katalhptikh;" fantasiva"

DAVID SEDLEY

fantasa
katalhptikhv ( ) - .
:
() ajpo; uJpavrconto~
() kat aujto;
ejnapesfragismevnh

to;

uJpavrcon

ejnapomemagmevnh

kai;

() oJpoiva oujk a]n gevnoito ajpo; mh; uJpavrconto~

fantash/a
J (),

.
"apov"v ,
, ,
, ,

. , ()
fantasiva" , () .

.

Zenos definition of
phantasia kataleptike
DAVID SEDLEY

enos epistemology was probably his most radical

philosophical innovation. His critic Antiochus, who interpreted
most of Zenos philosophy as derivative from the work of the early
Academy, allowed that his theory of katalepsis represented a clean
break from the anti-empiricist stance of the Platonists (Cic. Ac. I 402). It has been very plausibly argued by others1 that even in this case
Zenos meditation on a Platonic text, the Theaetetus, did in fact play
a significant part in the development of his theory. But he may well
have valued the Theaetetus less as a statement of Platos views than
as a guide to Socrates epistemology, to which, if so, he was
advocating a return.
It is amply attested that the commonly quoted three-part
definition of phantasia kataleptike is Zenos own. An infallible or
cognitive impression is one which is
(a) ajpo; uJpavrconto~ - from what is;
(b) kat aujto; to; uJpavrcon ejnapomemagmevnh kai;
ejnapesfragismevnh - moulded and stamped in accordance with
that very thing which is;
(c) oJpoiva oujk a]n gevnoito ajpo; mh; uJpavrconto~ - of a kind which
could not arise from what is not.
I hope it is not too much of a simplification to say that the
following type of interpretation currently holds sway. Zenos theory
is not just empiricist, but actually identifies its fundamental criterion
of truth, the cognitive impression, with a kind of sense-perception.

Ioppolo (1990); Long (this volume).

136

David Sedley

Moreover, the theory of sense-perception in question is a causal

one. Our direct sense-impressions, or rather a privileged subset of
them, gain their infallible hold on the world because they are directly
caused by (ajpov - from) the external things whose impressions they
are.
There is, it is true, pretty good ancient evidence that by the time
of Chrysippus Zenos epistemology was being interpreted and
promoted along just these lines. In particular, Aetius2 attributes to
Chrysippus an account of phantasia as a perceptual experience
which reveals both itself and that which has caused it (pepoihkov~ or
kinou`n). And this causal relation between the object and the
impression is associated with the same preposition, ajpov, as features
in Zenos definition. Thus a fantastovn or impressor is the external
object which causes the impression, while a fantastikovn or
imagining is an empty attraction (diavkeno~ eJlkusmov~) which arises
in the soul from no impressor (ajp oujdeno;~ fantastou` ginovmenon).
While it is not explicit, most readers are likely to assume that ajpov is
functioning as a causal preposition.
However, reasonable as this interpretation may seem, it has
gone largely unnoticed that it confronts us with an enormous
exegetical problem in interpreting Zenos definition of
phantasia kataleptike. Let us take it that the uJ p a' r con mentioned
in that definition is indeed the external object or state-of-affairs, so
that, by the first clause, the cognitive impression must be caused by

2
Aetius IV 12.1-5: Cruvs ippo~ diafevrein ajllhvlwn fhsi; tevttara tau`ta.
fantasiva me;n ou\n ejstiv pavqo~ ejn t yuc gignovmenon, ejndeiknuvmenon
auJtov te kai; to; pepoihkov~: oi|on ejpeida;n di o[yew~ qewrou`men to; leukovn,
e[sti pavqo~ to; ejggegenhmevnon dia; th`~ oJravsew~ ejn t yuc. kai; <kata;>
tou`to to; pavqo~ eijpei`n e[comen o{ti uJpovkeitai leuko;n kinou`n hJma~: oJmoivw~
kai; dia; th`~ aJfh`~ kai; th`~ ojsfrhvsew~. ei[rhtai de; hJ fantasiva ajpo; tou`
fwtov~: kaqavper ga;r to; fw`~ auJto; deivknusi kai; ta; a[lla ta; ejn aujt
periecovmena, kai; hJ fantasiva deivknusin eJauth;n kai; to; pepoihko;~ aujthvn.
fantasto;n de; to; poiou`n th;n fantasivan: oi|on to; leuko;n kai; to; yucro;n kai;
pa`n o{ti a]n duvnhtai kinei`n th;n yuchvn, tou`t e[sti fantastovn. fantastiko;n
dev ejsti diavkeno~ eJlkusmov~, pavqo~ ejn t yuc ajp oujdeno;~ fantastou`
ginovmenon kaqavper ejpi; tou` skiamacou`nto~ kai; kenoi`~ ejpifevronto~ ta;~
cei`ra~: t ga;r fantasiva uJpovkeitaiv ti fantastovn, t de; fantastik
oujdevn. favntasma dev ejstin ejf o} eJlkovmeqa kata; to;n fantastiko;n diavkenon
eJlkusmovn: tau`ta de; givnetai ejpi; tw`n melagcolwvntwn kai; memhnovtwn.

Zenos definition of phantasia kataleptike

137

something external, and by the second it must, as we might put it,

graphically portray that external thing. So far so good. But why did
Zeno go on to stipulate, reportedly under pressure from his
Academic critic Arcesilaus (Cic. Ac. II 77-8), his third clause, that it
must also3 be an impression of a kind which could not arise from
what is not? This, according to the same line of interpretation,
would have to mean of a kind which could not be caused by a nonexistent object (or state-of-affairs).
But that cannot be what Zeno meant, for two reasons. First,
Stoicism holds both that only bodies exist, and that only bodies can
be causes. It follows trivially that nothing whatsoever can be caused
by something non-existent. Hence, far from the cognitive
impression being distinguished by its inability to be thus caused,
nothing can be caused by something non-existent. Taken at face
value, Zenos third clause adds nothing that was not obvious all
along. One might respond by insisting that the clause means that the
impression cannot fail to have an externally existing cause. But then
we might expect Zeno to have written oJpoiva oujk a]n gevnoito mh; ajpo;;
uJpavrconto~, of a kind which could not arise without coming from
something which exists externally. To have put the same idea with
his actual formulation, of a kind which could not arise ajpo; mh;
uJpavrconto~, would be decidedly odd. Could any Greek have
understood ajpo; mh; uJpavrconto~ to mean from what is not something
external? Let us assume for the moment that they could. The
solution would still leave untouched the other major difficulty, which
is as follows.
The first clause, on the reading I have been criticising, should be
expected merely to establish the correct causal relation between the
external object and the impression, and it should be left for the
second clause to establish the complete veridicality of the
impression (as well as its clarity). In principle I could have an

By also I mean no more than that this, like the other two clauses, is true of
the phantasia kataleptike. I do not thereby mean to favour the strong reading
distinguished and criticised by Striker (1997), 266-72. I agree with her that Zeno
probably in fact intended a weak reading: that is, he presented his third clause
as merely making explicit something that was already in his view implicit in the
first two clauses. But nothing in the present paper turns on this question.

138

David Sedley

impression which is caused by something external, but which is false

(in that it misrepresents one or more of that things features), and a
fortiori not kataleptic. The Stoics were only too familiar with such
examples. Orestes in his madness saw Electra, i.e. had an impression
caused by her, but thought she was a Fury, i.e. had an impression
which did not accurately portray her as she was (SE M VII 170,
249). Likewise, all the familiarly cited optical illusions, such as the
straight oar seen in water as bent, satisfy the first clause but not the
second. They are from what is, in the causal sense, but fail to be
moulded and stamped in accordance with that very thing which is.
Many of our sources for Stoic epistemology analyse such cases in
just this way. In our fullest account, preserved by Sextus Empiricus
at M VII 248-52, the second clause of the definition is put at the
service of such an interpretation,4 even to the extent of separating it
into two halves: the impression must be (a) in accordance with its
object, and (b) in addition moulded and stamped. Here the first half
of the second clause is unambiguously singled out as capturing the
impressions veridicality. But the problem is as follows. If it is
correct to locate the impressions veridicality within the second
clause of the definition, why did Zeno add in his third clause that the
impression in question is of a kind which could not arise from what
is not? This means that the cognitive impression is of such a kind
that it not only satisfies both the first and the second clause, but,
further, could not have failed to satisfy the first clause. But how

S.E. M VII 249-51: deuvteron de; to; kai; ajpo; uJpavrconto~ ei\nai kai; kat
aujto; to; uJpavrcon: e[niai ga;r pavlin ajpo; uJpavrconto~ mevn eijs in, oujk aujto; de;
to; uJpavrcon ijndavllontai, wJ~ ejpi; tou` memhnovto~ Orevstou mikr provteron
ejdeivknumen. ei|lke me;n ga;r fantasivan ajpo; uJpavrconto~, th`~ Hlevktra~, ouj
kat aujto; de; to; uJpavrcon: mivan ga;r tw`n Erinuvw
v n uJpelavmbanen aujth;n
ei\nai, kaqo; kai; prosiou`san kai; thmelei`n aujto;n spoudavzousan
ajpwqei`tai levgwn mevqe~: miv ou\sa tw`n ejmw`n Erinuvwn j .j kai; oJ JHraklh`~
ajpo; uJpavrconto~ me;n ejkinei`to tw`n Qhbw`n, ouj kat aujto; de; to; uJpavrcon: kai;
ga;r kat aujto; to; uJpavrcon dei` givnesqai th;n katalhptikh;n fantasivan. ouj
mh;n ajlla; kai; ejnapomemagmevnhn kai; ejnapesfragismevnhn tugcavnein, i{na
pavnta tecnikw`~ ta; ijdiwvmata tw`n fantastw`n ajnamavtthtai. (wJ~ ga;r oiJ
glufei`~ pa`s i toi`~ mevresi sumbavllousi tw`n teloumevnwn, kai; o}n trovpon aiJ
dia; tw`n daktulivwn sfragi`de~ ajei; pavnta~ ejp ajkribe;~ tou;~ carakth`ra~
ejnapomavttontai t khr, ou{tw kai; oiJ katavlhyin poiouvmenoi tw`n
uJpokeimevnwn pa`s in ojfeivlousin aujtw`n toi`~ ijdiwvmasin ejpibavllein. Cf. also
D.L. VII 46.

Zenos definition of phantasia kataleptike

139

would that help? Even an impression which is such that it could not
fail to be caused by something external - i.e. could not fail to satisfy
the first clause - may be one which either does, or at least could,
misrepresent that external object or state-of-affairs, i.e. fail to satisfy
the second clause. For example, on this interpretation Zenos third
clause would be perfectly well satisfied by a waking impression
whose quality, unlike that of a dream, guarantees that I am definitely
seeing some external object, but where I misidentify that thing, or
identify it correctly but conjecturally. Such an impression, being
fallible or even false, can hardly be kataleptic on any possible
understanding of what the Stoics meant by this term.
This difficulty is problematic enough on any interpretation of the
theory, but for Chrysippus it is quite beyond the pale. He, according
to Aetius (loc. cit., n. 2), distinguishes a phantasia from a
phantastikon or imagining, defining the phantasia as having an
external cause while the phantastikon has none. Hence from his
point of view Zenos third clause, instead of providing the hallmark
of a cognitive phantasia, can do no more than guarantee that the
phantasia really is a phantasia.
Chrysippus was not stupid, and given his explicitly causal reading
of the theory he must have had some credible way of explaining
Zenos third clause. (I say explaining, because it was Chrysippus
practice not to contradict Zeno but to deal with any difficulties in
Zenos philosophy by reinterpreting his ipsissima verba; this
included, in the present context, his reinterpretation of what Zeno
must have meant by calling phantasia a printing (tuv p wsi") in the
soul, S.E. M VII 228-31.) At least two later attempts to deal with
the problem of interpreting Zenos third clause are extant, and
either or both might in principle be Chrysippean in origin.
One comes from a recently published papyrus fragment5 which
names the Stoic Antipater of Tarsus, head of the school a generation
after Chrysippus; it contains a classification of false phantasiai which
may well reflect Antipaters own work. The author classifies some

Journal of Juristic
PBerol. Inv. 16545, published by M. Szymanski,
Papyrology 20 (1990), 139-41. See now Backhouse (2000).

140

David Sedley

impressions as ajpov tino~, from something, others as oujk ajpov tino~,

not from something, and appears to mean by the latter those
hallucinatory impressions associated with dreams and insanity, ones
with no external impressor at all. It seems a good bet that this
expression is his attempt to make sense of Zenos category ajpo; mh;
uJpavrconto~: not caused by a non-existent thing, but not caused by
something (i.e. by some existent thing). That the interpretation
carried some weight in the school is suggested by D.L. VII 46, where
it seems to lie behind a small but significant rewording of the
canonical formulation: a phantasia which fails to be kataleptic
because it is hallucinatory is there described, not as ajpo; mh;
uJpavrconto", but as mh; ajpo; uJpavrconto".
Such a rewriting of Zenos terminology would, if accepted, at
best deal with my first listed difficulty - the objection that the
incapacity to be caused by something non-existent is not exclusive
to cognitive impressions but equally applicable to everything. It
would, however, leave untouched my second difficulty, namely that
even an impression of a kind which could not fail to have an external
cause might still, in Stoic terms, be non-kataleptic.
This difficulty and the need to circumvent it are presumably what
lie behind an alternative exegesis of Zenos third clause, which is
preserved in Ciceros Academica. Ciceros speaker Lucullus
sometimes shows his awareness that the third clause should in effect
specify the following: a cognitive impression is such that, not only is
it true, but neither it nor any impression exactly like it could be a
false one (e.g. Ac. II 18, 34, 42, 57; cf. SE M VII 152). And it is
perhaps in order to show how the third clause could amount to this
that Lucullus paraphrases it as follows at Ac. II 18: ... if it [a
cognitive impression] was such as Zeno defined it, ... an impression
stamped and moulded from what it was from, of a kind which could
not be from what it was not from (impressum effictumque ex eo
unde esset, quale esse non posset ex eo unde non esset).6 Thus
Zenos third clause, of a kind which could not arise ajpo; mh;

6
It is important to note that this is in part an interpretation, not a straight
translation. Cicero had at his disposal a perfectly good translation of ajpo; mh;
uJpavrconto~ as ab eo quod non est (Ac. II 77).

Zenos definition of phantasia kataleptike

141

uJpavrconto~, is being read as if it meant of a kind which could not

have been from anything other than the specific existing thing that it
is in fact from. This way of reading the clause has, in Ciceros wake,
won widespread support among modern interpreters of Stoicism.
This interpretation of the cognitive impression can be summed
up as follows. If you have an impression that the thing before you is
X (where X may be either a type or a token), that impression is a
cognitive one if and only if (i) the impression is being caused by X,
(ii) the impression accurately and graphically portrays X as X, and
(iii) neither this impression nor any impression exactly like it could
have been caused by Y, Z, or any other object apart from X.
Naturally enough, the normal propositional content for such an
impression would typically be of the form This is X - since it is the
actual recognition of X as X which one allegedly cannot be wrong
about - rather than some more complex proposition concerning X.
This is likely enough to represent Chrysippus explication of
Zenos definition, and I see nothing philosophically incoherent
about it. Note, for example, that it succeeds in shifting veridicality
back into the first clause, where it belongs if the third clause is to
make adequate sense. And it points towards the characteristic use of
the theory as we meet it in Stoic-Academic debate, in which the
successful identification of one or more individuals is indeed the
standard type of example invoked - distinguishing between
individual eggs, snakes or twins, Admetus failure to recognise his
own wife Alcestis, and so on.7 But at the same time, I cannot believe

7
See esp. SE M VII 401-10 and Cic. Ac. II 83-90, where the arguments of the
Carneadean Academy against (presumably Chrysippean) Stoicism turn on
either (a) dreams and hallunications, or (b) misidentifications of individuals,
with no obvious cases of (c) misdescriptions. No doubt ajpo; mh; uJpavrconto~ is
being assumed to amount to caused by no external object in (a), and to
caused by what is not that external object in (b). Cf. Rist (1969), 136-8.
Striker, (1997) mainly uses the Ciceronian identificatory formulation such as
could not arise from what is not that existing thing (pp. 265-70), but also such
that it could not arise from what is not so (p. 260). The words which I have
emphasised - whether taken to mean (a) from what is not as described, or
simply (b) from what is not the case - would clearly extend kataleptic
impressions beyond identificatory cases, but would bring us back to one or
other of my original difficulties. (a) would not very naturally be expressed by

142

David Sedley

that it captures what Zeno himself intended by his third clause. If he

had meant of a kind that could not be caused by any object (or stateof-affairs) other than the one in fact causing it, he could easily have
said so, but his actual choice of words, of a kind which could not
come ajpo; mh; uJpavrconto~, is not at all a plausible way of saying it.8
There is, I believe, a better alternative, to which I now turn.
It has as far as I know gone unnoticed that in Hellenistic
epistemology there is another, non-causal sense of ajpov. When
Orestes misperceived Electra as a Fury, his impression is in our
Stoic sources described as being from Electra (SE M VII 170). But
on other occasions it is treated as an outright hallucination, not a
misperception of Electra, and in that case it is described as from the
Furies (ib. VIII 67). This latter use of ajpov is clearly not causal, since
there were no Furies to do the causing. Rather it is what I would like
to call a representational use of the preposition ajpov. The
impression was one which represented the Furies. It is on this noncausal use of ajpov to mean representing that I wish to focus
attention.
Here is a further Stoic example (SE M VII 244-5):
Impressions which are true and false are like the one which
befell Orestes, in his madness, from (ajpov) Electra. In so far as
it befell him as from something existing (wJ~ ajpo; uJpavrcontov~
tino~) it was true, since Electra did exist, but in so far as it
befell him as from a Fury (wJ" ajpo; Erivnuo"), it was false, since
she was not a Fury. Another example is if someone dreaming,
from Dion who is alive (ajpo; Divwno~ zw`nto~) dreams a false
and empty attraction [yeudh` kai; diavkenon eJlkusmovn - the
Stoic technical expression for a delusion] as from one
standing beside him.

ajpo; mh; uJpavrconto~; and on (b), what would it mean for any impression to be
caused by what is not the case?
8
The best attempt to explain it that I have come across is Frede (1983/1987) p.
165 (in the reprint), In what sense could such an impression be said to have its
origin in what is not? The answer seems to be that the impression does not as a
whole have its origin in what is; part of it ... is made up by the mind and is not
due to the object... [I] t is characteristic of perceptual impressions that all their
representational features are due to the object. This still, in my view, suggests
that Zeno would have done better to write mh; ajpo; uJpavrconto~.

Zenos definition of phantasia kataleptike

143

The contrast between the two cases discussed is instructive. The

first, that of Orestes, follows the version of the story where he saw
Electra as a Fury. Hence its use of ajpov is causal, as we can see by
means of the following paraphrase. The impression was caused by
Electra. In so far as it seemed to be caused by something existing (wJ~
ajpo; uJpavrcontov~ tino~), it was a true impression. But in so far as it
seemed to be caused by a Fury (wJ~ ajpo; Erivnuo~), it was false. In this
series of locutions, ajpov functions as a causal term, and the
representational content of the impression is supplied instead by the
conjunction wJ~.
Now contrast the second case. Dion - whom, we may surmise, I
believed to be dead - is in fact alive. I dream, however, not merely
that he is alive, but also that he is standing beside me. Is my dream
impression true or false? It is both, according to the Stoic
classification, since it implies both the true proposition that Dion is
alive and the false one that he is standing beside me. The impression
that Dion is alive is expressed as from Dion who is alive. This time
the from cannot possibly be causal, since the Stoic dream-theory
differs from the Epicurean one precisely in classing dreams among
empty delusions, which, like the hallucinations of the insane, are
devoid of any external causation by their putative objects.9 Thus in
terms of its causal origins the dream impression of Dion is directly
comparable to the version of the Orestes story, which I considered
slightly earlier, where Orestes impression of the Furies was an
outright hallucination (as distinct from the version where he saw
Electra as a Fury). Here too, then, as in the case of Orestes
delusion, the preposition from in from Dion who is alive means
not that the impression is caused by the living Dion, but that it
represents the living Dion, or perhaps rather, more explicitly, that it
represents the living Dion as living.
The impression from Dion who is alive is further described as
being as from one standing beside the dreamer. The apparent
symmetry with the preceding part of the passage may mislead.

9
For the technical significance of yeudh` kai; diavkenon eJlkusmovn as a
hallucinatory impression with no external cause, see Aetius, loc cit. n. 2 above;
cf. Diog. Oen. fr. 10 Smith, S.E. M VII 241, VIII 67.

144

David Sedley

There, as we saw, ajpov indicated the causal origin of the impression,

and wJ~ ajpov added the representational content, some of it true, some
of it false. In the Dion case, however, ajpov already in itself introduces
the true representational content otherwise the true content would
fail to be mentioned at all - and the wJ~ ajpov locution adds the further,
false representational content.
This discrepancy is awkward, and must probably remain so on
any interpretation of the passage. But I think we can at least see how
it has come about. If the dream impression had merely been
described as ajpo; Divwno~, that would have allowed the possibility that
this representational content was false - Dion might have been a
purely imaginary figure, like the Furies of Orestes hallucination.
But the expression chosen, ajpo; Divwno~ zw`nto~, somehow serves to
inform us that the informational content of the impression is true,
thus leaving the further wJ~ ajpov locution to add merely the false as
if content of the impression. And if we ask how the words ajpo;
Divwno~ zw`nto~ convey the truth of the impression, the answer must
surely be that they have been formulated as a specific application of
the generic concept ajpo; uJpavrconto~. That is, an impression ajpo;
Divwno" zw`nto~ is one specific member of that class of impressions
which are ajpo; uJpavrconto".
The representational usage of ajpov, though rare, occurs too often
to be dismissed as mere carelessness. What we have to appreciate, it
seems, is that in Hellenistic Greek an impression from something
functions much as an impression of something does in (as far as I
know) most modern European languages. If we say that Orestes had
an impression of Electra (namely as a Fury), this is a causal of, but
if we describe a complete hallucination by saying that he had an
impression of a Fury, that is a representational of. Greek ajpov
seems to function with the same flexibility, sometimes varying even
from one sentence to the next. The rule appears to be as follows. If
I perceive X as Y, my impression will normally be described as
from X, in a fundamentally causal sense. If, on the other hand, I get
an impression of Y, where no attention is being paid to the direct
cause of the impression but merely to its phenomenology, then it is
perfectly acceptable to call this an impression from Y. Thus if I
hear a bell ringing, it would be normal to call my impression one
from a bell. But if I simply hear a ringing sound, my auditory

Zenos definition of phantasia kataleptike

145

impression is properly described as from something ringing,

regardless of whether the actual cause is a bell, a medical condition,
or a state of dreaming.
Here is a further example of representational ajpov, this time a
non-Stoic one. The Cyrenaics, as reported by Sextus, defend the
privacy of our sensations by making a distinction between our
having common names for sensibles and our having common pathe.
From the fact that you and I both use the name white it need not
follow that my sensory experience, that prompts me to use the word,
is the same as the one which prompts you to use the word. Thus we
read (M VII 196-7):
For everybody in common calls something white and
sweet, but they do not have something white or sweet in
common. For each person grasps his own experience, but
whether this experience arises in him and in his neighbour
from (what is) white (ajpo; leukou`), neither he himself can tell,
since he does not register his neighbours experience, nor can
the neighbour tell, since he does not register that persons
experience. And since no common experience occurs in us,
it is rash to say that what appears thus to me appears thus to
my neighbour too. For perhaps I am so constituted as to be
whitened by the object affecting me from outside, while
someone else has his sensory equipment so structured as to
be in a different condition.
If we were to insist on the causal interpretation of ajpov, the words
which I have emphasised would mean that I cannot tell whether the
external object causing the experience to which I apply the word
white, and also causing the experience to which my neighbour
applies the word white, is itself white. But that question is entirely
beside the point in this paragraph, having already been fully dealt
with earlier (ib. 191-5), with a different vocabulary for the
experiences causal relation to the external object (uJpov plus
genitive, and to; ejmpoihtikovn / poihtiko;n tou` pavqou~, but not ajpov).10
In the present context, the point made is that, despite the fact that

10
In the sequel (198), on the other hand, the causal ajpov does put in an
appearance. Once again this illustrates the ease with which Greek moves
between the two uses of this preposition.

146

David Sedley

you and I both agree verbally that (e.g.) snow is white, we have no
criterion for establishing that when we look at snow we are both
experiencing the same thing, whiteness - or, in the characteristic
Cyrenaic parlance which emerges in the last sentence, that we are
both being whitened by it. It is the unverifiable situation of our
both experiencing it as white that is conveyed by the ajpov locution.
A shared experience from what is white means a shared experience
which represents the thing as white.11
These non-causal uses of ajpov, to mean roughly representing,
may well echo a usage which had some currency in the early third
century BC, a time when Zeno was forging the Stoic theory of
cognition and when Cyrenaic epistemology was enjoying its final
phase before the schools disappearance. For surely this is the sense
of ajpov which we need to recognise in Zenos definition of phantasia
kataleptike.
Let us return to the three clauses of that definition. A cognitive
impression is, first, ajpo; uJpavrconto~. That is, on the proposed
reinterpretation, it represents what uJpavrcei. What does this mean?
As has often been noted, the verb uJpavrcein is not used in our Stoic
sources as a mere synonym for ei\nai, the being or existence
which bodies alone possess. Even incorporeal predicates are said to
uJpavrcein, merely because they are actually instantiated in
something, and the present is said to uJpavrcein, not because it is a
body, but because unlike the past and the future it is actual (SVF II
509). Thus uJpavrcein conveys the kind of actuality which can belong
not just to bodies which currently exist, but also to currently actual
predications and states-of-affairs. If a cognitive impression
represents what is actual, this may in different cases mean that it
conveys an object which actually exists (e.g. Dion), or a complete
state of affairs which actually obtains in the world, consisting of one
or more objects actualisation of specified predicates (e.g. the
complex fact that Dion is walking and Theon is sitting).
If we add the representational sense of ajpov, and thus take ajpo;
uJpavrconto~ to mean representing what is actual, it gains a far

11

For a close examination of this passage (incorporating my current suggestion

as to how ajpov is to be interpreted), see now Tsouna (1998), ch. 7.

Zenos definition of phantasia kataleptike

147

richer meaning than on the causal reading. An impression which

satisfies this description is not necessarily one caused by an external
object or state-of-affairs (although usually it will be thus caused), but
one which represents how it is, and thus already counts as true.12 We
have already encountered an example of such an impression. An
impression from Dion who is alive, although not even caused by
Dion, represents a uJpavrcon, namely the fact that Dion actually is
alive.
Why, if so, did Zenos first clause not simply specify that the
cognitive impression must be true? Because truth is, according to
Stoicism, primarily a property of propositions, one which is at best
no more than derivatively or loosely applicable to the impressions
which convey those propositions (cf. SE M VIII 10). Strictly
speaking, what an impression has, or aspires to, in its own right
should not be truth, but correspondence to reality. There is
evidence (S.E. M VII 154) that Arcesilaus in his debates with Zeno
especially insisted on this restriction of truth to propositions, and
Zenos cautious phrasing may to some extent reflect that adversarial
context.
I shall nevertheless for convenience, if a little loosely, continue
to use the term veridicality to describe this correspondence to the
way things are.
If, as I am arguing, the first clause already establishes the
cognitive impressions veridicality, its representation of the way
things are, what is added by the second clause, which should now be
translated moulded and stamped in accordance with that very thing
which is actual? We should no longer expect to locate in this clause
the impressions veridicality. Rather, it limits itself to describing the
graphic qualities with which the cognitive impressions
representation of how things are is carried out. It does not just
convey in barest outline how things are, but vividly portrays the

12
Frede (1983/1987) rightly takes the first clause already to specify truth; my
reinterpretation of ajpov may, I hope, help to show how it can, and also why
some Stoic texts nevertheless located truth either exclusively (S.E. M VII 24852, quoted in n. 4 above) or partially (the Berlin papyrus, cited in n. 5 above) in
the second clause.

148

David Sedley

thing or situation in panoramic detail. Zeno is trying to capture the

richness and absolute clarity which distinguish a totally reliable
impression from an indistinct and therefore unreliable one.
So far I have simply assumed the representational sense of ajpov,
in order to display its results in the reinterpretation of Zenos first
two clauses. But it is its application to the third clause which, I think,
confirms that it has every chance of being the meaning that Zeno
intended. We have seen already the severe difficulties that the causal
sense of ajpov generates in the interpretation of the third clause. By
contrast, the representational sense is readily and unproblematically
intelligible there. The clause of a kind which could not arise ajpo; mh;
uJpavrconto~ will, on this account, mean simply of a kind which
could not represent what is not actual. If the first clause already
stipulates that the impression should be veridical, the third adds that
it should be the kind of impression that could not fail to be veridical.
There is excellent evidence that this is exactly how Zeno was
understood in his own day. His contemporary critic Arcesilaus
reported the phantasia kataleptike to be one which was not only
true but also of a kind which could not turn out false (toiauvth oi{a
oujk a]n gevnoito yeudhv~, SE M VII 152). Or, to express this
infallibility requirement more accurately (in accordance with
Arcesilaus own insistence that truth and falsity should belong to
propositions, not to phantasiai): if I have a cognitive impression that
Dion is walking, the stipulation in the third clause is that it must be
an impression such that its representational content, namely that
Dion is walking, could not fail to correspond to an actual state-ofaffairs in which Dion is walking.
There remains an obvious difficulty for this interpretation. If
Zeno intended his ajpov to mean representing, why did he not find a
less misleading locution to express the idea? I confess that I can
think of no credible antecedent for his representational usage of the
term, and that even his own followers in the school tended to
mistake it for a causal usage. I can offer no more than a guess.
Cicero (Ac. II 76-8) seems to believe that Zenos definition
originally consisted of just the first two clauses, and that the
third was added only when he came under pressure from
Arcesilaus. Arcesilaus did not become head of the Academy until

Zenos definition of phantasia kataleptike

149

a few years before Zenos own death (somewhere in the years 2684; Zeno died in 262/1). There is no need to doubt that the debate in
question could have preceded Arcesilaus headship, perhaps by
many years, but their respective ages, with Zeno the senior by some
18 years, still make it plausible that Zenos original two-clause
formulation was in circulation for some time before he encountered
Arcesilaus and was forced to refine it by adding the third clause.
It may then well be the case that in his original formulation he
did intend the ajpov in a primarily causal sense, much as it was indeed
to be understood by later Stoics, who could no doubt call on the
evidence of his own writings. If Zeno was drawing his ideas partly
from the Theaetetus, it is very likely both that he, like Platos
Socrates, focused on perceptual cases of cognition, and that he took
the wax-impression model which he was borrowing as an obviously
causal one, according to which the form of the external object is,
more or less literally, imprinted on the soul.13
If so, we must conclude that, at any rate by the time that he was
challenged by Arcesilaus and decided to add the third clause, Zeno
had shifted to a primarily representational sense of ajpov, and that it
was this sense which in consequence he assumed in his formulation
of the third clause. Of course, even when still functioning causally,
ajpov must have already carried some representational connotations,
since the external object from which the phantasia arises was being
assumed to cause it not in just any way (e.g. in the way that the
person pushing the doorbell is the cause of my hearing it ring), but
specifically by transmitting its own perceptible properties to the
phantasia. Viewed in this light, Zenos new move was to not to
introduce these representational properties, but rather to emphasise
them at the expense of the causal ones.14

13
Actually the Theaetetus model does include purely conceptual imprints in the
wax (191d5), but neither Socrates in the dialogue nor Platos interpreters (cf.
already Alcinous, Didaskalikos 154.40-155.13 Whittaker-Louis) seem to make
much of them.
14
I exclude the alternative of letting the revised use of ajpov retain both the
causal and the representational senses. If that were so, the third clause would
mean of a kind which could not (a) be caused by and (b) represent a non-actual
thing or state of affairs, and that would leave untouched the original difficulty:

150

David Sedley

Just what might have impelled him to such a shift would then
become an urgent question. An attractive possibility is that, having
originally conceived the cognitive impression as directly and
exclusively perceptual (cf. Cic. Ac. I 40-1), he came in time to see an
indispensable role for non-perceptual cognitive impressions.
Gods existence and providence - doctrines to which Zeno
devoted many of his own arguments - are explicitly said in Diogenes
Laertius report of Stoic epistemology (VII 52) to be objects of nonsensory katalepsis. We must not jump too hastily to the assumption
that the cognitions in question are the outcome of non-perceptual
phantasiai kataleptikai. They might, it has been suggested, be
considered by the Stoics to be adequately grounded by earlier
perceptual cognitive impressions of the worlds functioning, and to
acquire their status as cognitions in that way.15 However, Zenos
recorded arguments on this theme16 can hardly be said to invite such
an analysis, and it is hard to believe that he considered their
conclusions cognitive without assigning the same status to such
salient premises as The rational is superior to the irrational, and
Nothing lacking sensation can have a sentient part17 - premises
which are not easily reducible to, or even derivable from, direct data
of sensory experience.
Again, Zeno defined a tevcnh as a system of cognitions
(katalhvyei~) unified by practice for some goal advantageous in life
(Olymp. In Gorg. 12.1), and these cognitions were themselves
standardly identified with the theorems constituting the art. Zeno
can hardly have meant to insist on the exclusively perceptual
content of the theorems. Once again there is the possible reply that
he nevertheless saw them as fully grounded by past perceptual
phantasiai kataleptikai. But we have no reason for attributing to him
so impoverished a conception of techne, especially when we bear in

thanks to the inclusion of (a), the clause would be satisfied by any phantasia
whatsoever.
15
See especially Striker (1974). Contra, see Brennan (1996) 324-5.
16
For Zenos theological and other syllogisms, see esp. Schofield (1983) and K.
Ierodiakonou (this volume).
17
SE M IX 104; Cic. ND II 22.

Zenos definition of phantasia kataleptike

151

mind that at least some virtues are technai, and that their theorems
must be or include moral principles.18 Zeno is well known to have
argued syllogistically for moral principles, and clearly did not think
that they were founded exclusively in sense-perception.
Finally, what about the cognition of fundamental laws of
thought? Consider such cognitions - ones fundamental to Stoicism as that every event has a cause, and that every magnitude is infinitely
divisible. Cognitions like these could hardly be thought either to be
caused by the facts which they record, or for that matter to be
adequately derived from a past series of directly sensory phantasiai
kataleptikai.
Later Stoics, at any rate from the time of Chrysippus, would no
doubt have sought to present these and similar intuitions as the
content of common conceptions (often treated as equivalent to
prolepseis), which came to function as an independent criterion of
truth alongside phantasia kataleptike; but the identification of these
as criteria of truth is associated explicitly with Chrysippus in our
sources,19 and I know of no evidence that would justify our tracing
that theory back to Zenos generation. When the sources attribute a
criterion of truth to Zeno and his generation, they speak exclusively
of katalepsis.20
There is, then, reason to think that Zeno assumed katalepsis to
include a variety of fundamentally non-sensory cognitions. Does it
follow that there are also, corresponding to these, non-sensory
phantasiai kataleptikai? It probably does. Although we have no
formal record of the Stoic definition of katalepsis, Arcesilaus
reported it as assent to a phantasia kataleptike (SE M VII 151-3).
Whether or not this ever became a canonical definition,21 Arcesilaus

18

Cf. SVF III 280.

D.L. VII 54; Alex Mixt. 216.14-218.6 Bruns = SVF II 473.
20
S.E. M VII 152; Cic. Ac. I 42, where katalepsis is norma scientiae, while
conceptions (which must include prolepseis) have a different, apparently
derivative, status.
21
Striker (1974) is probably wise to treat with caution the other passages where
this equivalence is attributed to the Stoics. However, if I am right, an
alternative to her way of accounting for those passages may lie in Zenos own
writings.
19

152

David Sedley

at the very least provides strong evidence as to how a contemporary

critic understood Zeno himself to be using the term. And if he is
right, it is hard to see how the phantasia kataleptike to which a nonsensory katalepsis is an assent can itself be anything other than a
non-sensory one. It is, indeed, well-established Stoic usage to speak
of a class of non-sensory phantasiai, and of phantasiai which are of
incorporeals despite the fact that they cannot be caused by those
incorporeals.22 There is no reason to doubt that these could include
kataleptic phantasiai.
If this reconstruction is right, Zeno must have come to accept
that philosophical understanding relies at least in part on nonsensory phantasiai kataleptikai. Assuming further that by this time
at least the first two clauses of his famous definition of phantasia
kataleptike were already established, perhaps even the subject of
inter-school debate, he had good reason to present that definition as
one not intended to be interpreted in a narrowly causal or
perceptual sense: the ajpov relation is one expressing accurate
representation of reality, without necessarily in every case also
implying causal derivation from that reality.
My suggestion, then, is that it was in the course of his theorys
evolution beyond the crude model offered by the Theaetetus that
Zeno found himself treating the ajpov relation less as a causal
derivation than as a representational one. In so far as he was
conscious of this semantic shift, he may be imagined as justifying it
by reflecting that the ajpov element present in his second clause,
ejnapomemagmevnhn kai; ajnapesfragismevnhn, already served to
convey this aspect of the impression: it is so moulded and stamped
in the mental wax as to represent its object accurately. Be that as it
may, a result of this new realisation was that the first clause, rather
than the second, came in Zeno's own usage to be the chief one for
conveying the cognitive impressions basic representational

22

See D.L. VII 51 on non-sensory phantasiai. They explicitly include those of

incorporeals, and S.E. M VIII 409 attests a Stoic attempt to show that these
could involve non-causal representation. Chrysippus narrower, causal
definition of phantasia (Aetius loc. cit. n. 2 above) should perhaps be regarded
as adding a specific sense of phantasia to this generic one.

Zenos definition of phantasia kataleptike

153

accuracy - its veridicality. When, consequently, the debate with

Arcesilaus finally came to a head, it was natural that Zenos newly
specified infallibility clause should borrow its materials from this
first clause.
If this suggestion is right, Zenos successors, reading his writings
as a single corpus rather than diachronically, must have failed to
recognise his drift away from a causal account of cognition.
Consequently, they were impelled to emphasise those cognitive
impressions whose derivation did indeed lie in the direct causal
action of the object upon the perceiving subject. That, at all events,
is what the Stoic theory of phantasia kataleptike eventually became.
But we should not be too confident that the theorys thoroughgoing
empiricism fully captures Zenos own mature intentions.23

________________________

BIBLIOGRAPHY
Backhouse, T. Antipater of Tarsus on false phantasiai (P Berol inv.
16545), in Studi e testi per il Corpus dei papiri filosofici greci e latini
10, Papiri filosofici, Miscellanea di Studi III (Florence 2000) 7-31.
Brennan, T. Reasonable impressions in Stoicism, Phronesis 41
(1996), 318-34.

23

The basic thesis of this paper, concerning the meaning of Zenos three
clauses, is one which I have already sketched briefly in my (1998), p. 152. In
developing it further, I have benefited from the comments of Thamer
Backhouse, Gisela Striker and Michael Frede, and from discussion with
participants at the September 1998 Larnaca conference Zeno and his Legacy
(especially Malcolm Schofield), at a Florence seminar in December 1998, and at
the Sminaire Lon Robin, Paris, in February 1999. It should not be assumed
that they all endorse my conclusions.

154

David Sedley

Frede, M. Stoics and Skeptics on clear and distinct impressions, in

M.F. Burnyeat (ed.), The Skeptical Tradition (1983) 65-93; repr. in M.
Frede, Essays in Ancient Philosophy (Minneapolis 1987), 151-76:
Ierodiakonou, K. 'Zeno's Arguments.' (this volume).
Ioppolo, A.M. Presentation and assent: a physical and cognitive
problem in early Stoicism, CQ 40 (1990), 433-49.
Long, A.A. 'Zeno's epistemology and Plato's Theaetetus'
volume).

(this

Rist, J.M. Stoic Philosophy (Cambridge 1969).

Schofield, M. The syllogisms of Zeno of Citium, Phronesis 28 (1983),
31-58.
Sedley, D. Article Stoicism in E. Craig (ed.), The Routledge
Encyclopedia of Philosophy (London 1998), vol. 9, 141-61.
Striker, G. Appendix to Krithvrion th`~ ajlhqeiva~ (original German
version Gttingen 1974; repr. in English in her Essays on Hellenistic
Epistemology and Ethics, 22-76).
Striker, G. Academics fighting Academics in B. Inwood, J. Mansfeld
(eds), Assent and Argument: Studies in Ciceros Academic Books
(Leiden 1997) 257-76.
Tsouna, V. The Epistemology of the Cyrenaic School (Cambridge
1998).