SciELO - Scientific Electronic Library Online

Home Pagelista alfabética de periódicos  

Serviços Personalizados




Links relacionados



versão impressa ISSN 0100-6045versão On-line ISSN 2317-630X

Manuscrito vol.42 no.4 Campinas out./dez. 2019  Epub 25-Nov-2019 


The aporia of ἢ ἐϰ παντὸς in Posterior Analytics II.19

1University of California, Los Angeles Department of Philosophy. Los Angeles, CA, USA.


This article sketches, and works to motivate, a controversial approach to Posterior Analytics II.19. But its primary goal is to recommend a novel solution to one particular interpretive aporia that’s especially vexed recent scholars working on Post. An. II.19. The aporia concerns how to understand the enigmatic ē ek pantos... (≈ “or from all...”) in the genealogical account of foundational knowledge at II.19 100a3-9. Our proposed solution to the aporia is discussed in connection with a number of larger philosophical issues concerning Aristotle’s theory of epistē mē.

Keywords: Aristotle; Posterior Analytics II.19; Episteme; Ancient Greek philosophy; Ancient Greek epistemology

1 Introduction

For better or worse, Post. An. II.19 is today the most widely studied chapter in the whole of Aristotle’s Posterior Analytics. The present article sketches, and works to motivate, a controversial approach to Post. An. II.19. Our primary goal, however, is to recommend a novel solution to one particular interpretive aporia that’s especially vexed recent scholars working on Post. An. II.19.

The aporia in question concerns how to understand the enigmatic ē ek pantos... (≈ “or from all...”) that we read at 100a6 when we study the genealogical account of foundational knowledge at II.19 100a3-9. The philosophical interpretation of 100a3-9 is, of course, highly controversial. And many competing readings of 100a39 in particular-and Post. An. II.19 in general-turn on how one responds to this aporia about how to interpret 100a6’s ē ek pantos. The solution proposed below doesn’t seem to have been elsewhere discussed. But it strikes me as significantly more attractive than any alternative solution to the aporia I’m aware of.

We’ll be turning to 100a3-9 and the aporia of ē ek pantos soon, in Sections 3-5 below. But I begin with some preliminary remarks about Post. An. II.19 as a unit and how 100a3-9 is embedded within it.

2 The context, structure, and philosophical goals of Post. An. II.19

Post. An. II.19 is the concluding chapter of (what we now call) the Posterior Analytics. But it must also be emphasized that Post. An. II.19 characterizes itself as the concluding chapter of the Analytics as a whole.1

The unifiying aim of Aristotle’s “two” Analytics (i.e. Prior plus Posterior) is to elaborate and defend an account of a particular type of knowledge which Aristotle calls demonstrative episteme [epistēmē apodeiktikē].2 To have episteme [epistasthai]-in the technical sense at issue in the Analytics-is to have a rationally unshakeable, perfected knowledge of why some inexorable aspect of reality is in fact as it is and cannot be otherwise (Post. An. I.2 71b9-12, cf. 72b2-3). To possess demonstrative [apodeiktikē] episteme of (some particular epistēton) X is to have episteme of this X exactly in virtue of one’s understanding-and assenting to the premises of-a demonstration that, through proving its conclusion, fully accounts for X in terms of prior causes. A demonstration [apodeixis], in this context, is much more than a valid argument (cf. Post. An. I.2 71b16-72a8): it’s a syllogistic-proof [sullogismos] of a truth T which fully explains why T is the case by deriving it from starting-points that are, by nature, (i) explanatorily primitive in themselves and (ii) explanatorily prior to T. These demonstrative starting-points, for Aristotle, are real explanatory primitives that prior causes fail to account for; in effect, they are the ultimate whys from which demonstrations demonstrate. Aristotle himself calls them the foundations [archai] of demonstrations (cf. I.2 72a7-8). Since such archai aren’t themselves fully accounted for by prior causes, Aristotle contends that demonstrative episteme of a demonstrative foundation is impossible in principle. But since demonstrative episteme of X is always based on one’s assent to a corresponding demonstration’s premises, Aristotle also insists that knowing demontrative foundations is foundational for demonstrative episteme.

Aristotle’s psychological works discuss episteme from the perspective of a theory of soul; his ethical works discuss episteme from the perspective of a theory of virtue and the human good. The distinctive strategy of Aristotle’s Analytics is to account for episteme from the perspective of a theory of logoi. I myself (see below) take this investigative approach to stand behind the 99b17 words t’auton gar esti (“for they’re the same”) in the opening sentence of Post. An. II.19. But be this as it may, it’s time for us to turn to the text of II.19 itself. The chapter’s elegent opening (99b15-19) can be translated3 as follows4:

And so, regarding syllogistic-proof and demonstration-what each of them is and how each comes to be-these things [are now] clear; also, regarding demonstrative episteme-[what it is and how it comes to be]-at the same time this [is now] clear. For, they’re the same [i.e. the latter illumination is the same as the former one].5 But regarding the foundations [archai]-(Q1) how they come to be known, and (Q2) what is the state of knowing [them]-from where we are [enteuthen] these things are6 manifest to those who have first gone through the aporiai. (Post. An. II.19 99b15-19)

The Analytics proposes to investigate demonstrative episteme by investigating demonstration and to investigate demonstration by investigating syllogistic-proof [sullogismos] (cf. Pr. An. I.1 24a10-11, I.4 25b27-31). Post. An. II.19 opens with the announcement that this investigative programme has reached a philosophically successful conclusion and proceeds to articulate a pair of questions about foundations: (Q1) and (Q2).

The ‘foundations’ [archai] whose knowledge is at issue in questions (Q1) and (Q2) are foundations of a demonstration: i.e. demonstrative starting-points in the sense rehearsed above (archai apodeixeōs). In asking how the foundations of a demonstration become ‘known’ [gnōrimoi], (Q1) is specifically asking how animals like us come to know such objects if/when we do come to know them and achieve the “state of knowing” [hē gnōrizousa hexis] mentioned in (Q2).7 To be in the “state of knowing” mentioned in (Q2) is to know the foundations of some demonstration. More specifically, as II.19 later puts it, the type of knowledge at issue in (Q2) is a knowledge of archai (i.e. archai apodeixeōs) that, qua knowledge, is itself an archē (an archē epistēmēs) because it furnishes the ultimate basis [archē] on which episteme of some X is founded (100b9-16). In the language of Post. An. I.2 (cf. esp. Top. VI.4 142a9-11), question (Q2) asks us to identify the kind of knowledge that’s prior to demonstrative episteme “by nature” (in the order of reasons and causes); question (Q1) inquires of the knowledge that’s prior to demonstrative episteme “for us” (in the temporal order of human learning).

So, the opening lines of Post. An. II.19 articulate a pair of questions; and these questions ask

Regarding the foundations F (D) of an arbitrary demonstration D:

(Q1) How do we come to know F (D)?

(Q2) What kind of knowledge is the knowledge of F (D) which grounds demonstrative episteme secured though D?

These questions’ level of generality bears emphasis. For the responses to (Q1) and (Q2) which Post. An. II.19 presents are supposed to hold for any value of D. But from Aristotle’s perspective, what D demonstrates could be a theorem of mathematics, a theorem of physics, or a theorem of metaphysics.8

Whatever else it might be, Post. An. II.19 is centrally some kind of attempt to address questions (Q1) and (Q2). The chapter’s opening lines (99b15-19) present it as such. And these same remarks manifestly imply that the Analytics’ preceding chapters do not suffice for philosophical clarity about (Q1) and (Q2). Emphasizing this latter point, Brunschwig has argued that students of the Post. An. shouldn’t suppose “que le problème de la connaissance des principes a été traité avant son ouverture officielle au début du chapitre II.19” (1981, p. 96). However, the Analytics’ preceding chapters have in fact given extensive attention to many substantive issues concerning demonstrative foundations and their epistemology. Moreover, before “officially introducing” (Q1) and (Q2), II.19’s opening sentence asserts that philosophical clarity regarding the nature and coming to be of demonstrative episteme has already been achieved (99b15-17). In connection with this assertion, I think the indicative mood of (Q1) and (Q2) is important to appreciate. As raised in Post. An. II.19, questions (Q1) and (Q2) presuppose demonstrative foundations exist and presuppose that demonstrative foundations-though impossible to demonstrate-are nonetheless capable of being known with a kind of knowledge that suffices to ground genuine episteme. I submit that Aristotle took himself to be entitled to these presuppositions in Post. An. II.19 because earlier chapters of the work argue for these claims at length.9Post. An. II.19 is not-as it’s sometimes assumed to be-an obscure and philolosophically underwhelmimg attempt to defend the claim that the indemonstrable foundations of demonstrations are knowable by a knowledge which sufficies to ground episteme.10

I further submit (NB note 6 above) that the opening lines of Post. An. II.19 shouldn’t be read as ambitiously promising a philosophically sufficient account of the complete truth concerning (Q1) and (Q2). For, 99b15-19 needn’t be be so read; moreover, it must be admitted-and Aristotle couldn’t possibly have been unaware-that II.19 leaves its reader with quite a few more philosophical puzzles about (Q1) and (Q2) than it adequately resolves. But be this as it may, what the opening of Post. An. II.19 most precisely tells us (99b18-19) is that to achieve philosophical clarity regarding how to answer (Q1) and (Q2), we must first work through certain aporiai [proaporēsasi prōton]. And this said, Aristotle immediately (99b20-30) proceeds to lay out two such aporiai:11

Now, it’s previously been said12 that having episteme through a demonstration is impossible if one doesn’t know the primitive unmediated13 foundations. But regarding the [requisite] knowledίge of the unmediated [foundations of a demonstration], one might press the aporiai [diaporēseien]:

(a2) Is it the same-or not the same-[kind of] knowledge? I mean [kai]: Is episteme [the knowledge] of each [i.e. not only what’s demonstrable but also indemonstrable foundations]? Or not?-Is there rather episteme of the one [i.e. what’s demonstrable] and another kind of knowledge of the other [i.e. indemonstrable foundations]?

(a1) Are the[se foundation-knowing] states engendered [in us] without [always already] existing within [us]? Or, [always already] existing within, do they go unnoticed? On the one hand it’d be strange if we [always already] have these [states of knowledge]-for this means our having of knowledges more exacting than demonstration isn’t noticed. On the other hand, if we acquire [these states of knowledge]-being previously without them-how could we get this knowledge and learn it from non-prior knowledge? For this is impossible (as we were also saying concerning demonstration). (Post. An. II.19 99b20-30)

The remainder of Post. An. II.19 is generally agreed to consist of two discourses: 99b30-100b5 and 100b5-17. Rounding out II.19’s stylistically pleasant double chiasmus,14 99b30-100b5 address aporia (a1) and question (Q1) while 100b5-17 addresses aporia (a2) and question (Q2).

In the philosophical practice of Aristotle, the work of discovering, developing, and solving the relevant aporiai plays a central role. An aporia, is this context, is any outstanding problem or unresolved philosophical disagreement that stands in the way of philosophical progress in the investigation at hand.15 The language (99b19, 99b23) with which Post. An. II.19 introduces (a1) and (a2) invites us to view (a1) and (a2) as aporiai in this sense-as aporiai the Analytics’ preceding chapters do not solve, as aporiai that (qua unsolved) impede us from giving philosophically adequate answers to questions (Q1) and (Q2).

As I myself read II.19, the unit 99b30-100b5 works towards progress on (Q1) by sketching (what Aristotle takes to be) the truth concerning (a1); the unit 100b5-17 works towards progress on (Q2) by sketching (what Aristotle takes to be) the truth concerning (a2). I submit, in other words, (i) that in the first instance 99b30-100b5 and 100b5-17 set out views which respond to (a1) and (a2) respectively, and (ii) that Aristotle’s goal in so doing is to draw out what the truth concerning (a1) and (a2) teaches us about (Q1) and (Q2) respectively.16 There are aspects of questions (Q1) and (Q2) on which Post. An. II.19’s handling of (a1) and (a2) sheds pretty much no light whatsoever-e.g. the vexing problem of how one can come to know of a foundation that it is a foundation. I think that all such matters are quite outside the scope of Post. An. II.19. The chapter does not, in fact, assert that aporiai (a1) and (a2) are the only yet-to-be-resolved aporiai that impede us from giving philosophically adequate answers to (Q1) and (Q2). In writing Post. An. II.19, Aristotle seems to have thought that beyond (a1) and (a2) the most important additional aporiai that impede us from giving philosophically adequate answers to (Q1) and (Q2) are aporiai about the soul and how cognition actually works.17

Our main focus below will be 99b30-100b5’s discussion of (a1)/(Q1). But before turning to 99b30-100b5, it will prove useful to rehearse a few basic points concerning 100b5-17’s discussion of (a2)/(Q2). Regarding the kind of non-demonstrative knowledge K on the basis of which a person with demonstrative episteme assents to an episteme-yielding demonstration’s premises, (a2) asks whether K is episteme or some different kind [genos] of knowledge. 100b5-17 famously opts for the latter option, contending that K is not episteme but another type of knowledge which Aristotle calls nous. On the final analysis, the English translation that best captures the basic sense and semantic range of the term nous in Aristotle’s Greek quite generally is probably ‘understanding’. Aristotle sometimes uses nous to indicate (what we might call) the understanding: i.e. the faculty of understanding/rational intelligence. But he also uses nous to mean a (perfected) understanding (of something): i.e. an acquirable intellectual virtue that a rational intelligence might, or might not, achieve. The Analytics univocally employs nous in the latter sense as a label for a quite specific perfection of intellect which (according to Aristotle) is epistemically superior to the intellectual virtue of episteme.18 A form of knowledge, paradigmatic instances of this intellectual virtue of nous consist (at least on my own reading of Aristotle) in perfected conceptualizations of essences through their real definientia.

In a word, then, nous is Aristotle’s 100b5-17 answer to question (Q2). According to 100b5-17: nous is the type of knowledge on the basis of which a person with demonstrative episteme assents to an epistemeyielding demonstration’s premises. The intellectual virtue of nous is the “foundation of episteme [archē epistēmēs]” (100b13). Now, getting clear on the exact account of nous that the Analytics’ theory of episteme presupposes would require an extended discussion, and is quite unnecessary for present purposes. It will, however, prove useful to have an unambiguous name for the kind of knowledge question (Q2) asks us to identify. So for ease of expression in what follows, let the (English neologism) ‘nous’ be just such a noun-a label for the kind of knowledge (Q2) asks us to identify and nothing more. Setting other senses of (the Greek word) nous aside, the basic thrust of (Q1) can then be restated as: How do we develop nous of demonstrative foundations?

3 From (a1)/(Q1) to the aporia of ἢ ἐϰ παντ ὸς

It is not, of course, uncommon for humans to have knowledge that they presently can neither consciously access nor in any way exercise. (Consider, e.g., an expert cobbler or geometer who is severely drunk or in the midst of a panic attack). What’s at issue in aporia (a1) is whether embodied human beings (i) are born lacking and need to acquire [lambanein 99b28] nous of foundations, or (ii) always already have [echein 99b26] this knowledge but initially lack-and must work to later develop-the ability to access and fully exercise it.19

Post. An. II.19 famously opts for alternative (i). Scholars have tended to identify alternative (ii) with Plato’s theory that humans learn (mathematical and other kinds of philosophical) knowledge by a kind of recollecting [anamnēsis]. Yet it must be admitted that the letter of II.19 hardly requires this identification. Aristotle was, moreover, a careful student of Plato’s Phaedo. But in an attempt to clarify the Meno’s far swifter 85b8-86b5 presentation of the Platonic theory of recollection, the Phaedo (73b3-76e7, esp. 75c776d6) manifestly takes great pains to distinguish the theory of recollection from the kind of strong innatism embodied in (ii). For recall that on the theory of recollection, our pre-embodied souls come into some sort of direct contact20 with imperceptible unchanging essences-the so-called “Forms”-and, as a result, acquire [lambanein] the correct understanding of each of them. Phaedo 75c7-76d6 explains that although our pre-embodied souls did indeed come to possess this kind of perfected knowledge of Forms, the process of embodiment must somehow destroy it [apollunai] because the vast majority of embodied humans souls do not in fact have [echein 75c7, d9, 76d2] any such knowledge. On the theory of recollection, the Phaedo explains, we embodied humans really do need to acquire [lambanein 74b4, c9, cf. 73d1] knowledge of Forms; we need, more precisely, to re-acquire [analambanein 75e4, e6] such knowledge as a good our souls previously did- but at birth do not-possess. We do this (according to the theory) by drawing on innate “true opinions” [alētheis doxai] which, while not themselves knowledge, importantly constitute surviving remnants of the no-longer-extant knowledge our pre-embodied souls used to have.21

So on Plato’s own interpretation of the theory of recollection, its proponents should actually agree with Post. An. II.19’s claim that we don’t always already have [echein 99b26] nous of foundations and do need to acquire [lambanein 99b28] this knowing state [hexis]. To this extent, Plato and Aristotle effectively respond to aporia (a1) in the same way: by rejecting alternative (ii) and developing alternative (i). The Platonic theory of recollection, however, would maintain that for humans to successfully acquire nous of a given foundation X we crucially need to access innately present X-specific information. And this is something that Aristotle will adamantly deny. Human beings, according to Aristotle, are naturally endowed with innate faculties of knowledge reception (like perception and a “blank” faculty of understanding); we are born with innate learning-conducive behavioral dispositions (like a compulsion to memetic imitation), and even an innate attraction to knowledge.22 But nature does not, he insists, supply us with any kind of innate contentful grasps.

On these grounds, among others, Aristotle will contend that the Platonic theory of recollection has to be false. And yet, while the details of II.19’s response to (a1)/(Q1) are, to be sure, at odds with the theory of recollection, scholars are quite wrong to characterize Post. An. II.19 as attempting to argue against the Platonic theory of recollection and in favor of some alternative theory of Aristotle’s. The chapter does not characterize itself as giving such an argument. Moreover, any philosophically serious Aristotelian attempt to do so would centrally involve explaining why the theory of recollection isn’t needed for solving the philosophical problems that motivated Plato to introduce it. But these are hard problems tied up with complex issues of metaphysics and mind.23 And neither Post. An. II.19 in particular-nor the Analytics in general-really works to explain how Aristotle proposes to handle these problems without giving up on the various rationalist and realist commitments he shares with Plato. Post. An. II.19, as we shall see, claims that humans can acquire nous of foundations by a process that begins in sense-perception; but an intelligent Platonist who fails to see how this could occur unless something like the theory of recollection is true will hardly find much help in II.19. Post. An. II.19’s responses to (a1)/(Q1) are most definitely not Plato’s; but in truth, the chapter does vanishingly little to show why its responses are better than Plato’s, or why we should believe the theory of recollection is false. That Post. An. II.19 aspires to be some kind of profound critical engagement with the Platonic theory of recollection is, I submit, a myth.24

But let us return to the aporia of ē ek pantos (ἢ ἐϰ παντὸς). This aporia (to repeat) arises in connection with 100a3-9 which, in turn, lies at the heart of 99b30-100b5. The task of 99b30-100b5 (as we’ve said) is to make progress on (Q1) by sketching what Aristotle takes to be the truth regarding (a1). Now, aporia (a1) concerns the question of whether humans (i) need to acquire nous of foundations, or (ii) always already have this knowledge. Aristotle himself endorses the former alternative. And the basic problem (a1) poses for alternative (i) is that acquiring new knowledge seems to essentially involve exercising prior knowledge we already have (99b28-30). For, having argued in Post. An. I.1 (71a1-2 ff.) that “all teaching and all learning, of the intellectual kind [dianoētikē], comes to be from prior knowledge [ek prohuparchousēs gnōseōs]”, 99b28-30 asks: “if we acquire [nous of foundations]-being previously without [it]-how could we get this knowledge and learn it from nonprior knowledge [ek mē prohuparchousēs gnōseōs]?” Aristotle’s Post. An. II.19 response to (a1) centrally emphasizes two points (99b30-34, cf. 100a10-11). The first is that in cases where we do acquire new knowledge by exercising knowledge we already have, the old knowledge needn’t be more epistemically valuable than nor prior by nature to the new knowledge. The second is that, like other animals, humans in fact possess primitive faculties of knowledge acquisition whereby we can gain new knowledge without exercising any pre-existing knowledge at all. All animals, Aristotle explains (99b34-35), have at least one such primitive faculty of knowledge acquisition: “the innate [sumphutos] faculty of discrimination [dunamis kritikē] that people call perception [aisthēsis]”. Some animals, he adds, have no knowledge [gnōsis] of perceptible objects when they’re not perceiving them; but other animals are so structured that exercises of perception suffice for bringing about lasting dispositional knowledge and developing further epistemic powers (99b36100a3). Thus some animals are of such a nature that by simply perceiving they acquire memories. And other animals are so constituted that after much perceiving and forming many memories, rationality [logos] is engendered within them.25 I take 100a1-3 to be saying that humans are not born with rationality, but rather acquire rationality from the world by perceiving and remembering what we perceive. As for how this occurs-and why one would be wrong to think that the theory of recollection is needed for explaining this kind of occurrence-such matters are passed over without comment.

Thus, in sum, 99b30-100a3. The remainder of Post. An. II.19’s response to (a1)/(Q1) comprises 100a3- 100b5. Turning away from knowers in general, and towards human knowers in particular, 100a3-100b5 sketches a highly abstract account of the multi-stage epistemic progression whereby we humans would proceed from ignorance to nous. It bears emphasis that what 100a3100b5 describes isn’t a justificational procedure or heuristic method that one might either adopt or fail to adopt. 100a3-100b5 concerns the type of intellectual progression that (according to Aristotle) is always successfully traversed when humans acquire nous of demonstrative foundation(s). At issue in 100a3-100b5 is something which “the [human] soul is capable of undergoing [paschein]” (100a14).

And so finally we arrive at Post. An. 100a3-9, the passage wherein our problematic ē ek pantos (100a6) is embedded. To work up the aporia of ē ek pantos, let us start by laying out Ross’ Greek text for 100a39 together with Barnes’ influential translation of the passage:

ἐϰ μέν οὖν αίσϑήσεως γίνεται υνήμη, ώσπερ λέγομεν, έϰ δέ μνήμης πολλάϰις τοῦ αὐτοῦ γινομένης έμπειρία, αί γάρ πολλαὶ μνῆμαι τῷ άριϑμῷ έμπειρία μία έστίν. έϰ δ’ έμπειρίας ἢ ἐϰ παντὸς ἠρεμήσαντος τοῦ ϰαϑὸλου ἐν τῆ ψυχῆ, τοῦ ἐνὸς παρὰ τὰ πλλά, ὃ ἃν ἐν ἃπασιν ἓν ἐνῆ ἐϰείνοις τὸ αύτό, τέχνης ἀρχὴ ϰαί ἐπιστήμης, ἐἀν μἐν περί γένεσιν, τέχνης, ἐἀν δἐ περὶ τὸ ὄν, ἐπιστήμης.

Thus from perception [aisthēsis] there comes memory [mnēmē], as we call it, and from memory (when it occurs often in connection with the same item) experience [empeiria]26; for memories which are many in number form a single experience. And from experience, or from all the universal which has come to rest in the soul (the one apart from the many, i.e. whatever is one and the same in all these items), there comes a principle [archē] of art [technē] or of understanding [epistēmē]-of art if it deals with how things come about, of understanding if it deals with how things are.

Taking pantos to modify tou katholou, Barnes translates 100a6’s ek d’ empeirias ē ek pantos ēremēsantos tou katholou en tēi psuchēi as “And from experience, or from all the universal which has come to rest in the soul”. This mereological reading of pantos (‘all’ in the sense of ‘the entirety of’) is arguably a bit linguistically difficult due to the position of the article. But Barnes’ construal is grammatically possible. And because ē ek pantos ēremēsantos tou katholou en tēi psuchēi simply cannot mean “from every universal which has come to rest in the soul”, the pantos in question is widely assumed to modify tou katholou and yield the meaning ‘all the universal’, i.e. ‘the entire universal’.

This much is basically uncontroversial.27 But it does leave us with the question as to why Aristotle is being so obscure in speaking of ‘all the universal’ rather than just ‘the universal’. Should all the universal be opposed to some or most of it? Aristotle simply does not use such locutions elsewhere, and the passage at issue does not seem to furnish any disambiguating hints. So why does he write ‘all the universal’ here? Why not simply write ‘the universal’? Various conjectures can be, and have been, given. But no-one seems to have convinced anyone else to adopt their own proposed conjecture; and most of us don’t find any of the proposals terribly promising.

So, part of the obscurity of 100a6’s ē ek pantos is due to the pantos (‘all’). But there’s also, of course, the ambiguity of the Greek the particle ē (‘or’). In fact, the debate about how to interpret ek d’ empeirias ē ek pantos ēremēsantos tou katholou en tēi psuchēi (“from experience or from all the universal which has come to rest in the soul”) is best viewed as centered around the question of how one should read this ē (‘or’). In an influential passage of his 1992 book, McKirahan explains the issue as follows:

After saying that memory arises from perception and experience arises from memory, [Aristotle] says

II.19 100a6-9 ‘From experience, or from the whole universal at rest in the soul, the one beside the many, which is one identical thing in all of them (the many), [arises] the principle of art and science.’

The issue in interpreting this passage is how to take the conjunction ‘or’ (ē). Is it (a) disjunctive (the principle of science comes either from experience or from the universal in the soul), (b) explicative (it comes from experience, that is to say, from the universal in the soul), or (c) progressive (it comes from experience, or rather from the universal in the soul, which is the next stage after experience)? The first can be eliminated straight off. [...] The second is adopted by most translators and commentators, but it contradicts the Metaphysics view that experience remains on the level of particulars. The parallels between Metaphysics A.1 and APo. II.19 are so close that we should be loath to find any significant disparity in the conceptions of experience present in the two chapters. Accordingly I prefer the third interpretation of ‘or’, on which ‘the universal at rest in the soul’ is a stage intermediate between experience and scientific knowledge. (McKirahan 1992, p. 243)

McKirahan’s terminology of “disjunctive”, “explicative”, and “progressive” has been taken up in many subsequent discussions of our passage.28 McKirahan’s contention that the “explicative” reading is the one “adopted by most scholars and translators” is less obviously true today than it was in 1992 when McKirahan wrote it. For quite a few recent authors-e.g. Charles (2000), Tuominen (2010), Bronstein (2016), Gasser-Wingate (2016)-have followed McKirahan in defending the “progressive” reading against the “explicative” construal adopted (e.g.) in influential publications of LeBlond (1939/1970), Ross (1949/1957), Barnes (1975/1993), and Bolton (1976). Hasper and Yurdin (2014)-who defend the “explicative” construal against the “progressive” alternative-aren’t obviously wrong to characterize the “explicative” construal as (still) the reading favored by “most scholars” (p. 122-123, n.7). But my own impression of the present scholarly landscape is that researchers who’ve worked on Post. An. II.19 are more-or-less evenly split on the “explicative” vs. “progressive” question. (For excellent reasons, the highly unlikely reading McKirahan calls “disjunctive” has never been seriously endorsed).

So scholars are divided as whether 100a6’s ē ek pantos...psuchēi (“or from all the universal having come to rest in the soul”) should be given a “explicative” or “progressive” reading. On the “explicative” reading, ē ek pantos...psuchēi is saying that experience somehow constitutes a grasp of “the entire universal”; and 100a3-9 would then be describing a four stage cognitive progression whereby nous is acquired:

perception → memory → experience → nous

On the “progressive” reading, ē ek pantos...psuchēi is saying that a grasp of “the entire universal” is some kind of intermediate cognitive achievement superior to experience but inferior to nous; and 100a3-9 would then be describing a five stage cognitive progression whereby nous is acquired:

perception → memory → experience → grasp of all the universal → nous

Neither construal is obviously incorrect. And nothing in the remainder of Post. An. II.19, or elsewhere in the Corpus Aristotelicum, is thought to decisively rule out either interpretive possibility.

Now, I don’t myself think that 100a3-9 is supposed to present some boldly inventive piece of anti-Platonic epistemology; I don’t myself think that 100a3-9 articulates a centrally important component of the theory of episteme that the Analytics works to develop; and I don’t myself believe of Post. An. II.19 that 100a6-8 “makes the central claim of the chapter as a whole” (Adamson 2010, p. 9). But on all of these counts, many recent scholars have thought otherwise. One result of this is that Post. An. scholarship has come to contain a great multiplicity of philosophicallyinteresting, albeit highly speculative, interpretive proposals developed out of the “progressive” and “explicative” readings of 100a6’s ē ek pantos respectively. The “progressive” vs. “explicative” question has become highly contentious. And many competing interpretations of 100a3-9 in particular-and Post. An. II.19 in general-now crucially require adopting one of these two construals of 100a6’s ē ek pantos and rejecting the other.29

In the next section I’ll be proposing a novel response the aporia of ē ek pantos. But before bringing this section to a close, it’s worth rehearsing the main and most powerful arguments that have been adduced for the “explicative” and “progressive” construals of 100a6’s ē ek pantos respectively.

The chief argument in favor of the “explicative” reading and against the “progressive” alternative is that the “explicative” construal of ē ek pantos looks quite a bit linguistically easier than the “progressive”.30 Now, proponents of the “explicative” reading are (I think) correct on this point of language. But proponents of the “progressive” reading of ē ek pantos are reasonably unmoved by the argument. For as a matter of Greek, the “progressive” construal really does seem to be linguistically possible. And in the interpretation of Aristotle, considerations of linguistic naturalness need to be weighed against many other considerations, hermeneutical and philosophical.

The chief argument in favor of the “progressive” reading and against the “explicative” alternative is that the “progressive” coheres better with a closely parallel text in Metaphysics A.1. In language impressively similar to that of Post. An. II.19, Met. A.1 980a27 ff. explains the development of technē (craft-knowledge) as follows:31

By nature, animals are born with perception. And from perception, memory is engendered in some of them, while in others it’s not. [...] The other animals live by impressions and memories, yet have little share of experience. But the human race lives [not only by impressions and memory but] also by technē and reasonings. And experience comes to humans from memory. For many memories of the same thing furnishes the power of one experience. Experience, indeed, seems a bit similar to the science of technē.32 But in fact, the science of technē result for human beings through experience. For experience, as Polus says (speaking correctly), makes technē, while inexperience makes luck [tuchē]. And technē arises when from the many conceptions of experience one universal grasp [mia katholou hupolēpsis] concerning similar things comes about. For [what’s characteristic] of experience is to have a grasp that when Callias gets ill with this disease this helps, and Socrates too, and similarly in many particular [cases]. But [what’s characteristic] of technē is to have a grasp that this helps everyone of a certain constitution-in one demarcated class-who gets ill with this disease. [...] The reason [why the experienced are more successful than unexperienced people with the correct accounts] is that experience is knowledge of the particulars [gnōsis tōn kath’ ekaston] while technē is knowledge of the universals [tōn katholou]. (Met. A.1 980a27-980b21, 980b25-981a16)

The linguistic, narrative, and philosophical congruences between Post. An. II.19 99b34-100a9 and this Met. A.1 parallel are quite remarkable. Now, according to the latter text, what the experienced person without technē characteristically lacks is a grasp [hupolēpsis] and knowledge [gnōsis] of the universal [tou katholou]. But on the “explicative” reading of Post. An. II.19 100a6, ē ek pantos...psuchēi serves to identify experience with grasping “the entire universal”. And this, it is alleged, tells against construing 100a6’s ē ek pantos...psuchēi as “explicative” and supports construing it as “progressive”.

Proponents of the “explicative” reading of ē ek pantos haven’t found this argument terribly persuasive. And neither do I. For note that our Met. A.1 parallel describes a cognitive progression in four, and not five stages:

perception → memory → experience → technē

So there’s also some tension between Met. A.1 and the “progressive” reading of ē ek pantos...psuchēi in Post. An. II.19. Is the alleged tension between Met. A.1 and the “explicative” reading of the phrase really all that worse? If their appeal to Met. A.1 is to be helpful, proponents of the “progressive” reading need us to grant (i) that Met. A.1 and Post. An. II.19 provide compatible, parallel accounts of the same general progression, and (ii) that the passages’ differing contexts can easily explain why one (Met. A.1) gives a four stage analysis while the other (Post. An. II.19) gives a five stage analysis. But if we are to grant (ii) to proponents of the “progressive” reading, shouldn’t we also grant proponents of the “explicative” reading that the differing contexts of Post. An. II.19 and Met. A.1 could easily have motivated Aristotle to use “universal” [katholou] in two different senses- meaning one thing by katholou in Post. An. II.19 (100a3-9) and another by katholou in the Met. A.1 parallel?

4 A new solution the aporia of ē ek pantos

We are left, then, with the following aporia concerning how to understand ē ek pantos... (“or from all...”) at Post. An. II.19 100a6. Numerous competing interpretations of Post. An. II.19 in general-and 100a3-9 in particular-require adopting one and not another construal of this ē ek pantos. Some scholars read the relevant ē (‘or’) as “progressive” while others read it as “explicative”. Arguments have been adduced in favor of both possibilities. But neither of the two cases is terribly strong; and it’s far from clear that either case is significantly stronger than the other. Why 100a6 speaks of “all [pantos] the universal coming to rest in the soul”-rather than simply “the universal coming to rest in the soul”-is anyone’s guess.

My proposed solution to this aporia is more a dissolution than a resolution of the puzzle. It sets out from a simple observation I haven’t found elsewhere discussed: the words ē ek pantos that we read in modern printings of 100a6 are unattested in some quite important witnesses to the text of Post. An. II.19.

4.1 Evidence of “direct” witnesses to the text of 100a6

Given the present state of scholarship on the Greek Post. An. manuscript tradition, the most evidentially important “direct” witnesses to the text of Post. An. II.19 look to be (i) manuscripts A, B, d, and n consulted by Ross (1949), and (ii) manuscripts D, V, and G of which Ross was ignorant.33 , 34

As for the relative genealogy of the group of manuscripts {A, B, d, n, V, G, D} it is readily shown that:

  1. our Post. An. manuscripts A, B, d, V, and G all descend from a (no-longer-extant) common ancestor α from which neither D nor n descend

  2. our Post. An. manuscripts D and n descend from a (no-longer-extant) common ancestor β from which none of A, B, d, V, or G descend

It’s with reference to the above that we speak of ‘α-family’ and ‘β-family’ Post. An. manuscripts.

At the 100a6 text-location empeirias...ēremēsantos in our α-family manuscripts, we find that A, B, d, V, and G furnish a number of variant readings:

έμπειρίας ἢ ἐϰ παντὸς ἠρεμήσαντος BG

έμπειρίας ἢ ἐϰ παντὸς ἠρεμίσαντος A

έμπειρίας ἡ ἐϰ παντὸς ἠρεμήσαντος V

έμπειρίας ἢ ἐϰτὸς ἀριϑμήσαντος d

But at the same text-location in both our β-family witnesses, n1 and D1, the words ē ek pantos (ἢ ἐϰ παντὸς) are omitted. For 100a6’s empeirias...ēremēsantos:

ὲμπειρίας ἠρεμήσαντος nD

This latter variant, as we shall see, is also found in significant “indirect” witnesses to the text of 100a6.

4.2 Evidence of “indirect” witnesses to the text of 100a6

The oldest Post. An. commentary of whose contents we have any substantial knowledge is that of Alexander of Aphrodisias (fl. 2-3rd cent. CE). Originally a quite massive work, commenting on both of the Posterior Analytics’ two books, the bulk Alexander’s Post. An. commentary is no longer extant. Numerous fragments from Alexander’s Post. An. commentary do, however, happily survive. The Analyticorum Posteriorum Paraphrasis of Themistius (ca. 317-390 CE) survives in full, and in paraphrasing “the meaning” [ta boulēmata] of Post. An. I-II often follows Aristotle’s Greek quite closely. As for Neoplatonic commentators, while there can be little doubt that Proclus (d. 485 CE) gave influential lectures on the Post. An. in Athens, it is far less clear that these lectures were ever circulated in the form of a book. The Post. An. lectures of Proclus’ student Ammonius, however, did form the basis for a written commentary by Philoponus (fl. 6th cent. CE). And Philoponus’ commentary on Post. An. I is fully extant. Philoponus, conceivably, might also have commented on Post. An. II. But the commentary on Post. An. II attributed to Philoponus in the Commentaria in Aristotelem Graeca (CAG ) series is the work of a medieval Byzantine author (= Leo Magentinus, according to Ebbesen) and not Philoponus himself. The so-called “Anonymous Commentary” on Post. An. II (i.e. Anonymi in Analyticorum Posteriorum Librum Alterum Commentarium in CAG 13.3) isn’t exactly a commentary in the traditional sense. Commenting on much, though not all, of Post. An. II, the “Anonymous Commentary” presents itself as a sequence of excerpts. Due to the important study of Moraux (1979), we now know the primary-and perhaps exclusive-source of these excerpts to be the largely lost Post. An. commentary of Alexander.35

Now, among the excerpts one finds in the “Anonymous Commentary” on Post. An. II, there turn out to be two discussions (602,2 ff. and 600,26 ff.) that actually quote 100a6. And on both of these occasions, the “Anonymous Commentary” quotes 100a6 in a version that omits the 3 words ē ek pantos (ἢ ὲϰ παντὁς). The text for 100a6’s ek d’ empeirias...katholou that whoever wrote these discussions read in their copy/copies of the Post. An. was:

έϰ δ’ έμπειρίας ἠρεµήσαντος τοῦ ϰαϑόλου

and not, as one reads in modern editions:

έϰ δ’ έμπειρίας ἢ έϰ παντὸς ἠρεµήσαντος τοῦ ϰαϑόλου

So the “Anonymous Commentary” testifies to the same 100a6 variant as our β-family manuscripts n and D.

Boethius’ 6th cent. translation of the Post. An. into Latin seems to have been lost not long after it was produced. And we have no evidence that the Post. An. was re-translated into Latin before the 12th cent. The Syriac Post. An. translations of Athanasius of Balad (d. 687) and especially Isḥāq Ibn Ḥunayn (d. 910/911) seem to have faired much better than Boethius’. But neither of them, nor any other Syriac Post. An. translation, presently survives. The oldest extant Post. An. translation in any language is the Arabic Post. An. due to Abū Bishr Mattā (d. 940). While the oldest Greek Post. An. manuscripts that survive today date to the 9-10th cent., the no-longer-extant Greek Post. An. manuscript(s) on which Abū Bishr’s Arabic most immediately depends are early 9th cent. at the latest and likely much older.36

If we attend to Abū Bishr’s Arabic translation of Post. An. 100a3-9, we find no trace of the 3 words ē ek pantos in his rendering of 100a6.37 The Greek that corresponds to Abū Bishr’s Arabic for 100a6’s ek d’ empeirias...tou katholou is

έϰ δ’ έμπειρίας ἠρεµήσαντος τοῦ ϰαϑόλου

Abū Bishr’s translation presupposes a Greek source that contained the same version of 100a6 which we find in our β-family Post. An. manuscripts-the same version of 100a6 which the “Anonymous Commentary” quotes.

Themistius’ 4th cent. CE reworking of Post. An. II.19 in his Paraphrasis furnishes no clear evidence as to whether he read the words ē ek pantos in the version of 100a6 known to him. As far as the Greek Post. An. commentary tradition goes, the earliest commentators that we do positively know to have read the words ē ek pantos at 100a6 turn out to be 12-13th cent. Byzantine scholars like Eustratius (CAG 21.1 264,8 ff.) and Pseudo-Philoponus (CAG 13.3 436,2 ff.). The enormously influential 12th cent. Latin Post. An. translation due to James of Venice (see AL IV.1) renders 100a6: ex experimento autem aut ex omni quiescente [=ἢ ἐϰ παντὸς ἠρεµήσαντος] universali in anima. And this was the version of 100a6 known (e.g.) to Aquinas.38

4.3 The text of 100a6 in some notable printed editions

Moving forward a few centuries, at 100a6 in the 1495 Aldine Organon one reads (sic)

ἐϰ δ’ έμπειρίας, ἡ ἐϰ παντὸς ἠρεµήσαντος τοῦ ϰαϑόλου

But if one turns to Post. An. II.19 in Pacius’ influential 1592 Organon, one reads at 100a6 (the better) α-family variant

ἐϰ δ’ έμπειρίας ἢ ἐϰ παντὸς ἠρεµήσαντος τοῦ ϰαϑόλου

By the 19th cent., the version of 100a6’s ek d’empeirias ...katholou printed by Pacius seems to have decisively established itself as the vulgate text. Bekker-who for Post. An. II.19 consulted the 3 manuscripts: A, B, and D-prints Pacius’ text for 100a6 in his 1831 edition without recording the β-family alternative that omits ē ek pantos in his apparatus. While Ross (1949) and Waitz (1844-46) had no access to Abū Bishr’s Arabic and were unable to consult manuscript D, both correctly note the omission of ē ek pantos at 100a6 in the first hand of manuscript n. But Ross (1949) and Waitz (1844-46) both print the vulgate text of Bekker/Pacius for 100a6. Due largely (I suspect) to their incomplete knowledge of the available evidence, neither editor bothers to justify doing so.

4.4 The proposal

It turn outs to be clear that in the early medieval (and probably late ancient) world, two quite different versions of 100a6 were in circulation: a version that contained the words ē ek pantos and a version which didn’t.39 The Post. An. II.19 that Aristotle actually wrote wouldn’t have contained both versions of 100a6. So text-critical reasoning is needed to determine what we ought to believe the authentic 100a6 to read.

We can speak, in particular, of the following two versions of 100a6’s ek d’...katholou:

Version 1

ἐϰ δ’ έμπειρίας ἢ ἐϰ παντὸς ἠρεµήσαντος τοῦ ϰαϑόλου

Version 2

ἐϰ δ’ ἐμπειρίας ἠρεµήσαντος τοῦ ϰαϑόλου

We’ve noted that Version 1 is supported by (i) the “direct” testimony of certain α-family manuscripts, as well as (ii) the “indirect” testimony of certain 12-13th cent. Byzantine Post. An. commentaries and certain 12-13th cent. Latin Post. An. translations. We’ve noted that Version 2 is supported by (i) the “direct” testimony of our β-family manuscripts, as well as (ii) the “indirect” testimony of Abū Bishr’s 10th cent. Arabic Post. An. translation and the so-called “Anonymous commentary” on Post. An. II (whose discussions Moraux has shown to be excerpted largely, and perhaps exclusively, from Alexander). Version 1 is printed by Pacius, Bekker, Waitz, and Ross and presupposed by all modern discussions of 100a3-9 I’m aware of. But significant evidence-some quite unknown Pacius, Bekker, Waitz, and Ross-supports Version 2.

Ultimately, I find it difficult to resist the following conclusion. Careful scrutiny of the relevant textgenealogical, linguistic, hermeneutical, and philosophical considerations tells strongly in favor of taking Version 2 to preserve what Aristotle in fact wrote, and rejecting Version 1 as spurious. I contend that the words ē ek pantos one reads at 100a6 in all modern editions of the Post. An. were not put there by Aristotle; I contend that in the transmission history of the text these 3 words were erroneously interpolated into the α-family exemplar. This, in a nutshell, is my proposed response to the aporia of ē ek pantos.

To establish the basic viability of this, let’s start with the most pressing text-genealogical issue it raises. If one adopts the hypothesis that Version 2 of 100a6 is correct, the erroneous appearance of ē ek pantos in α-family manuscripts needs accounted for. Now, this can in fact be done in several plausible ways. But for present purposes, it suffices to rehearse two such plausible explanations-both appealing to one of the most common and widely recognized mechanisms of accidental interpolation.40

If Version 2 of 100a6 is correct, the erroneous emergence of Version 1 could easily have resulted from a process like one of the following. (Option 1) The closeness of these uncial letter patterns is striking:

This suggests the following hypothesis. In the early transmission history of the Post. An., a scribe erred in transcribing 100a6 by erroneously writing ἢ ἐϰ παντὸς where s/he ought to have written ἠρεµήσαντος (perhaps the text of the source-manuscript was semi-legible at 100a6). As a result, ἢ ἐϰ παντὸς enters the manuscript tradition as a variant for ἠρεµήσαντος. (NB that ἐϰ δ’ ἐμπειρίας ἢ ἐϰ παντὸς τοῦ ϰαϑόλου ἐν τῆ ψυχῆ isn’t non-sense). Later, a diligent scholar records ἢ ἐϰ παντὸς as a variant for ἠρεµήσαντος in some α-ancestor (perhaps in supralinear position, perhaps in the left margin beside a line initial ἠρεµήσαντος). Finally, the scribe of some subsequent α-ancestor conflates this reported variant for a proposed insertion and erroneously writes ἢ ἐϰ παντὸς in front of ἠρεµήσαντος. (Option 2) Scholars in the Greek world have long used the particle ἢ to introduce explanatory glosses and the like. This suggests the following hypothesis. At some point in the transmission history of the Post. An., a scholar transcribes ἢ ἐϰ παντὸς into her copy of the Post. An. as a meta-textual comment on 100a6 (perhaps as a kind of gloss on ϰαϑόλου, perhaps to indicate the manner of ἠρεµήσαντος τοῦ ϰαϑόλου ἐν τῇ ψυχῇ that’s requisite for the emergence of τέχνης ἀρχἠ ϰαὶ έπιστήμης). Later, the scribe of some α-ancestor confl this meta-textual comment for a proposed insertion and erroneously writes ἢ ἐϰ παντὸς in front of ἠρεµήσαντος.

For present purposes, we needn’t dwell further on Options 1 and 2. Both can be tweaked in various ways; it can be argued that one of them is more likely than the other; and the same explanatory work can undoubtably be done otherwise. All that matters here is that it needs to be admitted as a not-at-all-unlikely historical possibility that Version 1 entered the Post. An. manuscript tradition through some such ē ek pantos interpolating event. And this much should now be clear.

At a minimum, the last few pages have shown that the collective testimony of our most important surviving text-witnesses to 100a6 no more disfavors the hypothesis that Version 2 is correct than it disfavors the competing hypothesis that Version 1 is correct. I submit, moreover, that there are strong linguistic, hermeneutical, and philosophical considerations that support adopting Version 2 and rejecting Version 1. To start with the obvious, Version 1 gives us a text with an unfortunate ambiguity (ē) and unnecessary obscurity (pantos). Version 2, in contrast, reads much more smoothly. In fact, with ē ek pantos out of the picture, the 100a3-9 we are left with is a wellbalanced, and kind of beautiful, piece of Aristotelian prose.

To dig deeper, let’s take a closer look at 100a3-9. We need to ask how 100a3-9 will need to be interpreted if Version 2 of 100a6 is adopted. And we’ll do well by starting with some points of grammar. The first thing to note is the prominence of the ek men oun aisthēseōs...ek de mnēmēs...ek d’ empeirias structuring of 100a3-6. The finite verb ginetiai that’s explicit in the ek men unit will need to be supplied twice: both with empeiria (100a5) to get the main clause of the first ek de unit, and also with archē (100a8) to get the main clause in the second ek de unit. (NB that 100a7-8’s tou auto is in apposition to 100a6’s tou katholou). Note finally, that in view of the text’s pronounced structure we’ll want to take the participial clauses pollakis tou autou ginomenēs (100a4) and ēremēsantos tou katholou en tēi psuchēi (100a6) as grammatical parallels. Given the genders of the relevant participles, it seems that mnēmēs will have to be the (implied) subject of 100a4’s ginomenēs, and that tou katholou will have to be the subject of 100a6’s ēremēsantos.

Adopting Version 2 of 100a6, I therefore propose translating 100a3-9 as follows:

So, from perception (as we say) memory comes to be. And from memory-when it comes to be of the same thing many times-experience comes to be (for, the memories, numerically many, [constitute] one experience). And from experience-when the universal has come to rest in the soul (the one over the many, whichever one it is that’s present in all of those things the same)-a foundation [archē] of episteme or of craft-knowledge [technē] comes to be. (It’s of craft-knowledge if it’s [an archē] for generating; it’s of episteme if it’s [an archē] for being).

έϰ μέν οὖν αίσϑήσεως γίνεται υνήμη, ὥσπερ λέγομεν, ἐϰ δὲ μνήμης πολλάϰις τοῦ αὐτοῦ γινομένης ἐμπειρία, αί γὰρ πολλαὶ μνῆμαι τῷ ἀριϑμῷ ἐμπειρία μία ἐστίν. ἐϰ δ’ ἐμπειρίας ἠρεμήσαντος τοῦ ϰαϑόλου ἐν τῇ ψυχῇ, τοῦ ἐνὸς παρὰ τὰ πολλά ὃ ἂν ἐν ἅπασιν ἓν ἐνῇ ἐϰείνοις τὸ αὺτό, τέχνης ἀρχὴ ϰαὶ ἐπιστήμης, ἐὰν μὲν περί γένεσιν, τέχνης, ἐὰν δὲ περὶ τὸ ὄν, ἐπιστήμης.

If the authentic 100a3-9 is the one I’ve printed above, Aristotle’s basic train of thought would look to be the following.

100a3-9 is sketching a highly abstract account of the intellectual progression whereby humans would acquire nous-a knowledge of demonstrative foundations that’s suitable for grounding demonstrative episteme. The account is highly abstract because it’s supposed to be general enough to cover all possible cases of demonstrative episteme. The account aims, that is, to capture not only how humans would develop states of nous needed to ground demonstrative episteme regarding (say) the natural world, but also how humans would develop states of nous needed to ground demonstrative episteme concerning truths studied in mathematics and metaphysics. (This bears emphasis because as NE 1142a16-20 makes clear there’s a sense of “from experience” [ek empeirias] on which Aristotle will deny that nous of mathematical foundations is “from experience”). In fact, 100a3-9 presents its account as so abstract that it even captures how humans would acquire the kind of foundational knowledge on which craftknowledge [technē] is based. But putting technē aside, we can say that 100a3-9 introduces four knowledgefurnishing states [hexeis] that humans can possess: (i) perception [aisthēsis], (ii) memory [mnēmē], (iii) experience [empeiria], and (iv) nous: i.e. the “foundation of episteme” [archē epistēmes].

If the authentic 100a6 is Version 2, 100a3-9 will be presenting a highly abstract sketch of an intellectual progression in four stages:


100a4’s pollakis tou autou ginomenēs (“when [memory] comes to be of the same [thing] many times”) evidently describes a necessary condition for the occurrence of transition-2. And 100a6’s ēremēsantos tou katholou en tēi psuchēi (“when the universal has come to rest in the soul”) will likewise describe necessary a condition for the occurrence of transition-3. 100a5’s parenthetical hai gar pollai mnēmai tōi arithmōi empeiria mia estin (“for, the memories, numerically many, are an experience that’s one”) strongly suggests that 100a4’s pollakis tou autou ginomenēs is intended as a rough-and-ready characterization of a necessary and sufficient condition for the occurrence of transition-2. And if this is correct, our text’s pronounced structure will suggest that ēremēsantos tou katholou en tēi psuchēi correspondingly names a necessary and sufficient condition for the occurrence of transition-3. In fact, given a parallel passage in the Phaedo (96a6-b8)41 as well as Aristotle’s well-known epistemic uses of the verb hērmein elsewhere,42 I think that our default assumption needs to be that on Version 2 of 100a6, ēremēsantos tou katholou en tēi psuchēi does name a necessary and sufficient condition for the occurrence of transition-3. The noun phrase tou katholou (“the universal”) would then refer to the object of nous. Now, there are scholars who take “the universal” mentioned at 100a6 to be a universal proposition. But this proposal is severely undermined by Aristotle’s own 100a78 gloss of the relevant phrase.43 For 100a7-8 describes the universal in question as a “one over many” [hen para ta polla]: as something that can be uniformly “present within” [enēi ] a plurality of things; and this is simply very hard to square with Aristotle’s usual ways of thinking/speaking about universal propositions [protaseis]. Far more likely, I submit, is that “the universal” that “comes to rest in the soul” is an eternal and unchanging universal nature-roughly, an Aristotelian analogue of a Platonic Form. This very conceptual analogy, I suggest, is the basic philosophical point behind 100a7’s self-consciously Platonic turn of phrase “one over the many” (cf. Post. An. I.11 77a5 ff., I.22 82a32-35).

Before drawing this section to a close, permit me a few final remarks on the relationship between Post. An. 100a3-9 and the Phaedo passage mentioned in the preceding paragraph (Phd. 96a6-b8). The opening of the so-called “philosophical autobiography”, the text of Phd. 96a6-b8 might be translated as follows (for Plato’s Greek see footnote 41):

“O Cebes,” said [Socrates], “when I was young, I became incredibly passionate for the kind of wisdom which people call natural history [historia peri phuseōs]. For I thought it sublime-to know the explanations [tas aitias] of each thing: [to know] why each thing comes to be, why it perishes, and why it is. And many times I would turn myself back and forth, examining first of all the following sorts things: Is it the case, as some were claiming, that living things are assembled precisely when the hot and the cold start to decay? And it is blood by which we are wise [phronoumen]? Or is it air ? Or fire? Or is it none of these, but it is rather the brain that furnishes the senses [aisthēseis] of hearing, seeing, and smelling? And is it the case [as some were claiming]44that from the senses, memory [mnēme] and opinion [doxa ] come to be, and from memory and opinion-when it has acquired fixedness [to hērmein ]-along these lines epistēmē comes to be?’ (Phd. 96a6-b8)

The linguistic parallels between Phd. 96b6-8 (bolded) and Post. An. 100a3-9 are hard to exaggerate. Given the context and grammar (NB footnote 44) of the Phaedo text, it’s quite likely that Phd. 96b6-8 and Post. An. 100a3-9 are both making conscious reference to some earlier Presocratic source. On the final analysis, I suspect Aristotle himself would have viewed Post. An. 100a3-9 as distilling a small handful of more-or-less widely accepted endoxa [≈ “reputable opinions”] about how humans develop intellectual virtue-endoxa drawn (I suspect) not only from Presocratic natural philosophy, but also from ancient debates about the epistemology of technē.45

Our Phaedo passage (96b6-8) speaks of epistēmē emerging when an opinion/view [doxa] “has acquired fixedness” [labousēs to hērmein]. This latter idea is famously developed in the Meno (97e-98a) which explains that states of epistēmē come about when correct opinions/views [orthai doxai ] are tied down. Above we translated the Post. An. 100a6 phrase hērmēsantos tou katholou en tēi psuchēi as “when the universal comes to a rest [hērmēsantos] in the soul”. But the verb hērmein from which hērmēsantos derives can indicate unmovable presence (“standing firm”) as well as unmoving presence (“standing still”). It’s arguable that an alternative translation like “when the universal has become fixed in the soul” better captures the thought behind 100a6’s hērmēsantos tou katholou en tēi psuchēi. At any rate, Aristotle’s considered view in the Analytics is that because a person’s episteme through a demonstration D must be rationally unshakeable [ametapeiston], the nous from which this person assents to D’s premises must itself be no less unshakeable (cf. Post. An. I.2 72a25-b3 to which the II.19 text 99b20-22 apparently refers).

5 Conclusion

The preceding sections have sketched, and worked to motivate, a controversial approach to Post. An. II.19-one considerably more deflationary than most. But we have not rehearsed a complete reading of the chapter.46 For our main goal has been to recommend a novel solution to the aporia of ē ek pantos. The proposed solution, as I’ve said, is more a dissolution than a resolution of the underlying puzzle. But this dissolution I propose seems to me considerably more philosophically, hermeneutically, and textually satisfying than any alternative solution I’m aware of. For let us return to the chief arguments marshaled in favor of the “explicative” and “progressive” construals of 100a6’s ē ek pantos respectively.

The chief argument in favor of the “explicative” reading and against the “progressive” alternative was that the “explicative” is more linguistically natural than the “progressive”. Be this as it may, I submit that the variant ek d’ empeirias ēremēsantos tou katholou is more linguistically plausible Aristotelian Greek than ek d’ empeirias ē ek pantos ēremēsantos tou katholou . So with respect to considerations of linguistic naturalness, our proposed solution to the aporia of ē ek pantos in an important way looks to fare better than the “explicative” reading of ē ek pantos.

The chief argument in favor of the “progressive” reading of 100a3-9 and against the “explicative” alternative was that the “progressive” reading more closely coheres with a parallel text in Metaphysics A.1: i.e. 980a27-981a16.But this paper’s proposal gives us a reading of 100a3-9 that coheres even better with 980a27-981a16 than the “progressive” reading does. For, recall that 980a27-981a16 describes a four stage intellectual progression wherein the object grasped at stage-four is called ‘the universal’. Our proposed solution to the aporia of ē ek pantos likewise has 100a3-9 describing a four stage progression wherein the object grasped at stage-four is called ‘the universal’; the upshot of the “progressive” reading of 100a3-9, in contrast, is a five stage intellectual progression wherein ‘the universal’ names the object grasped at a mysterious penultimate stage which Aristotle simply doesn’t seem to discuss anywhere else.

5 References

Adamson, P. “Posterior Analytics 2.19: A dialogue with Plato?”. Bulletin of the Institute of Classical Studies, 54(S107), pp. 1-19, 2010. [ Links ]

Badawī, A. Manṭiq Arisṭū (Organon Aristotelis) Vol. I-III. Cairo: Maktabat Dār al-Kutub almis.riya, 1948-52. [ Links ]

Barnes, J. Aristotle: Posterior Analytics: Translation and Commentary. Oxford: Clarendon, 1975/1993. [ Links ]

Bekker, I. Aristoteles Graece Vol. I. Berlin: Reimer, 1831. [ Links ]

Berti, E. Aristotle on Science: The Posterior Analytics. Padua: Antenore, 1981. [ Links ]

Biard, J. Raison et Démonstration: Les Commentaires Médiévaux sur Les Seconds Analytiques. Turnhout: Brepols, 2015. [ Links ]

Bolton, R. “Essentialism and semantic theory in Aristotle: Posterior Analytics, II.7-10”. Philosophical Review, 85(4), pp. 514-544, 1976. [ Links ]

Bolton, R. and Code, A. “Aristotle on Knowledge”. In S. Hetherington (ed.) (2012), pp. 55-78. [ Links ]

Brockmann, C. “Das Papyrusfragment und die ältesten byzantinischen Textzeugen der Analytika des Aristoteles”. Philologus, 148(1), pp. 50-63, 2004. [ Links ]

Bronstein, D. “Aristotle’s critique of Plato’s theory of innate knowledge”. Logical Analysis and History of Philosophy, 19, pp. 126-139, 2016. [ Links ]

Bronstein, D. Aristotle on Knowledge and Learning: The Posterior Analytics. Oxford: OUP, 2016. [ Links ]

Brunschwig, J. “L’Objet et la structure des Seconds Analytiques d’apres Aristote”. In E. Berti (ed.) (1981), pp. 61-96. [ Links ]

Charles, D. Aristotle on Meaning and Essence. Oxford: Clarendon , 2000. [ Links ]

de Haas, F. and Leunissen, M. and Martijn, M. Interpreting Aristotle’s Posterior Analytics in Late Antiquity and Beyond. Leiden: Brill, 2010. [ Links ]

Detel, W. Aristoteles: Analytica Posteriora: Übersetzt und Erläutert Vol. I-II. Berlin: Akademie Verlag, 1993. [ Links ]

Ebbesen, S. “[Review of] F. de Haas, M. Leunissen and M. Martijn, eds., Interpreting Aristotle’s Posterior Analytics in Late Antiquity and Beyond ”. Aestimatio, 9, pp. 355-366, 2012. [ Links ]

Ebbesen, S. “The Posterior Analytics 1100-1400 in East and West”. In J. Biard (ed.) (2015), pp. 1130. [ Links ]

Fine, G. The Possibility of Inquiry: Meno’s Paradox from Socrates to Sextus. Oxford: OUP , 2014. [ Links ]

Gasser-Wingate, M. “Aristotle on induction and first principles”. Philosophers Imprint, 16(4), pp. 1-20, 2016. [ Links ]

Gregorić, P. and Grgić, F. “Aristotle’s notion of experience”. Archiv für Geschichte der Philosophie, 88(1), pp. 1-30, 2006. [ Links ]

Goulet, R. Dictionnaire des Philosophes Antiques Vol. I. Paris: CNRS Editions, 1989. [ Links ]

Hasper, P. S. “A new independent group of manuscripts for Aristotle’s Sophistici Elenchi ”. Work-in-progress. [ Links ]

Hasper, P. S. and Yurdin, J. “Between perception and scientific knowledge: Aristotle’s account of experience”. Oxford Studies in Ancient Philosophy, 47, pp. 119-150, 2014. [ Links ]

Hayduck, M. Eustratii in Analyticorum Posteriorum Librum Secundum Commentarium (CAG Vol. XXI.1). Berlin: Reimer , 1907. [ Links ]

Hetherington, S. Epistemology: The Key Thinkers. London: Cambridge University Press, 2012. [ Links ]

Le Blond, J. M. Logique et Méthode chez Aristote. Paris: J Vrin, 1939/1970. [ Links ]

Moraux, P. Le Commentaire d’Alexandre d’Aphrodise aux Seconds Analytiques d’Aristote. Berlin: Walter de Gruyter, 1979. [ Links ]

Minio-Paluello, L. “Note sull’ Aristotele latino medievale”. Rivisita di filosofia neo-scolastica, 43, pp. 97-124, 1951. [ Links ]

Minio-Paluello, L. Analytica Posteriora: Transl. Ionnis et Anon. et Gerardi. et recensio Guillelmi de Moerbeka (AL Vol. IV.1-3). BrugesParis: Desclée de Brouwer, 1953-54. [ Links ]

Peters, F. Aristoteles Arabus: The Oriental Translations and Commentaries on the Aristotelian Corpus. Leiden: Brill , 1968. [ Links ]

Primavesi, O. “Aristotle’s Metaphysics A: A New Critical Edition with Introduction”. In C. Steel (ed.) (2010), pp. 385-516. [ Links ]

Reinsch, D. R. “Fragmente einer OrganonHandschrift vom Beginn des zehnten Jahrhunderts aus dem Katharinenkloster auf dem Berge Sinai”. Philologus, 145, pp. 57-69, 2001. [ Links ]

Ross, W.D. Aristotle’s Prior and Posterior Analytics: A Revised Text with Introduction and Commentary : Translation and Commentary. Oxford: Clarendon , 1949. [ Links ]

Smith, R. “Immediate propositions and Aristotle’s proof theory”. Ancient Philosophy, 6, pp. 47-68, 1986. [ Links ]

Sorabji, R. “The Ancient commentators on concept formation”. In F. de Haas, M. Leunissen, and M. Martijn (ed.) (2010), pp. 3-26. [ Links ]

Steel, C. Aristotle’s Metaphysics Alpha: Symposium Aristotelicum. Oxford: OUP , 2012. [ Links ]

Torraca, L. “Il Cod. Vat. Barberinianus Gr. 87 e il testo delle Categorie di Aristotele”. Bollettino del Comitato per la preparazione della Edizione Nazionale dei Classici Greci e Latini, 11, pp. 91108, 1963. [ Links ]

Tuominen, M. “Back to Posterior Analytics II 19: Aristotle on the knowledge of principles”. Apeiron, 43, pp. 115-143, 2010. [ Links ]

Waitz, T. Aristotelis Organon Graece Vol. I-II. Leipzig: Hahn, 1844-46. [ Links ]

Wallies, M. Themistii Analyticorum Posteriorum Paraphrasis (CAG Vol. V.1). Berlin: Reimer , 1900. [ Links ]

Wallies, M. Ioannis Philoponi in Aristotelis Analytica Posteriora Commentaria cum Anonymo in Librum II (CAG Vol. XIII.3). Berlin: Reimer , 1909. [ Links ]

Walzer, R.. “New light on the Arabic translations of Aristotle”. Oriens, 6, pp. 91-142, 1953. [ Links ]

Williams, M. Studies in the Manuscript Tradition of Aristotle’s Analytica. Königstein: Hain, 1984. [ Links ]

1With Post. An. II.19 99b15-19 cf. esp. Pr. An. I.1 24a10-11, I.4 25b27-31.

2A comment about my rendering of Aristotle’s epistemic vocabulary is in order. In the Analytics, the Greek noun epistēmē almost always functions as a technical term naming a quite specific variety of high-grade knowledge that English doesn’t have a name for. Here, as in other writings, I introduce the new English noun ‘episteme’ as stand in for epistēmē in the technical sense at issue in Aristotle’s Analytics. (Note, however, that I always use adjectives like ‘epistemic’ and ‘epistemological’ in the ordinary sense of contemporary philosophical English: e.g. ‘epistemic’ ≈ ‘of or related to justificational/knowledge phenomena [of some kind]’, not ‘of or related to episteme [specifically]’). The word ‘knowledge’, of course, is an etymological relative of the Greek noun gnōsis—and not epistēmē. The semantic range of ‘knowledge’ in ordinary English is fairly close that of gnōsis in the Greek of our period; and Aristotle never assigns to gnōsis (as he does to epistēmē) any kind of special technical sense beyond its everyday linguistic meaning. Thus, e.g.: episteme, nous, technē, and phronēsis are all (according to Aristotle) types of gnōsis; he thinks knowing what ‘triangle’ signifies is gnōsis; he thinks (human and non-human) animals can have gnōsis through exercising their perceptual faculties; he thinks that an experienced person, qua experienced, has gnōsis that an unexperienced person lacks. For these and other reasons, I deploy ‘knowledge’ to translate occurrences of gnōsis in Aristotle and render its verbal/adjectival correlates accordingly: gignōskein as ‘know’, gnōstos as ‘known’. In the Analytics, Aristotle tends to use (i) the verbs gignōskein and eidenai, as well as (ii) the adjectives gnōstos and gnōrimos, as freely exchangeable synonyms. So I also employ ‘known’ to translate gnōrimos and ‘know’ to trans late eidenai as well as instances of gnōrizein where it functions as a synonym for gignōskein.


Unless otherwise noted: (i) the translations I print are my own, (ii) the Greek text my Analytics translations translate is that of Ross’ 1949 edition. Ross’ text for Post. An. II.19 is primarily based on the testimony of four 9-11th cent. Byzantine Post. An. manuscripts—sigla: A, B, d, n. Subsequent research has shown there to be significant extant text-witnesses for the Post. An. that Ross failed to exploit. Here I’ll be making additional use of the following “direct” Post. An. text-witnesses:

D = Par. gr. 1843 (12-13th cent. Post. An. MS)

V = Barb. gr. 87 (9-10th cent. Post. An. MS)

G = Guelf. 24 Gud. graec. (12-13th cent. Post. An. MS)

All three manuscripts are digitized and can be accessed online via the relevant library websites.

4περὶ μὲν οὖν συλλογισμοῦ ϰαὶ ἀποδείξεως, τί τε έϰάτερόν ἐστι ϰαὶ πῶς γίνεται, φανερόν, ἃμα δὲ ϰαὶ περὶ ἐπιστήης ἀποδειϰτιϰῆς ταὐτὁν γάρ ἐστιν. περὶ δὲ τῶν ἀρχῶν, πῶς τε γίνονται γνώριμοι ϰαὶ τίς ἡ γνωρίζουσα ἓξις, ἐντεῦϑεν ἐστι δῆλον προαπορήσασι πρῶτον.

5The illumination is the same because the philosophical strategy of the Analytics is to illuminate epistēmē apodeiktikē by illuminating apodeixis. 99b17’s t’auton gar esti (‘for they’re the same’) can’t be saying, because Aristotle doesn’t think, that apodeixis and epistēmē apodeiktikē are the same: see, e.g., Post. An. I.4 73a23-4, I.2 71b16-23.

6Reading the lectio difficilior esti (‘is’/‘are’) with A, B, d, D, V, G, Themistius, Bekker, Waitz where Ross prints estai (‘will be’) which is attested by n alone. Translating—and perhaps over-translating—Ross’ text, Barnes renders 99b18-19 (1993 ed. p. 72, my emphasis): ‘[the truth regarding (Q1) and (Q2)] this will be plain from what follows [enteuthen estai dēlon], when we have first set out the puzzles [proaporēsasi prōton]’. I take 99b18-19 to be making a weaker claim than that which Barnes’ translation seems to suggests. Prima facie, Barnes’ construal of enteuthen might seem more linguistically likely than mine; but I think my alternative philosophically superior given the actual discussion the remainder of Post. An. II.19 contains.

7Post. An. displays an especially strong preference for gnōrimos over gnōstos, using gnōstos at 82b38 but nowhere else. I take it that articulating (Q2), Aristotle is using the verb gnōrizein as a synonym for gignōskein/eidenai; I take it that he prefers gnōrizein over gignōskein/eidenai in articulating (Q2) because the verbal correlative of gnōrimos is gnōrizein and he’s just used gnōrimos in articulating (Q1).

8I submit that one reason why II.19’s replies to (Q1) and (Q2) are highly abstract is that Aristotle is abstracting away from (what he himself takes to be) significant differences between the epistemology of foundational knowledge in mathematics, physics, and metaphysics.

9On my own reading of the Post An., the core of Aristotle’s case for the former presupposition (= that demonstrative foundations exist) is provided by Post. An. I.19-22 while the core of his case for the latter presupposition (= that demonstrative foundations can be known with the relevant type of knowledge) is provided by Post. An. II.1-10. I elaborate and defend this controversial interpretation of the text in my “Aristotle’s Finest Theorem and the Project of his Analytics” (draft available on request).

10For scholarly endorsements of the assumption about II.19 I’m rejecting here see, e.g., Smith (1986, p. 55) and Tuominen (2010, p. 115).

11ὅτι μὲν οὖν οὺϰ ἐνδέχεται ἐπίστασϑαι δι’ ἀποδείξεως μὴ γιγνώσϰοντι τὰς πρώτας ἀρχὰς τὰς ἀμέσους, εἴρηται πρότερον. τῶν δ’ ἀμέσων τὴν γνῶσιν, ϰαὶ πότερον ἡ αὺτή ἐστιν ἢ οὺχ ἡ αὺτή, διαπορήσειεν ἄν τις, ϰαὶ πότερον ἐπιστήμη ἑϰατέρου ἢ οὔ ἢ τοῦ μὲν ἐπιστήμη τοῦ δ’ ἓτερόν τι γένος, ϰαὶ πότερον οὺϰ ἐνοῦσαι αί ἓξεις ἐγγίνονται ἢ ἐνοῦσαι λελήϑασιν. εί μὲν δή ἒχομεν αὺτάς, ἄτοπον· συμβαίνει γὰρ ἀϰριβεστέρας ἒχοντας γνώσεις ἀποδείξεως λανϑάνειν. εὶ δὲ λαμβάνομεν μή ἒχοντες πρότερον, πῶς ἄν γνωρίζοιμεν ϰαὶ μανϑάνοιμεν ἐϰ μἡ προϋπαρχούσης γνώσεως· ἀδύνατον γάρ, ὥσπερ ϰαὶ ἐπὶ τῆς ἀποδείξεως ἐλέγομεν.

12The apparent back reference looks to be Post. An. I.2 72a25 ff.

13Here, as in more than a few other Post. An. passages, the adjective amesos (‘unmediated’, ‘lacking a middle [meson]’) is basically a synonym of anapodeiktikos (‘indemonstrable’, ‘lacking a demonstration [apodeixis]’).


That Aristotle structured II.19 so that its train of thought would weave an elegant ABBA ABBA pattern is easy to see:

A: What are syllogism and demonstration? (99b15-16)

B: How do syllogism and demonstration come to be? (99b16)

B: How do the foundations of demonstration come to be known? [= question (Q1)] (99b17-18)

A: What is the knowledge that knows these foundations? [= question (Q2)] (99b18)

A: An aporia that picks up on (Q2) is set out [= aporia (a2)] (99b22-24)

B: An aporia that picks up on (Q1) is set out [= aporia (a1)] (99b25-30)

B: A view addressing (a1)+(Q1) is set out (99b30-100b5)

A: A view addressing (a2)+(Q2) is set out (100b5-17)

15For Aristotle on the investigative significance of aporiai, the locus classicus is Metaphysics B 995a24-b4. It bears recalling that many of the aporiai Met. B proceeds to rehearse (cf. Met. K.1-2) are not simply puzzles but puzzlement-inducing philosophical disagreements (i.e. one group asserts φ and denies ψ; another group asserts ψ and denies φ; neither position is obviously mistaken; there are compelling arguments on both sides; so, who is right? and who is wrong?; the truth is far from clear).

16That 100b5-17 addresses (Q2) by addressing (a2) is not difficult to see: all that 100b5-17 really says about (Q2) is that the knowing state at issue is something called nous which isn’t episteme. To appreciate that 99b30-100b5 likewise addresses (Q1) by addressing (a1) compare 99b25-34 with 100a10-11. The latter looks to be the central conclusion that 99b30-100a9 works toward. And 100ba12-100b5 presents itself as an elaboration of 100a10-11.

17Insofar as Aristotle wrote an On the Soul, 100a13-14’s un-explained assertion “the soul exists as the sort of being capable of undergoing this [progression]” looks like a pretty obvious invitation to consult it.

18As far as Post. An. II.19 goes, occurrences of the noun nous are limited to 100b5-17. That it’s an intellectual virtue called nous which is at issue in 100b5-17 is readily verified by studying 100b5-17 in the light of NE VI.1-3, 6 (NB 100b5-9’s employment of the tag “always true” [aei alēthē]). Outside of 100b5-17, we find the term nous in only three further passages in the whole of Aristotle’s Analytics: 85a1, 88b35-89a1, 89b8. What Aristotle means by nous in the latter three passages is exactly what he means by nous in 100b5-17.

19With aporia (a1) cf. Met. A.9 992b24-993a2.

20With Meno 81c5-7, cf. Phaedrus 247c2 ff.

21Socrates’ authoritative and careful presentation of the theory of recollection at Phaedo 73b3-76e7 is supposed to correct —and should be contrasted with—Cebes’ overly swift 73a7b2 sketch of the theory. For the latter, which unsubtly refers us to the Meno, is instructively inaccurate on several important points. Contra Cebes (Phd. 73a7-8), when Meno’s slave is questioned by Socrates he makes mistakes and importantly does not αὺτὸς λέγει πάντα ᾖ ἒχει (cf. Men. 82b9-85b). Contra Cebes (Phd. 73a8-10), the argument of Meno 81e3-86b5 does not in fact show that epistēmē and the correct logos are already present within [enousa] Meno’s slave before Socrates questions him (cf. Men. 85b8-d4, 86a7-8: ἐνέσονται αὺτῷ [= the slave] ἀληϑεῖς δόξαι αἴ ἐρωτήσει ἐπεγερϑεῖσαι ἐπιστῆμαι γίγνονται). Most basically: in giving his swift Phaedo 73a7-b2 account of recollection, Cebes has forgotten about the crucial role of “forgetting” (= the destruction of pre-natal knowledge) in the theory. Socrates’ corrective discussion (esp. Phd. 75c7-76d6) works to clarify—and strongly emphasizes—this latter component of the Platonic theory of recollection. Insofar as τὸ είδέναι and τὸ ἐπίστασϑαι is λαβόντα του ἐπιστήμην ἒχειν ϰαὶ μἡ ἀπολωλεϰέναι (75d8-10), it is strictly speaking both false and at odds with the theory of recollection to say that epistēmē of Forms is present within [enousa] Meno’s slave from birth.

22Cf. e.g.: Met. A.1 980a21 ff., Poet. 4 1448b5 ff., DA II.5, II.12, III.4-8.

23E.g. “poverty of stimulus” problems connected to questions like Why does the human intellect produce reliably true intuitions about mathematical objects (e.g. perfect squares) and unchanging essences (e.g. the nature of justice)? (NB that I’m using the term ‘intuition’ here in the loose and non-loaded sense favored by contemporary Anglophone philosophers).

24I contend, in sum, that the rejected alternative (ii) of aporia (a1) is not the theory of recollection. And I contend that if II.19 were to be read an to attempt to refute the theory of recollection and show the superiority of an Aristotelian alternative, we should have to judge its discussion as thoroughly inadequate and highly question-begging. Scholars are wrong to suppose that 99b25-34 refers to the Platonic theory of recollection. And scholars are wrong to interpret 99b26-27 as an argument against the theory of recollection. To further appreciate the implausibly of the latter interpretative thesis, recall that Plato agrees with Aristotle that it would be “strange” [atopon] and “wondrous” [thaumaston] (Post. An. 99b26, Met. 993a1) if the theory of recollection were true. Plato’s basic idea in texts like the Meno and Phaedo is that the theory of recollection is a hypothesis worth adopting because (i) it solves certain hard philosophical problems, and (ii) upon investigation, it’s not as implausible as it may first seem. Finally, a remark about 71a29 is in order. The Post. An. I.1 71a29 phrase to en toī Menōni aporēma refers neither to the puzzle which Post. An. I.1 71a17-31 claims to solve [luein 71a31] nor to what we call “Meno’s paradox”. The latter is a puzzle argument—an apparently valid argument from apparently true premises to an apparently absurd conclusion; the puzzle which 71a17-31 claims to solve is a different puzzle entirely; and the referent of to en toī Menōni aporēma at 71a29 isn’t an argument at all but a paralyzing objection—an absurd consequence that allegedly results [sumbainei 71a30] if one rejects Aristotle’s proposed solution to the puzzle which Post. An.I.1 71a17-31 claims to solve. This absurd consequence is analogous to, but not identical with, the absurd conclusion of the “eristic logos” (Meno 80d-e) that we call Meno’s paradox. The sense of to en tōi Menōni aporēma needs to be understood accordingly. Neither Post. An. I.1, nor any other Analytics text, presents Aristotle’s response to the puzzle argument (we call) Meno’s paradox.

25Like Bronstein (2016) and others, I think the Met. A.1 parallel (980a27 ff.) tells strongly in favor of taking logos at 100a2 to mean reason in the sense of rationality (NB logismois at 980b28). Barnes’ case against this construal of logos at 100a2 is not convincing. For further discussion I esp. recommend Gregorić and Grgić (2006, pp. 21-4), Gasser-Wingate (2016, pp. 7-8).

26Recall that empeiria (“experience”) is Aristotle’s name for the type of knowledge we ascribe when we call a person “experienced” (empeirios) in or with respect to something.

27Or rather: today this much is basically uncontroversial. While the vast majority of 20-21st cent. scholars construe 100a6’s pantos with tou katholou in the very manner that Barnes does, Pseudo-Philoponus construes 100a6-9 in a strikingly different way (cf. the translation of 100a3-9 in Bolton and Code (2012) p. 67). Pseudo-Philoponus’ gloss on 100a6 explains (CAG 13.3 436,2-5): τὸ ἢ ἀντὶ τοῦ ϰαὶ ληπτέον. ἔστι δὲ τοιοῦτον· ἐϰ δὲ τῆς ἐμπειρίας ϰαὶ ἐϰ παντός αίσϑήματος τοῦ ἠρεμήσαντος ἐν τῇ ψυχῇ ϰαὶ ἑδραιωϑέντος γίνεται ἡ γνῶσις τοῦ ϰαϑόλου τοῦ ἐντός ὄντος παρά τὰ πολλὰ. Taking pantos to modify an implied aisthēmatos, and pantos aisthēmatos as the subject of ēremēsantos en tēi psuchēi psuchēi, Pseudo-Philoponus seems to effectively read tou katholou as an objective genitive (governed by an implied gnōsis if not technēs archē kai epistēmēs at 100a8)!

28Some authors prefer “corrective” as an alternative label for the construal that McKirahan calls “progressive”; nothing much turns on this.


For a taste of this, one can compare (e.g.) Bronstein’s version of the “progressive” reading with Sorabji’s version of the “explicative”. In a 2010 article Sorabji writes:

[In Post. An. II.19] Aristotle is providing an alternative to Plato by showing how concepts can be formed on the basis of sense perception. There is in that case, he thinks, no need for Plato’s alternative, argued in Phaedo 72e-77a, of concepts stored in the mind from the soul’s existence before birth. In 100a3-8, Aristotle says that many memories of the same type of perceived thing, let us say of oxen, constitute experience (empeiria) of oxen. And then, on my interpretation, he uses the word ‘or’ (ē), to equate experience, or many memories, with a rudimentary universal concept of oxen. At any rate, he speaks of experience, or (ē) the whole universal (katholou) in the soul. Admittedly, the word ‘or’ can mean ‘or rather’. But if he were here talking of experience or rather the whole universal [as Aristotle would be on the “progressive” reading], he would have left unexplained the very thing he is trying to explain, how, contrary to Plato’s view, remembered sense perceptions are enough to give us at least a rudimentary universal concept. The explanation is that to have a rudimentary universal concept of oxen just is to have enough memories of oxen to react with experience to them. It is not a further step beyond many memories. (Sorabji 2010, p.3)

Pace Sorajbi, his “explicative” construal of 100a6 does not in fact yield a text that (my emphasis) “explains how, contrary to Plato’s view, remembered sense perceptions are enough to give us at least a rudimentary universal concept”. If Post. An. II.19 was (which I dispute) written to accomplish the philosophical task Sorabji describes, the effect of adopting his “explicative” reading of ē ek pantos would be a flagrant begging of the question. Sorabji’s case in the passage above that the “explicative” reading ought to be preferred because it’s philosophically superior to the “progressive” is hardly compelling.

While Sorabji’s reading of Post. An. II.19 requires adopting the “explicative” reading of 100a6’s ē ek pantos and rejecting the “progressive” alternative, Bronstein’s depends on adopting the “progressive” reading and rejecting the “explicative” alternative:

[In 100a3-9] Aristotle identifies four stages prior to nous which he here calls the “principle of craft and scientific knowledge”: perception, memory, experience, and the grasp of “the entire universal”. By ‘universal’ Aristotle means, I take it, a proposition of the form ‘all As are B’. We reach this universal by induction which he discusses [at 100a14b5]. However, knowing this universal is not equivalent to knowing a first principle. Aristotle is clear that nous, the state we are in when we know first principles, comes after knowing this universal. [...] Nor is it the case that we reach knowledge of first principles directly after grasping this universal. Rather, I suggest that the universal we reach by induction is a preliminary account required for scientific inquiry. Aristotle omits the important stages between grasping the universal and nous, and so there is a gap in his account. However, this gap is not filled by further inductive activity alone or by the allegedly intuitive activity of [some faculty of] nous; rather, it is filled by the methods of inquiry he sets out earlier in Book 2. (Bronstein 2016, pp. 236-237)

Proponents of the “progressive” reading have set out various competing proposals regarding (i) what kind of cognition “the universal having come to rest in the soul” amounts to, (ii) how nous is supposed to emerge from this kind of cognition, and (iii) how this kind of cognition differs from experience. Bronstein’s own account of (i)-(iii) is highly speculative. Alternative accounts of (i)-(iii) given by other proponents of the “progressive” reading differ and tend to be no less speculative than Bronstein’s.


Hasper and Yurdin, for instance, develop this line as follows:

A sentence of the form ‘from x or from y, z comes about’ is more plausibly interpreted as meaning either that x and y are two separate sources for z or that ‘y’ is an alternative way of specifying or picking out the same thing as is picked out by ‘x’, one that makes it clear what aspect of x is relevant to its being a source for z. As the former does not apply here [i.e. the former gives an implausible reading of 100a6’s ē ek pantos...] Aristotle is using ‘experience or a universal having come to rest as a whole in the soul’ in the latter way. (Hasper and Yurdin 2014, pp. 122-3 n.7)

31I’m translating the Met. A.1 980a27-981a16 text of Primavesi (2012): φύσει μὲν οὖν αἴσϑησιν ἒχοντα γίγνεται τὰ ζῷα, ἐϰ δὲ τῆς αίσϑήσεως τοῖς μὲν αὺτῶν οὺϰ ἐγγίγνεται υνήμη, τοῖς δὲ γίγνεται. [...] τὰ μὲν οὖν ἄλλα ταῖς φαντασίαις ζῇ ϰαὶ ταῖς μυήμαις, ἐμπειρίας δὲ μετέχει μιϰρόν·τὸ δέ τῶν ἀνϑρώπων γένος ϰαὶ τέχνῃ ϰαὶ λογισμοῖς. γίγνεται δ’ ἐϰ τῆς μνήμης ἐμπειρία τοῖς ἀνϑρώποις· αί γὰρ πολλαὶ μνῆμαι τοῦ αὺτοῦ πράγματος μιᾶς ἐμπειρίας δύναμιν ἀποτελοῦσιν. ϰαὶ δοϰεῖ σχεδὸν ἐπιστήμῃ ϰαὶ τέχνῃ ὄμοιον εἶναι ἡ ἐμπειρία, ἀποβαίνει δ’ ἐπιστήμη ϰαὶ τέχνη διὰ τῆς ἐμπειρίας τοῖς ἀνϑρώποις·ἡ μέν γὰρ ἐμπειρία τέχνην ἐποίησεν, ὡς φησὶ Πῶλος ὀρϑῶς λέγων, ἡ δ’ ἀπειρία τύχην. γίγνεται δὲ τέχνη ὄταν ἐϰ πολλῶν τῆς ἐμπειρίας ἐννοημάτων μία ϰαϑόλου γένηται περὶ τῶν ὁμοίων ύπόληψις. γίγνεται δέ τέχνη ὃταν ἐϰ πολλῶν τῆς ἐμπειρίας ἐννοημάτων ϰαϑόλου μία γένηται περὶ τῶν ὁμοίων υπόληψις. τὁ μἐν γὰρ ἒχειν ὑπόληψιν ὃτι Καλλίᾳ ϰμάνοντι τηνδὶ τἡν νόσον τοδὶ συνήνεγϰε ϰαὶ Σωϰράτει ϰαὶ ϰαϑ’ ἓϰαστον οὕτω πολλοῖς, έμπειρίας ἐστίν· τό δ’ ὅτι πᾶσι τοῖς τοιοῖσδε ϰατ’ εἶδος ἓν ἀφορισϑεῖσι, ϰάμνουσι τηνδὶ τὴν νόσον, συνήνεγϰεν, οἶον τοῖς φλεγματώδεσιν ἢ χολώδεσι πυρέττουσι ϰαύσῴ, τέχνης. πρὸς μὲν οὖν τὸ πράττειν ἐμπειρία τέχνης οὺδὲν δοϰεῖ διαφέρειν, ἀλλὰ ϰαὶ μᾶλλον ἐπιτυγχάνοντας ὁρῶμεν τούς ἐμπείρους τῶν ἄνευ τῆς έμπειρίας λόγον ἐχόντων· αἴτιον δ’ ὃτι ἡ μὲν ἐμπειρία τῶν ϰαϑ’ ἓϰαστόν ἐστι γνῶσις ἡ δὲ τέχνη τῶν ϰαϑόλου...

32In view of the discussion’s practically exclusive focus on technē, I’ve translated epistēmē kai technē as a hendiadys. Effectively, I take the relevant epistēmē kai to emphasize that it’s intellectual/scientific aspect of technē which is at issue here.

33For sigla see note 3 above. The especially significant evidentiary value of D (= Par. gr. 1843) for the Post. An. has been stressed by Brockmann (2004) and is confirmed in my “Towards a Text History of Aristotle’s Analytics” (draft available on request). Among the most important results of Brockmann 2004 is that folios 127-207v of manuscript D give us a careful copy of the (now highly fragmentary) 9th-10th cent. manuscript Sin. gr. M138 newly discovered in 1975 (on which see Reinsch 2001). Folios 127-207v of D cover Pr. An. 30b37 Post. An. end; unfortunately, Post. An. II.19 is not extant in surviving fragments of Sin. gr. M138.

34The sense in which Ross was “ignorant” of manuscript D is qualified and merits comment. The Organon siglum ‘D’ actually goes back to the 1831 Organon edition of Bekker whose (very sparse) apparatus uses ‘D’ to index selected variants Bekker found in the BnF manuscript Par. gr. 1843. Unfortunately, instead of (correctly) stating that D = Par. gr. 1843, the front matter to Bekker’s edition incorrectly reports that D = Cois.170. But as Waitz pointed out in the introduction to his own 1844-46 enlarged critical edition, the latter BnF manuscript (i.e. Cois. 170) doesn’t in fact contain any of the Organon. The true codicological identity of Bekker’s D was to remain a scholarly mystery for next 100+ years. Now, as Waitz did in his 1844-46 Organon edition, Ross includes some of Bekker’s 1831 reports of D variants in the apparatus to his 1949 Analytics edition. But confusing matters even further, the front matter of Ross 1949 falsely prints that D = Cois. 157, thus misidentifying Bekker’s D with another BnF manuscript—one which, unlike Cois. 170, does at least contain the Organon. That the manuscript corresponding to Bekker’s siglum ‘D’ is in fact Par. gr. 1843 was finally discovered by H. D. Saffrey about a decade after Ross’ 1949Analytics edition was published. (Ross credits Saffrey with this discovery in his 1959 Topics OCT).

35For an insightful and pleasantly concise discussion of the Greek tradition of Post. An. commentaries, I esp. recommend Ebbesen 2012.

36Abū Bishr’s Arabic Post. An. translation is extant in the celebrated 11th cent. scholarly manuscript Par. ar. 2346 (edited by Badawī 1948-52). It was most immediately based on the Syriac Post. An. of Isḥāq Ibn Ḥunayn (d. 910/911). Unfortunately, no other medieval Arabic translations of Post. An.II.19 are presently extant in Arabic. As Walzer 1953’s comparative study of Abū Bishr’s translation and manuscripts A, B, d, n already suggests, the Greek manuscript(s) on which Abū Bishr’s Arabic Post. An. is ultimately based look to be β-family. For a concise introduction to 20th cent. scholarship on Arabic/Syriac Post. An. translations and related issues, see the entry “Aristote de Stagire, Organon—Tradition syriaque et arabe” in Goulet (1989).


Abū Bishr translates 100a3-9 as follows:


The 13th cent. revision of James’ translation by William of Moerbeke (see AL IV.4) gives us a Latin 100a6 perfectly identical with James’ original. The so-called “Translatio Anonyuma sive ‘Ioannis”’ (see AL IV.2) gives us a Latin 100a6 minimally different from James’: ex experimento autem illud quod aut ex omni quiescente universali in anima. We noted above that Abū Bishr’s is the only medieval Arabic translation of Post. An. II.19 presently extant in Arabic. The 12th cent. Latin Post. An. translation of Gerard of Cremona (see AL IV.3) is thought to be based on an Arabic text for the Post. An. substantially different from Abū Bishr’s. It’s not entirely clear whether the latter Arabic text is best regarded as a later reworking of Abū Bishr’s translation, or a fresh translation rendered in a far freer and more paraphrastic style than Abū Bishr’s. But in any case, it’s clear enough that compared to Abū Bishr’s Arabic Post. An., Gerard’s highly paraphrastic Arabo-Latin Post. An. is of significantly less evidential value for establishing the original Greek. Gerard’s rendering of 100a3-8 reads (NB the lack of an aut ex omni):

Quod est quia virtus ymaginativa et memorialis non suscipiunt eas nisi ex virtute sensibili; et virtus memorialis, quando servat formas rerum una vice post aliam, lubratur ex eis assimilationem rebus, ex huiusmodi assimilationibus comprehensis pervenit universale anime. Et universale quidem est forma una procedens ex formis multis, et est existens in unaquaque earum; et hec forma est una eadem, et est principium scientie et artis.

39For purposes of this sentence, ē ek pantos is a stand in for the letter string HEKΠANTOΣ of which both ἢ ἐϰ παντὸς and ἡ ἐϰ παντὸς would be instances.

40The following paragraph has greatly benefitted from conversations with David Blank.

41Phd. 96a6-c1 (NB 96b6-8 which I’ve bolded): ἐγὼ γάρ, ἔφη, ὦ Κέβης, νέος ὢν ϑαυμαστῶς ὡς ἐπεϑύμησα ταύτης τῆς σοφίας ἣν δὴ ϰαλοῦσι περὶ φύσεως ίστορίαν· ὑπερήφανος γάρ μοι ἐδόϰει εἶναι, είδέναι τὰς αἰτίας ἑϰάστου, διἁ τί γίγνεται ἓϰαστον ϰαὶ διἁ τί ἀπόλλυται ϰαὶ διἁ τί ἔστι. ϰαὶ πολλάκις ἐμαυτὸν ἄνω ϰάτω μετέβαλλον σϰοπῶν πρῶτον τὰ τοιάδε· ἆρ’ ἐπειδἁν τὸ ϑερμὸν ϰαὶ τὸ ψυχρὸν σηπεδόνα τινὰ λάβῃ, ὣς τινες ἔλεγον, τότε δὴ τὰ ζῷα συντρέφεται· ϰαὶ πότερον τό αἶμά ἐστιν ᾧ φρονοῦμεν, ἢ ὁ ἀὴρ τὸ ἢ πῦρ· ἢ τούτων μὲν οὑδέν, ὁ δ’ ἐγϰέφαλός ἐστιν ὁ τὰς αίσϑήσεις παρέχων τοῦ ἀϰούειν ϰαὶ ὸρᾶν ϰαὶ ὀσφραίνεσϑαι, ἐϰ τούτων δὲ γίγνοιτο μνήμη ϰαὶ δόξα, ἐϰ δὲ μνήμης ϰαὶ δόξης λαβούσης τὸ ἠρεμεῖν, ϰατὰ ταῦτα γίγνεσϑαι ἐπιστήμην· ϰαὶ αὖ τούτων τὰς φϑορὰς σϰοπῶν, ϰαὶ τὰ περὶ τὸν οὑρανόν τε ϰαὶ τὴν γῆν πάϑη, τελευτῶν οὕτως ἐυαυτῷ ἒδοξα πρὸς ταύτην τὴν σϰέψιν ἀφυὴς εῖναι ὡς οὑδὲν χρῆμα.

42See, e.g. DA I.3 407a32 ff., Phys. VII.3 247b9 ff.

43Further evidence against this proposal is that the items 100a15-100b5 calls “universal” are things like animal and human being.

44I’m supplying ὣς τινες ἔλεγον (“as some were claiming”) from 96b3 as is grammatically necessary. Note the accusativecum-infinitive of indirect speech γίγνεσϑαι ἐπιστήμην at 96b8, and that γίγνοιτο at 96b7 is an optative of indirect discourse from the past.

45NB Metaphysics A.1’s 981a3-5 reference to the historical rhetorician Polus and/or the character Polus at Gorgias 448c; with Metaphysics A.1’s 981a7 ff. remarks on the epistemology of medicine cf. esp. Laws IV 720a ff., IX 857c ff. In speaking of Post. An. 100a3-9 as distilling endoxa concerning how humans develop intellectual virtue, I don’t mean to suggest—for, I see no reason to think—that Aristotle’s endorsement of 100a3-9 is “merely dialectical” or qualified in any such way.

46We have not, e.g., discussed the important question of what’s meant by epagōgē at 100b4. For what it’s worth, I think epagōgē at 100b4 does not mean what it does in the Topics; and I think that while translating epagōgē at 100b4 as “induction” can help illuminate later episodes in the history of philosophy, this translation significantly obscures what Aristotle himself is trying to say.

Article Info: CDD:185

*Parts of this article has been presented to audiences at the Universidade Estadual de Campinas (UNICAMP), HumboldtUniverität zu Berlin, and the UCLA “Aristotle Bash”. The text that follows has benefitted greatly these discussions, as well as written comments from Lucas Angioni, Fernanda Izidorio, Whitney Schwab, and Breno Zuppolini.

Recebido: 03 de Outubro de 2019; Aceito: 07 de Outubro de 2019

Creative Commons License This is an open-access article distributed under the terms of the Creative Commons Attribution License