 vii 22-3 106-7
medical writer. 72b25-8 76 xxv 11-13 9'
72b25-J3a20 73 xxviii 2-3 '35
72b36-J 65 II XX 1 5 '34
J U L I A N U S: 2nd century AD, Methodic doctor attacked by GA LEN.
Met 98oa1 >J2 20 '34
Phys 256a16 44n6 Epicurus
NUMENIUS of Apamea: 2nd century AD, Platonist philosopher. Top 145b24-7 >8 ad Hdt 75 '34
Eusebius
Caelius Aurelianus PE I viii 14 6
PY RRHO of Elis: c.360-270 BC, founder and eponym of Pyrrhonian morb acut 11 8 19n22 IV iii 2-3 58
scepticism. Celsus XIV ii I 6
med proem 27-8 6 iii I 6
Cicero iv 16 J
ZENO of Elea: 5th century BC, inventor of celebrated paradoxes. leg 1 xx 53 2n4 xv xxxii 9 7
Luc xii 38 134-5 Jxii IJ 6
fin v x 27 135--6 '5 6
I5I
�;
Index of passages Index of passages

Galen 49 97n5 379 59 '78 27-8, 116, 117n3
adv Iul XVlllA 268 K 7, 26 320 1on10 400 H 178-9 n7
const art med 1 243-4 K 4, j III 90 436 rn8 ''5 24, 26
244-5 K 4 95 445 59 '90 rn5
di{( pu/s VIII 710 K '7 90 <63 rn4 '95 9
in Hipp morb acut xv 434-5 K 4�5 4 95-6 476 97n5 "5 rn
lib prop XIX 38 K 2n2 5 96, .t28nro IX 47 59 II 27, 125
45 K 2n2 6 95 XI 36 12n17 9 73
meth med x 469 K 4 7 96-7 PH 1 1-3 rn 17-18 '4
opt doct 1 48-9 K '32 7-17 99 3 rn 20 n5
ord lib prop x1x s 5 K 2n2 8 roe, 101 7 7, 9, IO 37 3on29
PHP v 723 K '33 9 rno rn 9, 105 45 47
724 K '33 ,,-w rno 24 '37 68 74
725 K '33 rn WO 1 16-17 62, 63 88 98n10
732 K '33 u rn2 '22 41, 42, 44, 45, 52 ,82 44
sect ingred 1 78-9 K 5-6 II-12. rno '64 16, 96, 98, 114n2 '97 59--60, 61
sub(emp 62..2--6 B 4 '2 102, 106 '65 17, 24 III 6 26n28
67.22--68.3 B 4 ,, rno ,66 39, 40, 44 22 63, 73
Thrasyb v 846-7 K '32 '4 rn9 ,68 96, 98, 101, 114 JO 'l
14-17 HO '69 59, 114, 114n2 76 45, 45n7
Hermogenes 27 9' '70 20-, '75 12n17
id I iii 69 99 59 '70-7 u4 233 25
stas m 70 VJ 10 '73 WO 280 '43
9402
[Hero] M vu 25 n5 173-4 99
de{ 138.8 96n3 27 '32 '74 rno Xenophanes
52 138-9 '77 117n3 frag 34 DK 14, 138-9
Lucian 228 '2
Hermotimus 70 40 252 '35
lcaromenippus 5 257 '34
259 '34
Nemesius 325 107-8
nat ham 70 M 36 369 27
376 9705
Philodemus 39' 137-8
rhet xxix 13-21 VII-VIII 27, 125
Philoponus VIII 16 48
in APr 243.15-24 90 26 rn8
Maximus Planudes 86 74
Rhet Gr v 295--6 Walz 7' us 20
297 Walz 70-' '20 rn8
Plato 183-4 JO
Phaedrus 2640 30029 325 '39
Theaetetus 2010-202c '25 343 w4
Porphyry 360 rn4
isag 4·4-9 66 369 96
Proclus 369-78 99
in Eucl 76.24-77,3 93-4 370 IOO, 105
202.9-II 78 37' IOO, IOI
252.5-23 90 372 '°'
372-3 rno
Seneca 373 rn6
apocol n 2 2 374 roo, 103
Sextus Empiricus 375 rn9
M I 28 15n18, 3on29 376-8 no
152 153
Index of persons
on words in -10s 17-18 on the criterion 261, 29-30, 125-6
also 69, 145--6 on Cyrenaics 10
L. Gellius Poplicola 2n4 on disagreement 8, 11-15, 16-17,
18-20, 23, 24-5, 26n28, 30, 47,
Index of persons Heraclitus (73)
Hermogenes 69--J2, 79, 148, 149
115-16, 118
on Dogmatists 104-5
Hero 148 on EiroxiJ 8�, 2on25
Hippocrates 7, 133, 150 on foundations of
knowledge 123n7
Julianus 7, 150 on geometry 90, 95-6, 106
on hypotheses 90, 91, 94n2, 95-112
Locke, John 33 on infinite regression 39, 40---2,
Lucian r , 40, 148 44-6, 47-8, 51, 56--'J, 73
on knowing that you know 138-9,
Marcellinus 149 141-3
and life l 37-8
Nemesius 36, 149 on reciprocity 5�0, 61-5, 64n4,
The index excludes modern scholars. Schools or sects are listed in the Index of Numenius 3, .r50 73-5
subjects, Casual or illustrative references are enclosed in parentheses. on CJKETITLKij 10-11
Olympiodorus 147, 148, 149 style of 8n8
Aenesidemus viii, ix n6, 40, r 50 Chrysippus (58, 65, 74-5, 82, 148), system of modes 113-14, 115,
Agrippa viii, viii n5, roo, 150 IjO Pherecydes (13) 116-19, 117n3
and Aristotle 65, 121-2 (see also: Stoics) Philodemus 3, 149 use of Dogmatic arguments 104-5
(see also: Modes) Cicero 134-5, 137 John Philoponus 90, 149 works vii-viii, 99-100, 145
Alexander of Aphrodisias 3n6, 58-9, Cleomedes 36--'J, 147 Maximus Planudes 70-1, 149 Socrates (59--60, 137)
94, 146 Plato (71), 125, 133, 149 Sopater 149
Ammonius, son of Hermias 38--9, 66, David 147 on hypotheses 92-3, 92n1, 94, 95, Julius Soranus 147> 150
76, 91, 146, 149 Dio (71-2, 76) rn9
Ammonius Saccas 36, 38, 147 Diogenes of Apollonia (13) Porphyry 66, 69, 69n6, 149 Thales
Amphinomus 78--9, 150 Diogenes Laertius 16-17, 40, 59, 99, Proclus 90, 93-4, 146, 149 Theo
Anatolius 96n3, 147 r48 Pyrrho vii, vii n1, 150
Anaximander (13) Diogcnianus 58�, 148
Anaximenes {21) John Doxapatres 69n6, 148 Wittgenstein, L. 125n8
Seneca 2, 149
Apellas viii n5, 150 Sextus Empiricus
Aristotle Elias 66�, 76, 148 on Academic scepticism lo Xeniades (14)
on circular arguments 65, 73, 75--'7 Epictetus 91, 106--'J, 132, 134, 135, on 6Ko:To:ATJ'f'lo: lO, 2on25 Xcnophanes (14), 138-9, 150
on conversion of propositions 58� 136, 148 biography of vii
foundationalism 122-3 (see also: Stoics) a copyist viii, viii n4, loo Zeno of Elea 36, 150
on hypotheses 93-5, 121, 123 Epicurus 134, 148
on infinity 37 (see also: Epicureans)
on infinite regression 44n6, 47 Epiphanius 148
on knowledge 132, 133 Euclid (40, 46, 93,) 150
and Magna Moralia (13-14, 21-2, Eusebius 6-7, 8, 12, 148
31, 34-5, 75)
method of 3-4 Favorinus 132, 150
on proof 76-7, 78�
on scepticism 1 I9-22 Galen
on words in �ToS 18 commentary on Cat 86-7
also 33, 86-7, 147 on disagreement 1-2, 2n2, 4-6, 7,
Asclepiades 96, 99, 128n10, 150 w
epistemology of
Caelius Aurelianus 19n22, 147 on hypotheses
Carroll, Lewis 53n9 on scepticism
Celsus 6, 147 style of 8n8
155
154
Index of Greek terms
i\aµj36.V€lV IOI OKETrT!K6S .10
Index of Greek terms A6yos 67, 97n9 0T6:cr1s

o-roxaoµ6s
8n8, 12
70
µ6:x11 8n8, 12, ro5 cruµqiwvia 6, 8n8
µE8o5os 97n9
µEvetv 19n22, 74n12 -LOS 17-18
µeTEwpos 2on24 -rp6rros 16-17, 59
µ€xp1 vVv 10, 19n22 (see also: Modes)
µox811p6s 58� -rVxwv 2on24
6µoi\oyia 8n8 Vrroypaqiij 67

6µoqiwv!a 8n8 Vrr68EcrlS 90--r , 93-5, 94n2, 96, 96n4,
6p1crµ6s 67 rn
Ocrov hrl -roV-r<y -43, 46, 65--6, 99, 104 (see also: Hypothesis)
&rroKelcr8w
rrCdlos 21 VTI6µv11cr1s
&511i\os 26, 26n28, 103 5e56oew 91
rro(Kti\os 2on24 VTI6crxecr1s
(see also: U nclear) 51' aiWvos 2on24
rr6i\Eµos 8n8, 12
6:5t6Kp1TOS 2on24 5i6:i\i\11Aos/5i' Cti\i\i)i\wv 58-9, 70, 72n9
rrpo- 48, 74, 74n12 qia1v6µevov 26
a'(pEO"lS 2 (see also: Reciprocity) (see also: Phenomena)
ato·e11cr1s 12015 5JOcrTao1s 8n8, 12
(see also: Priority)
rrp6s TI 66, 1 1 3 qiav-racr(a 12, 64n4
(see also: Perception) 5iaqiwvia x, 2, 3, 5, 6, 8n8, 11-15,
rrp6TEpov 74n12 qiOcris 97-8, 97n6, 97n7
ato611T6s r7-18 16n19, 21, 23-4, 116
(see also: Priority) qiwvoµaxeiv 8n8
aiTT)OlS 95, 106 (see also: Disagreement)
6:KaTai\t)lfl(a 9-1 1 , ronro, 2on25 8oyµanK6s vii, 104-5
crriµEiov 125 1.Vti\6s 97, 97n5, 97n6
6:Kpnos 98n10
(see also: Signs) (see also: Bare assertion)
O:µV6T)TOS 2on24 eis 6:Tie1pov 39
6:µcpiaj3i)T1)0"lS 8n8 EK[36:i\i\e1v 39
6:vOTI65EJKTOS 71-2, 98n10 EKTil1ITHV 39
6vE1Th<plTOS 5 , 17-20, 19n22, 20023, Evav-rlov 105
26, 98n10 Ev8o�os 3
6vfivvTos 2on24 E� E-roiµou 98010
6:VTIKE!µevov 105 ETI1Kpive1v 17, 30--2
6vTtrrapaj3oAi) 55n10 ETIJO"Ti)µri 120
(see also: Parity of reason) ETioxiJ 7, 8-11, 14, 16, 2on25, 21, 39,
6:vwµa�da 8n8 40--2, 65, 116
&TTeipos 201124, 45, 47 (see also: Suspension of judgement)
(see also: Infinite regression) EppWcrea1 109, 110
6:rr68e1�1s 38, 67, 72119, 76, 78, 9711 8, EcrTw 91
981110, 125 gTI 19n22
(see also: Proof) gqio8os 97n9
Crrropf1TtK6s 7 Eqi' Ocrov 19022, 20
6:Tiopos 59
Cnrogiaivsiv 97117
Cxpxf) 44-5, 48, s 6-J, 93, 95--6, 99, l;T]TTJTIK6S 10
III, IJ2
(see also: First principle,
Starting-point)
&arrElOTOS 12, r2n17, 201124 Ka8f]KOV 106
m'.1T68ev 94, 98n10, 101 KOTOi\T]1TTlK6S 6404
&qieap-ros 18 Ktve\cr8a1 137
Oxpl vVv 10, 19n22 Kplve1v 74
Kp11t'Jp1ov 97n9, 125
[:Hos 12, 17 (see also: Criterion)
156 157
Index of subjects
'parts' of 12 Hypothesis (Chapter 4 passim)
Index of subjects potential 24-5, 18
Principle of 20--3, z.5, 28, 31
)(argument 93, 94-5, 97-8, IOl,
rn9
psychological force of 16n19, 116 and assumptions 92, IOI
sceptics party to 14-15 and belief 22, 93, 95
spur to research 3-5, rr choice of 92-3, 95, 102., Io6-12,
sufficient for scepticism? 6-7, 7-8, I26---J
16-17, 20--3, 25-35, 43-4 and first principles 9I, 93-6, 99,
terminology for 8n8, 12, 23nz.4 IOO, 104, 109, III, 128n10
ubiquity of 17, 24-5, z.8, 103, 117 inconsistent 102, 104
and unclarity 5--6, 24, 26, 103-4 'Platonic')('Aristotelian' 92-5, 121,
undccided)(undecidable 18-20, n3
19022, 23 and proof 94-5, 97, 97n8, 98n10,
Academics vii, :z.-3, 10, I32, I34 Confirmation 59, 109-10 verbal 3, 8408 rn6
Adjectives in -TOS I7-20 Conflict 1, 6-7, 11, 11n12, 15, 105, and yardsticks 27-8, z.9-32 sorts of 90--I , 96n4
Adultery 70--I , 79, 8I Divination 58, 74-5 and truth 93, 95, 97, IOO-I, I02.
rn7
Algorithm 5I-3 (see also: Disagreement) Dogmatisms vii-viii, 34, 39, 41, 56, Hypothetical method 37, 90--1, 92.-3,
Ambiguity 17-18, 32, 90--1 59, 62, 63, 75, 88, 96, Io:z.-4, 92n1, 94, 95-6, I09, 1 11-12
Consuls 80
Any)(every 48-50 Conversion 5 8---9 I04--9> I09"-I:Z.1 II7-I8, I:Z.:z.-3, defended 109-'12, 122-3
Archers 139-40, 14:z. {see also: Logic) I:Z.5, I:Z.6(, I:Z.8-3I, 136-7, Hypothetical mode 92., 96, 97, 98,
Assumption 92, 101 Counting 48-9, 51 I40--I , I43-4 98--9, 103, I04, 107, 108, I09,
(see also: Hypothesis) 113-14, 114-15, 119-201 I2I-2,
Criterion 14-15, 26-7, 2_9-JO, 39,
Axiom 46, 93-4, III Empiric doctors 5-6, 8 126, 128, I 30, 143
62-3, 97n9, 115, 117, 125--6,
(see also: First principle, Epicureans :z.-3, 30, 58, 12.7, 134 arguments for 100-9
125n8, r32-5
Starting-point) natural)(arcificia! 132 Equipollence 9, 2.0n2.5, I08 expression of 96-8, IOI
Cyrenaics ro Ethics viii, 92., 106, I35-6, 137 and scepticism 98, 99, 108--9, 114
Balances 13:z., I34-5 Evident (see: Unclear) Hypothetical utterances 90--1, 93
Bare assertion 97, 100, 103, 107-8, Decision i 7, 30--2. Explanation
II4, II9, 12.I, I:Z.:z.-3, I:Z.6-7, (see also: Disagreement) inference to the best I34• 136, Impressions, apprehensive 64n4,
12.8, I30, I3I Definition I39n15 I34-5
(see also: Hypothesis) regresses of 39 in proofs 76, 78--9 Induction 594)0
Belief regresses in 39 Infinite regression (Chapter 2 passim)
)(outline 67
acquisition of 74-6, 83 (see also: Proof) epistemological 38--9, 44, 47-8,
reciprocal 63-4, 64n4, 66--9, 69
basic 123-4, 12.6-31, 136(, Disagreement (Chapter I passim) 'Externalism' 33-4, 56-?, 88--9, 50--7, 62-3, 98
I30--l, 136-r, 137-44 of arguments 46, 52-3, 87-8
139-40, 141-:z., 144 and fu:aTo:i\T]\j.l!a 5-6
natural 132-'7 among doctors I-2., 4-6 (see also: Justification) of criteria 39n3
systems of 80, 86(, I26, 131 among laymen 12., 17 of definitions 39n4
web of 83--9, 140 among qiavTao-(0:1 I2 Fate 58, 65, 74-5 of explanations 39n4
(see also: Justification, Knowledge, among philosophers 11 2-3, 4-5, First principle 1, 91, 93-6, 99, IOO, of proofs 38, 39n2., 41, 52-3
Rationality, Suspension of 6---J, 13, 2.7, 29-30, 117 104, I091 III, 132.o 135-6 of reasons 4I, 45-6, 47, 48, 50,
judgement) analysed 11-15 (see also: Axiom, Hypothesis, 52-'71 87-8, 120--1 , 140
awareness of 2.I-2., 34 Starting-point) of signs 39n4
Causes 6, 63, 67--9 ){debate I2-13, 30--I Flagging game 8:z.-4, &7. 1:z.4 first member of? 44-6, 48, 56-?, 73
and belief :z.1, 130--I , 136-7, 14:z. mere)(undedded 5, 17, 2.0, i.5, Formal)(informal differences 55, last member of? 46-?
)(signs 79 27-8, 29-30, I17 103-4, 107 mastery of 48-51, 51-3
Circular argument 61, 64-5, 73, 77, 'meta-disagreement' 27 Foundationalism 106, 1:z.:z.-6, ri.3n7, mode of 39, 40, 42-4, I08-9,
115, I:Z.I 128, 132 Il3-14, 115, JI7-19, 119-20,
mode of I7, 23-4, 28, 32, 43-4, 47, 120--I , 122.1 I28, 140
defended 66-73, 85 96, Io3, 114, I17-19, I22, 140, (see also: Knowledge)
large circles 64-5, 85 non-epistemological 39, 39n1, 45
'45 of numbers 37, 45, 48-50, 51-2, 80
(see also: Reciprocity) multilateral 11-12, IJ Genus, defined 66---J, 69
'Co-established inference' 70--:z. Geometry 4, 46, 78--9, 90--1 , 9i.n1, 93, reductio ad infinitum 37
necessary for scepticism? 16, II6
(see also: Logic, Rhetoric) of opinion)(attitude 13, 14-15, 23, 95-6, 100, 106, 122 and scepticism 39, 4I-4, 47, 99,
Coherence theories 84-5, 87, I:z.3-4 '44 Gold-seekers 138--9, 140, 142 108, 114
surveyable? 46-5I
Honey 127, 143 why bad? 44-57, 88
159
Index of subjects Index of subjects
'Internalism' (see: 'Externalism') Nature ) (circularity 61, 64-5, 73, 77, I15 Self-love 135-6, 137
and belief 132-7 defended 66-73, 85, 122, I24 Signs 6, 26-?, 29, 70, 79, I25, I29
Justification and causation 6, 136--'7 defined 60--1 Snakes, horned 135
)(acquisition 75-6 and criteria I32-3 in definition 63-4, 64n4, 66-9, 69 Soul 9, 36
analysed 76-7, 79-81 hidden 5-6, 140 generic 62-3, 67, 1I5 Species, defined 66-7, 69
and disagreement 21-2, 23-4, 31-2, in Pyrrhonism 137-8 mode of 59, 63, 64, 66, 108-9, Starting-point 44-6, 48, 56--'7, 73
114-15 I14-15, 1 I7-19, II<J-20, I21-2, (see also: Axiom, First principle)
'external' 32-4, 12']-30, 131, I37 Obscurum per obscurius 74 I28, 140 Stoics 2, 7, 30, 58, 6404, 71-3, 74-5,
and hypothesis 105-6, 108-9, III, Opposites loo, IOI, 105-6, 127, 128, and proof 58-9, 65, 67, 69-73 91, I06--?, I34-6, 137
126-7 •JO rejected 70--3 , 73-5, 76-?, 87-9 Suspension of judgement 6-?, I4-15,
and KK-thesis 141-2 and scepticism 59, 65-6, 99, 108, 39
objective)(subjective 7']-80 Pangolins 9 u6 'as far as that goes' 42-3, 65-6, 99
and reciprocity 59, 83-7 Parity of reason 54-5, 87-9, 105-6 in Stoics 58, 7I-3, 74-5 on part of Dogmatists 9, 15nI8,
and regression 40, 48 Perception 4, 17-18, 27, 30, 41, 74, and syllogism 59---60 '35
(see also: Infinite regression, Proof, 1I7, I33. 134, 137 (see also: Coherence theories) explained 8-IO
Rationality) Peripatetics 2, 3n6, 1z.1-3 Reductio ad absurdum 93 a Tiaeos 21
(see also: Aristotle) Relativity )(Co<a1o:Arilf!LO: 9-10, 1on10, 2on25
Knowledge Phenomena 5-6, 12, z.6, 40--1, 1 10--1 1, and definition 66----9 'up to now' l0--1 1, 19, 19n22, 20,
claims to 143-4 137-8 mode of I13 '44
dircct)(indirect 117, 121-2, 123-4, (see also: Unclear) Reputable opinions (see also: Scepticism)
125, 128-30 Philosophy viii, 96, I38-9 Rhetoric 3, 69-?3 Sweating (see: Pores)
KK-thesis 138-9, I41-3 sects in 2-3, 11-12, 30--1 Syllogisms 38, 59-60, 76
Physics viii, 39n1, 117 Scepticism (see also: Logic)
'Links with the world' 56-7, 88-9 Pores 59, 128-9, 1z.8n10 degrees of 9, 24-5, 28-9
Logic viii, 4, 33, 38, 58--9, 5 9---60, Postulate 94, 95, 102, Io6 Pyrrhonian)(Academic vii, ro Toads IT, 13
70--2, 76, 90, 92, 106-?, IIO (see also: Hypothesis) radical 9, 24, 26, 28-9
Priority 64, 74-82 and research 10--l I Unclear)(evident 5-6, 24, 26, 70--1 ,
Magna Moralia 13-I4, 21-2, 31, epistemic 45-6, 47, 48, 76-82, satirised l, 40
34-5, 75 83-4, 85-7, Ill 94, 103-4, 125
victory of 23, 24, 144 (see also: Phenomena)
Medicine 96, 99, 133 asymmetrical? 77--<J, 81-2, 84, 85 (see also: Disagreement,
disagreements in 1-2, 2n2, 2n3, 4--6 defined 79-8 I Hypothetical mode, Infinite
sects of 2, 5-6, 8 transitive? 77--9• 81-2, 84, 85 regression, Reciprocity)
and scepticism 7, 9 temporal 74-6, 82 Schools
(see also: Asclepiades, Galen) terminology for 74n12 of philosophy 2-3, 11-12, 30--I Warranty (see: Justification)
Metaphors 78, 82 of medicine 2, 5-6, 8
military 7, 12-13, 12n16, 30 Probability 22, 80 Scope distinctions 49-50 Yardsticks 26-?, 29-35, 132, I40
Method, philosophical 3-4 Proof 39, 40, 41, 71-3, 93, 103 Scripture, Holy, coherence of 6 (see also: Criterion, Proof, Signs)
Methodic doctors 2n3, 7 Aristotelian 65, 67, 73, 76-8, 78-9,
Modes of suspension 16-I7 94
of Aenesidemus 40--1 )(criterion z.6-7, 29, 125
of Agrippa 16-17, 23, 25-6, 39, 59, and justification 75-6
96, 99, 113, 1I7n3 as psychological event 74-5
systems of (Chapter 5 passim), 26, and syllogism 38, 59-60
28, 32, 42-3, 66, 98-9, 108--9 (see also: Infinite regression,
Five Modes i14 Justification)
Four Modes 114-16 Pyrrhonism vii-ix, lo, 24, 29, 35,
Three Modes 119-20 137-8
'Two' Modes 116-19, 117n3 {see also: Pyrrho, Sextus)
sources for 65, 72, lOO, 114, 1 14n2,
117n3, 120--2 Rationality 21, 22, 23-4, 3I-2, 34-5,
(see also: Disagreement, 86-?, 143-4
Hypothetical mode, Infinite Reasons (see: Justification, Proof,
regression, Reciprocity) Rationality)
Motion, paradox of 45 Reciprocity (Chapter 3 passi1n)
160 r6r

Barnes-The Toils of Scepticism

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Barnes-The Toils of Scepticism

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Barnes-The Toils of Scepticism

Uploaded by

Copyright:

Available Formats

Epistemology

EDITED DY Stephen E\'crson

The Modes of Scepticism

'A lively and accessible introduction'

CAMBRIDGE UNIVERSITY PRESS

© Cambridge University Press 1990

First published 1990

Printed in Great Britain by

British Library catnlog11i11g i11 p11l1/icatio11 data

Barnes, Jonathan, 1942- Introduction page Vtt

Greek philosophy knew two main vanet1es of scepticism, one

1For Pyrrho himself (c.360-270 nc) sec the fundamental study by

Sextus once had the status of an influential thinker, and he was

ln a dialogue written by the satirist Lucian, Menippus relates how

To which Menippus' companion ironically replies:

Faction and disagreement, conflict and dispute, were endemic

sides); rather, I shall be concerned with the arguments for and

Infinite regression Infinite regression

Thus Aristotle, too, allegedly argued from P, to P, and then back

The reciprocal mode is the fifth of the Five Modes of Agrippa.

Spengel]) . Later commentators defended him: they insisted that

surely Y can not be prior to X? And i f X is prior to Y and Y is prior

The Dogmatist replies that, on the contrary,

what is assumed without proof they call an hypothesis (or

In the four preceding chapters I have discussed four Pyrrhonian

/ System of 'Two' Modes: all that happens is that the position of

/\ on knowledge of some R;. In that case, of two things one. Either

The scepticJs net The sceptic's net

The passage is strikingly similar to the Epictetan text I quoted a

alim fac Nutritive Powers of 'Neoplatonism'.

a penchant for Academic scepticism.

disc cycl Elementary Theory

in lnt Commentary on Aristotle's On Interpretation in Porph isag Commentary on Porphyry's Introduction

summary account of the later Pyrrhonian philosophy.

E L I A S : 6th century AD,Platonist philosopher, pupil of O L y M P I O D O R u s , in Cat Commentary on Aristotle's Categories

PORPHYRY of Tyre: 233-305 AD, pupil and biographer of Plotinus,

of Tarsus: c. 160-:-225 AD, rhetorical theorist, his works

X E N O P H A N E S of Colophon: c.580-480 BC, peripatetic poet and hedge­

(4) OTHER AUTHORS MENTIONED:

Index of passages Index of passages

Index of Greek terms A6yos 67, 97n9 0T6:cr1s

6µoi\oyia 8n8 Vrroypaqiij 67

You might also like

X E N O P H A N E S of Colophon: c.580-480 BC, peripatetic poet and hedge