1. Status quaestionis: two recensions or two appendiXes?

The non-standard version is attested in the oldest extant Greek manuscript to preserve Aristotle’s De anima, Parisinus E. The manuscript, commonly dated to the tenth century, has been acknowledged since Bekker’s seminal 1831 edition as a key witness.Footnote ¹ The text it contains for Book 2 of De anima has a peculiar and complicated history. Random isolated bits of Aristotle’s De anima 2 are found at the very beginning in the present-day manuscript (ff. 1–2), while the title and beginning of Book 2 appear twice (ff. 186^v and 187^r), as well as its end (ff. 195^rv and 196^rv): these are sure signs that the codex has undergone some major reorganization. The process can be reconstructed as follows:Footnote ²

• t ₀ , middle of the tenth century: The manuscript as originally produced has one single recension for the whole of Aristotle’s De anima, copied by the same hand (E I).

• t ₁ , second half of the tenth century: An early owner of the codex (hand E ² ) notices that there is something unusual about its text for Book 2 and leaves some annotations in the margin around its beginning (f. 186^v: the beginning of the text in the vulgate version is written down in the external margin, along with a reminder to look for its standard version after the title of the second book).

• t ₂ , second half of the tenth century: This same owner decides to have the original recension of Book 2 in the codex supplemented by a different one reflecting the vulgate text, copied by a different hand (E II).Footnote ³ The original recension is moved to the end of the treatise (as signalled by a new annotation in the internal margin of f. 186^v). The codex now presents the following sequence: original recension of Book 1 (now ff. 175^v–86^v), original recension of Book 2 (now ff. 187^r–95^v), original recension of Book 3 (ff. 196^v–202^v), new recension of Book 2 (after what now corresponds to f. 202^v).

• t ₃ , beginning of the fourteenth century: A later owner of the codex finds that having two recensions of Book 2 in the same codex is a waste of precious parchment, moves the new recension of Book 2 between the original recensions of Book 1 and 3, and has the original recension of Book 2 recycled or eliminated.

As a result, the recension of Book 2 of Aristotle’s De anima which was originally present in E (henceforth ‘the original recension of E’) has almost completely disappeared from the codex in its present state. It only survives in what Ross calls ‘four scraps’: two folia have been recycled during t ₃ as a cover with no regard for the text they contain (now ff. 1–2, 414b13–416a10 and 421a5–422a24), while the beginning of a torn-down folium subsists where the first (on the recto) and final (on the verso) letters of each line are still visible (f. 187^bis, 412a12–413b1). The beginning (f. 186^v, 412a3–12) and end (f. 196^rv, 423b8–424b18) of Book 2 of Aristotle’s De anima have also been preserved in the original recension because they had been written on the same folia as the end of Book 1 and beginning of Book 3 and could not be removed without textual losses.

The two recensions in E offer significantly different texts, at least for the sections for which both have been preserved. There is general agreement among editors on the observation that the recensio recentior which has come to replace the original one reflects the same text that is attested in the vast majority of the manuscript transmission (including all other manuscripts used by nineteenth-century editors) and on which ancient authors rely (including Alexander of Aphrodisias, Themistius, ‘Simplicius’,Footnote ⁴ Philoponus). Two opposite attitudes have prevailed regarding the unusual text offered by the original recension.

This theory was first put forward by Torstrik in 1862. Out of the two recensions found in E, Torstrik regards the original and now fragmentary one as equally genuine. His main (and questionable) argument is that the original recension must be granted the same authority as the recensions of Books 1 and 3 by the same hand in the codex, assuming that they all have the same source. Since Torstrik regards E as the most important manuscript witness for the other two books (as Bekker and Trendelenburg had done before him), the conclusion is a ringing endorsement of its original recension.Footnote ⁵ Consequently, the manuscript would reflect a ‘double recension’ of Book 2 of De anima, the ancient coexistence of two versions both authored by Aristotle in the course of his revisions of his own work.

This cautious strategy is used by Ross in his two editions (1956 and 1961), following the 1912 edition of Förster.Footnote ⁶ Ross does not explain his decision in his terse 1956 Latin preface or in his 1961 English introduction. Förster’s preface first provides a description of the situation in E with ample reference to Torstrik, until the editor explicitly suspends judgement about the status of the original recension of Book 2 in E in the very last sentence.Footnote ⁷

The same appendix-strategy is used again in Förster’s edition and in Ross’s 1961 edition for a different manuscript, Vaticanus graecus 1339 P, now dated to the early fourteenth centuryFootnote ⁸ . Manuscript P had already been examined by Bekker and Brandis, without being admitted in the critical apparatus of Bekker for De anima. The manuscript was brought to light only when its complete text for Book 2 (and only for Book 2) was published in 1891 by Rabe.Footnote ⁹ Again, Förster remains tentative and non-committal in his preface (P ‘occupies a sort of intermediary position’ between the two main families of the transmission [page xvii]) and is closely followed by Ross.Footnote ¹⁰

Rabe had not been as vague. His preface offers a clear-headed assessment of the status of P, claiming that its text of Book 2 of De anima is an elaborate mixture of the vulgate text and of the same version that subsists in the fragments of the original recension of E.Footnote ¹¹ If Rabe is right, a proper edition of Aristotle’s De anima should start with a clear statement of the editor’s view regarding the status of the version of the text of Book 2 found in the fragments of E and partly in P. If it is found to be valuable for reconstructing Aristotle’s text, their readings belong, if not to the main text, to the critical apparatus; if not, they should at most be relegated to one single appendix.

The text found in P for Books 1 and 3 was first studied by de Corte in 1933, who reached a different conclusion from Rabe, claiming that, far from being a mixture of different strands, the text of P reflects for the whole treatise an early state of the tradition predating its main splits, thus presenting P as the most important manuscript witness to Aristotle’s text, above even E.Footnote ¹² Siwek, aiming at an exhaustive study of the manuscripts of Aristotle’s De anima, examined the issue once more in 1965. He places P within the same family (siglum λ) as E and four other manuscripts for Book 3 and notes (with Rabe) peculiar agreements between the original recension of E and the text of P for Book 2.Footnote ¹³ His final verdict, however, is confused, as Siwek ends up claiming that one of the fragments of the original recension of Book 2 in E derives from P,Footnote ¹⁴ although he had ruled out this hypothesis for obvious chronological reasons.Footnote ¹⁵ A full report of the readings of P is made available for the first time in the 1966 Budé edition of Jannone, who does not venture to make any claim about its status.Footnote ¹⁶

2. Sources for the non-standard version

Even a quick comparison between the texts of De anima 2 found in E and P is sufficient to establish that there are important commonalities. The text of Book 2 in P, even if it is not completely identical with the text preserved in the fragments of the original recension of E, still has numerous points of agreement with it against the vulgate. Here are two examples.

The same text which is found in P, for these two passages as well as for the whole of Book 2 (barring scribal errors), is preserved in another manuscript hitherto neglected by scholars, now held in Moscow in the library of the Historical State Museum (Государственный Исторический Музей) with the shelfmark Sinodicus graecus 240 (M^o).Footnote ²² M^o has been copied in the third quarter of the fifteenth century (shortly after the Ottoman conquest of Constantinople) by Matthaios KamariôtesFootnote ²³ and is thus much more recent than P. It is highly likely, however, that the two manuscripts are mutually independent, due to each having errors of some separative value against the other,Footnote ²⁴ and that they both derive from a common ancestor for their texts of De anima 2 that they do not share with any other extant manuscript.

Sophonias’ paraphrase (late thirteenth century)Footnote ²⁵ regularly includes bits from Aristotle’s text that correspond to the unique readings of P and M ^o and, when available, of the original recension of E. Compare, for example, the text of Sophonias’ paraphrase in Hayduck’s CAG edition for the first passage quoted above; Sophonias seems to be generously relying on the non-standard version attested in manuscripts E, P and M ^o .Footnote ²⁶

This may give rise to the suspicion that the non-standard version could be a mixture of the vulgate text with Sophonias’ paraphrase.Footnote ³¹ Careful analysis shows that the most likely scenario is that Sophonias is reacting to the non-standard version, occasionally making alterations and adding improvements of his own. Consider the following passage from the end of Book 2, for which the original recension of E happens to be still available.

The non-standard version adds an unexpected supplement to the vulgate text, printed above in bold: ὡς τῆς αἰσθήσεως οἷον μεσότητός τινος οὔσης τῆς ἐν τοῖς αἰσθητοῖς ἐναντιώσεως (‘as sense-perception is like a sort of mean within the contrariety in the sense-objects’, meaning here that our tactile sense of temperature is such as to be activated only by tangible objects that are colder or warmer than its norm) becomes ὡς ἂν τοῦ αἰσθητηρίου τῆς αἰσθήσεως οἷον μεσότητός τινος οὔσης τῆς ἐν τοῖς αἰσθητοῖς ἐναντιώσεως, with three extra words. As it is hardly possible to make sense of the words ἂν τοῦ αἰσθητηρίου within the clause, they must be an intrusion.Footnote ³²

This supplement is found in Sophonias’ paraphrase, where the clause has been rearranged: it becomes ὡς ἂν τοῦ αἰσθητηρίου διὰ τὴν αἴσθησιν οἷον μεσότητός τινος ὄντος τῆς ἐν τοῖς αἰσθητοῖς ἐναντιώσεως, ‘as if the sense-organ, due to sense-perception, were a sort of mean within the contrariety in the sense-objects’. The participle has become in Sophonias the neuter ὄντος, agreeing with τοῦ αἰσθητηρίου, while τῆς αἰσθήσεως, which governs the participle οὔσης in both the vulgate and the non-standard version, is turned into διὰ τὴν αἴσθησιν. Chances are that Sophonias was embarrassed by the supplement found in the non-standard version and decided to slightly rewrite the text in order to make it syntactically palatable.Footnote ³³ Sophonias, then, is probably reacting to the non-standard version, which shows that the testimonies of manuscripts E, P and M ^o do not derive from his paraphrase, but that Sophonias, despite his innovations, ultimately depends on the same source.

How should one explain the agreements between these three manuscripts and Sophonias’ paraphrase? Assuming for now that the non-standard version is a by-product of the vulgate text and given that E is by far the oldest of these four witnesses, one plausible hypothesis would be that P, M ^o and a (presumably lost) manuscript used by Sophonias all derive from the original recension of E before it was eliminated from the codex. This, however, would be too simple. I shall leave aside the testimony of Sophonias, because there are clear signs that Sophonias had access to various source materials.Footnote ³⁴ Many features of the original recension of E that should have been reproduced in a normal copying process are not found in manuscripts P and M ^o , which therefore cannot be immediate copies of it.

The texts of P and M ^o are regularly found to be closer to the vulgate than to the original recension of E.Footnote ³⁵ For instance, in 414b28 P and M ^o read παραπλησίως δ’ ἔχει τῷ περὶ τῶν σχημάτων with the vulgate while E reads παραπλησίως δὲ ὥσπερ καὶ ἐπὶ τῶν σχημάτων ἔχει; in 415a27 P and M ^o (and Sophonias, 56.35) read ὅσα τέλεια καὶ μὴ πηρώματα with the vulgate while E reads ὅσα μὴ ἀτελῆ ἢ πηρώματά ἐστιν; in 421a27 P and M ^o (and Sophonias, 91.31) read οὕτω καὶ ὀσμαί with the vulgate while E reads καὶ ὀσμαὶ τὸν αὐτὸν ἔχουσαι τρόπον. Such evidence is not incompatible with the hypothesis that the exclusive ancestor of P and M ^o derives from the original recension of E; it may be that this ancestor, while still stemming from E, has been contaminated with readings from the vulgate.

The assumption of contamination is strengthened by the sporadic presence in M ^o of variant readings copied by the original hand in the margin or above the line, which usually record the vulgate reading against the non-standard one.Footnote ³⁶ There is no such phenomenon in P which had originally almost no paratext, but in view of the seamless way in which these variant readings have been integrated in M ^o , it is likely that the contamination occurred in their ancestor. Some readings shared by P and M ^o may accordingly be explained as combinations of the non-standard reading with the vulgate one. For example, in 415a9, the vulgate reads οἷς μὲν γὰρ ὑπάρχει λογισμὸς τῶν φθαρτῶν, τούτοις καὶ τὰ λοιπὰ πάντα. By contrast, the original recension of E reads οἷς μὲν γὰρ ὑπάρχει λογισμός, καὶ τῶν ἄλλων ἕκαστον τῶν εἰρημένων. P and M ^o both read οἷς μὲν γὰρ ὑπάρχει λογισμός, τούτοις καὶ τὰ λοιπὰ πάντα τῶν εἰρημένων,Footnote ³⁷ sharing with E the absence of τῶν φθαρτῶν and the presence of τῶν εἰρημένων at the end while sharing with the vulgate the beginning of the main clause τούτοις καὶ τὰ λοιπὰ πάντα.

As the original recension of E is only available for limited sections of text, it is difficult to determine with sufficient certainty whether what has been combined with the vulgate in the ancestor exclusive to P and M ^o goes back to the text of the original recension of E itself or to something related.Footnote ³⁸ The general upshot is that the absence of the original recension of E for the bulk of Book 2 of Aristotle’s De anima is no reason to panic. Three other witnesses (P, M ^o and, to a lesser extent, Sophonias’ paraphrase) that are available for the whole text have been observed to testify to the same non-standard version. As they have been contaminated with the vulgate, there might exist additional peculiarities of the non-standard version that are not reflected in these three sources (but would have been in the original recension of E). Be that as it may, the extant evidence is still enough to form an informed judgement about its nature.

3. The non-standard version in the light of the vulgate and of the ancient exegetical tradition

Although I have focussed so far on some of the most spectacular discrepancies, the non-standard version is not wholly different from the vulgate text. Even when the original recension of E is available, some stretches of texts remain rigorously identical in the two versions (for instance 422a10–11 καὶ τὸ σῶμα … ἁπτόν τι), or at least close, the differences amounting only to a matter of word order (αἰσθανοίμεθα τῶν ἁπτῶν ἁπάντων in the vulgate vs ἁπάντων αἰσθανοίμεθα τῶν ἁπτῶν in the non-standard version, 423b9) or to the presence or absence of a few short words (ὅλως δ’ ἔοικεν ἡ σὰρξ καὶ ἡ γλῶττα in the vulgate vs ὅλως δ’ ἔοικε καὶ ἡ σὰρξ καὶ ἡ γλῶττα in the non-standard version, 423b17).Footnote ³⁹ Given that the non-standard version thus regularly coincides with the vulgate, the two versions must have a shared origin. Even when the two versions do not coincide, the differences typically have no bearing on the text’s general structure or meaning.

Various theoretical possibilities could be envisaged as to the relationship between the two versions. Without any contamination hypothesis, the main options are that one version stems from the other, or that the two versions are mutually independent and derive from a common source.

The most likely option in view of the evidence is option II above: the non-standard version derives from the vulgate. The main reason is that the non-standard version tends to offer a smoother and somewhat simpler text, suggesting that it results from some sort of processing of the vulgate text. For instance, in 424b3–5 the sentence ἀπορήσειε δ’ ἄν τις εἰ πάθοι ἄν τι ὑπ’ ὀσμῆς τὸ ἀδύνατον ὀσφρανθῆναι ἢ ὑπὸ χρώματος τὸ μὴ δυνάμενον ἰδεῖν becomes in the non-standard version (attested in E, P and M ^o ) ἀπορήσειε δ’ ἄν τις ἆρα πάθοι ἄν τι ὑπ’ ὀσμῆς τὸ μὴ δυνάμενον ὀσφρανθῆναι ἢ ὑπὸ χρώματος τὸ μὴ δυνάμενον ἰδεῖν. Aristotle’s slight wavering from τὸ ἀδύνατον ὀσφρανθῆναι to τὸ μὴ δυνάμενον ἰδεῖν has been smoothed over. In 421a7 (quoted above), the vulgate has περὶ δὲ ὀσμῆς καὶ ὀσφραντοῦ ἧττον εὐδιόριστόν ἐστι τῶν εἰρημένων· οὐ γὰρ δῆλον ποῖόν τί ἐστιν ἡ ὀσμή and the non-standard version περὶ δὲ ὀσμῆς καὶ τοῦ ὀσφραντοῦ οὐκ ἔστι ῥᾴδιον διορίσαι ὁμοίως τοῖς εἰρημένοις αἰσθητοῖς τί ἐστιν ἡ ὀσμή: the elaborate highlighting of the difficulty of the topic (ἧττον εὐδιόριστόν ἐστι) is replaced with a more straightforward turn of phrase (οὐκ ἔστι ῥᾴδιον διορίσαι), and the vague τῶν εἰρημένων is made explicit.

The phenomenon is more pronounced in the original recension of E than in P and M ^o (and in Sophonias’ paraphrase, when it remains sufficiently close to Aristotle’s text), in all likelihood because the latter have been contaminated with the vulgate: in 414b31 οἷον in P, M ^o and the vulgate becomes λέγω δὲ ὥσπερ in E; in 414b32 ζητητέον in P, M ^o and the vulgate becomes δεῖ ζητεῖν in E; in 415a29–415b1 ᾗ δύνανται in P, M ^o and the vulgate becomes ὃν δύνανται τρόπον in E. The phenomenon can still be detected in the absence of the original recension of E. Here is a case in point.

Several differences between the two texts can, again, be attributed to the same tendency to smooth over and regularize the text on the side of the non-standard version, as preserved in P and M ^o . In the very first sentence, the γάρ-clause (οὐσία γὰρ ἡ κατὰ τὸν λόγον, 412b10) which supplies the content of the speech-act evoked in the previous clause (εἴρηται τί ἐστιν ἡ ψυχή) is turned into a dependent ὅτι-clause (ὅτι οὐσία ἡ κατὰ τὸν λόγον, to be construed with εἴρηται), thus dissolving however little ambiguity was attached to the particle γάρ. The main term is moved to the end in the non-standard version so as to produce a slightly easier syntax (ἔχοντος ἀρχὴν κινήσεως καὶ στάσεως → ἔχοντος κινήσεως καὶ στάσεως ἀρχὴν, b17; similarly, οὐ γὰρ τοιούτου σώματος ἡ ψυχὴ τὸ τί ἦν εἶναι καὶ ὁ λόγος → οὐ γὰρ τοιούτου σώματος τὸ τί ἦν εἶναι καὶ ὁ λόγος ἡ ψυχή, b16). The variation between the phrases ἀλλ’ ἢ ὁμωνύμως (412b14–15) and πλὴν ὁμωνύμως (412b21) is eliminated in the non-standard version which has instead ἀλλ’ ἢ ὁμωνύμως twice. A resumptive pronoun is added in b22 (δεῖ δὴ λαβεῖν → τοῦτο δεῖ λαβεῖν).

The most systematic difference has to do with the examples: the vulgate version has as its example of a tool an axe, πέλεκυς (four occurrences); the text preserved in P and M ^o has as its example a saw, πρίων (four times as well).Footnote ⁴¹ It seems unlikely that one of these two readings arose a corruption of the other, especially as the example is used repeatedly. But why would one have taken the place of the other? Both πέλεκυς and πρίων are standard (and even somewhat famous) Aristotelian examples of tools:Footnote ⁴² it does not make any difference for the argument whether Aristotle is speaking of the one or of the other. The only reason I can think of comes from the corresponding nomina actionis. What one does with an axe, πέλεκυς, is the act of cutting, τμῆσις; what one does with a saw, πρίων, is the act of sawing, πρίσις. In the latter case, the nomen actionis is obviously related to the noun for the tool, which is not the case in the former. Substituting the couple πρίων/πρίσις for the couple πέλεκυς/τμῆσις makes the connection between the two unmistakable on a morphological level. Again, the most likely explanation for the difference between the two versions assumes that the non-standard version has been produced on the basis of the vulgate.

One other reason to favour option II above according to which the non-standard version derives from the vulgate comes from comparing its peculiarities with what has been preserved of the ancient exegetical tradition. There is at least one clear case of interpolation in the non-standard version. It is found in 424a2–5, quoted extensively above, where the non-standard version differs from the vulgate by the additions of the words ἂν τοῦ αἰσθητηρίου. As Sophonias seems to have noticed, they wreak havoc on the syntax and meaning of the clause, so that they can hardly have been intended to stand there as such.

Where does the supplement come from? The ancient exegetical tradition sheds some light on the issue.Footnote ⁴³ Themistius, ‘Simplicius’ and Philoponus all rely on the vulgate text. That being said, both ‘Simplicius’ and Philoponus consider explaining the passage by claiming that τῆς αἰσθήσεως in 424a4 means τοῦ αἰσθητηρίου, that is, the sense-organ and not the sense-capacity. This exegetical move is only evoked by ‘Simplicius’;Footnote ⁴⁴ it is more decidedly endorsed by Philoponus.Footnote ⁴⁵ Themistius’ take on the passage is somewhat ambiguous about this issue, but his paraphrase still seems to go in the same general direction.Footnote ⁴⁶

This interpretation aims at solving a problem that immediately arises from the context. Aristotle has just claimed that the sense-organ for touch is potentially the differences of bodies it perceives (423b26–424a1). But if it is a body, it must have some of these tangible properties: it must for instance be located somewhere in the hot/cold spectrum. This seems to already provide an explanation of the empirical fact that we do not feel the temperature of objects that are as cold or hot as our body when we touch them. In virtue of the principle that like is not affected by like, the sense-organ for touch will thus not be affected by objects that are similarly hot or cold. Why then would Aristotle make a new claim about the capacity for touch of one’s soul, if this is what τῆς αἰσθήσεως means in 424a4? Ancient commentators tend to avoid the issue by claiming that the clause is still about the organ of the body, and not about the capacity of the soul.Footnote ⁴⁷

Taking one’s cue from the way in which this interpretation is presented in ‘Simplicius’ (ἢ ἀντὶ τοῦ αἰσθητηρίου εἰπὼν τῆς αἰσθήσεως, 165.11–12), the most likely explanation for the reading of the non-standard version is that it results from an unfortunate interpolation on the basis of a similar paratext. Since a scribe is unlikely to misinterpret her own script and intervention in a manuscript, I assume that the scribe responsible for the paratext is different from the scribe who added the corresponding words to the main text.

• Step 1: The words ἀντὶ τοῦ αἰσθητηρίου are added by scribe 1 to a copy of Aristotle’s De anima near the words τῆς αἰσθήσεως in the main text (424a4) as an exegetical annotation.

• Step 2: The word ἀντί in the annotation is misread as the particle ἄνFootnote ⁴⁸ (due probably to an abbreviation). The annotation is mistaken for the correction of an omission by scribe 2, who proceeds to insert the words ἂν τοῦ αἰσθητηρίου in the main text.

One could object that accidents of this type may happen at any point in the transmission as long as some paratext is present, without being connected to the production of a new version. There are, however, clues that support the hypothesis that the ancient exegetical tradition played a role in the process of which the non-standard version is the outcome.

There are traces of the influence of Philoponus’ commentary in the texts of P and M ^o . In 414a1, both manuscripts read τοῦτο δὲ ποιεῖ διαφορὰς τῶν ζώντων instead of τοῦτο δὲ ποιεῖ διαφορὰν τῶν ζῴων. The switch from ζῴων to ζώντων comes in all likelihood from Philoponus’ commentary, in which it is explained that τῶν ζῴων here means τῶν ζώντων (243.14–17). In 417a1–2, the text of P adds to Aristotle’s cross-reference εἰρήκαμεν ἐν τοῖς καθόλου λόγοις περὶ τοῦ ποιεῖν καὶ πάσχειν (which is to De generatione et corruptione 1) the words ῥηθήσεται δὲ καὶ νῦν, which according to Philoponus must be implied since Aristotle is going to address this very topic, even if they are not present in the text he reads (although Alexander, according to Philoponus, knew of a copy with the words λεκτέον δὲ καὶ νῦν).Footnote ⁴⁹

The most notable case is by a distance that of Themistius. There are often what may sound like faint echoes of his paraphrase in the non-standard version or at least in the original recension of E: 416a6 πρὸς δὲ τούτοις vulg. P M ^o : ἔτι δὲ E Themistius (51.10–11); 421b9–10 καὶ γὰρ τὰ ἔνυδρα δοκοῦσιν ὀσμῆς αἰσθάνεσθαι vulg.: καὶ γὰρ τὰ ἔνυδρα φαίνεται αἰσθανόμενα ὀσμῆς E: καὶ γὰρ τὰ ἔνυδρα φαίνεται αἰσθανόμενα καὶ ὀσμῆς P M ^o : καὶ γὰρ τὰ ἔνυδρα τῶν ζώων ὀσμῆς αἰσθάνεσθαι φαίνεται Themistius (69.6–7). The connection with Themistius’ paraphrase is especially striking in 420b2 (for which the original recension of E is no longer available). Aristotle is comparing what ‘sharp’ (ὀξύ) and ‘blunt’ (ἀμβλύ) mean in the haptic realm with what the ‘sharp’ or ‘high’ (ὀξύ) and ‘flat’ or ‘low’ (βαρύ) mean in the acoustic realm. He points out the reason why the haptic ‘sharp’ and the acoustic ‘sharp’ have the same name: both are associated by means of an analogy with the property for something perceptible of moving fast (in different environments, against one’s skin or towards one’s ear, resulting in both cases in acute sensations, 420a30).

In the vulgate text,Footnote ⁵² Aristotle starts by explaining what sharpness and bluntness mean for touch by matching each property with its corresponding action: what is sharp somehowFootnote ⁵³ ‘stings’ (κεντεῖ), what is blunt somehow ‘pushes’ (ὠθεῖ). These actions are then brought under a more abstract general description that reveals the analogy: stinging occurs because the object moves through one’s flesh in little time and so produces a vivid haptic sensation (διὰ τὸ κινεῖν τὸ μὲν ἐν ὀλίγῳ, sc. χρόνῳ; presumably because sharp objects are only in contact with a small amount of skin, which offers less resistance) and pushing occurs because the object requires, comparatively, a greater amount of time to trigger the same result (διὰ τὸ κινεῖν … τὸ δὲ ἐν πολλῷ, sc. χρόνῳ; presumably because blunt objects do not pierce through one’s flesh and are felt in a much duller way). The contrast between sharp and blunt turns out, at this point, to be captured by the same description that applies to the contrast between sharp and flat for sounds, which had just been shown to be a matter of comparative speed (420a30–31).

In both Themistius’ paraphrase and the non-standard version, the verbal step in which each property is associated with its action (κεντεῖ vs ὠθεῖ) occurs at the acoustic level, and not at the haptic one. What is now said to ‘sting’ and ‘push’ are initially sharp and flat sounds. The difference may not look like much, but it still brings a major (and, in my view, catastrophic) change to the way the analogy is built. ‘Stinging’ and ‘pushing’ are metaphors when applied to sounds in a way in which they are not when applied to tangible bodies.Footnote ⁵⁴ Moreover, the result is that the text says almost nothing about the haptic properties of sharpness and bluntness. Instead, it needlessly repeats the association of acoustic sharpness and flatness with a difference in speed that had already been stated (420a30–1), whereas the claim that sharpness and bluntness are also connected with a similar difference would call for some justification.

Themistius avoids the issue by adding a sentence before the verbal step that makes precisely this point (‘sharpness’ means for a figure ‘moving one’s flesh or body by a great amount in little time’, 66.4–6), so that the ground for the analogy is already secure by the time ‘stinging’ and ‘pushing’ are introduced as metaphors for acoustic processes. There is no such remedy in the non-standard version, which I therefore regard as corrupt. The mistake requires little explanation: in the course of an analogy between ὀξύ/ἀμβλύ and ὀξύ/βαρύ, given that the first term is linguistically identical in both cases, there is always a risk of misunderstanding the couple to which a sentence applies.

It is difficult to believe that the two innovations of the non-standard version, ἀμβλύ → βαρύ and οἷον → ὥσπερ, could not be connected with Themistius’ paraphrase of the passage, which features the same first change and then adds precisely a ὥσπερ-clause to the original sentence. Finding a historical path from the one to the other, however, requires using a philological shaker: it could be that some annotation inspired by Themistius was badly misinterpreted in an ancient copy of Aristotle, or that Themistius is relying on a similarly deviant version of Aristotle’s text of uncertain origin. The existence of independent evidence in 424a4 that the non-standard version (as it is transmitted) stems from a manuscript with annotations drawn from the ancient exegetical tradition which have been subsequently misunderstood increases the plausibility of the first option.

Returning to the issue of the substitution of πρίων/πρίσις for πέλεκυς/τμῆσις in the non-standard version in 412b9–413a3, it can hardly be a coincidence that Themistius briefly switches when paraphrasing this very passage to the example of the saw, πρίων, instead of the axe, πέλεκυς, before introducing the corresponding nomen actionis: ὥσπερ γὰρ ὁ σίδηρος ὁ ἀμόρφωτος δυνάμει πρίων, ὅτι δέξασθαι δύναται τοιόνδε σχῆμα, ἀλλ’ ὅμως οὐδέπω πρίων, … (43.10–11, see also 14 and 19). There is, however, little doubt that Themistius read the couple πέλεκυς/τμῆσις in his copy of Aristotle which remains his first example (42.19–25, 43.12). It seems that Themistius decided to come up with a pair of terms of his own making with a clearer morphological connection. Chances are, therefore, that the use of the couple πρίων/πρίσις in the non-standard version results from Themistius’ efforts, assuming that the words were present somewhere in the paratext in a copy of Aristotle and eventually found a way into the main text.

Let us take stock. There are many subtle clues that the ancient exegetical tradition, most of all the paraphrase of Themistius and the commentary attributed to Simplicius, exerted an influence on the text of the non-standard version. There is no conclusive evidence that Themistius, ‘Simplicius’ or Philoponus had any awareness of the non-standard version. This influence is pervasive and deeply entrenched in the text of the non-standard version that can be reconstructed on the basis of the extant witnesses. In some cases, there is evidence in favour of the hypothesis that the innovations of the non-standard version result from a two-step process in which paratext related to ancient exegetical works ended up having an impact on Aristotle’s text.

These observations make option II above, according to which the non-standard version derives from the vulgate, much more probable than option III, according to which the vulgate derives from the non-standard version. They are not on their own decisive when adjucating between option II and option I, according to which both versions stem from a common source independently from each other. Option I, however, entails that the non-standard version should regularly preserve the original text against the vulgate (and conversely) throughout Book 2. To judge from current editions of the text, this is not the case: editors have not felt any need to rely on the non-standard version to reconstruct Aristotle’s text.Footnote ⁵⁵

4. The genesis of the non-standard version. Some hypotheses about De anima BOOK 2 and De partibVs animaliVm 4.11–12

Claiming that the non-standard version derives from the vulgate generates the expectation that there should be a plausible way of understanding the genesis of the former version on the basis of the latter. Picturing a historical process of which the vulgate would be the input and the non-standard version the output is a difficult matter. The amount and nature of the innovations in the non-standard version go far beyond the deviations usually induced by the copying process: some deliberate intervention must have taken place. Yet these innovations mostly affect limited aspects of the text on a micro-scale (lexical choices, syntaxis of a given clause) while preserving its general structure and meaning. It is striking how little improvement they bring to Aristotle’s text, be it in terms of legibility or philosophical precision, especially when the non-standard version is contrasted with other texts that have been produced on the basis of Aristotle’s in the genre of the paraphrase or the commentary. As a result, it is difficult to grasp what the motivations behind this alleged deliberate intervention would be.

Comparisons with similar cases may shed some light on the issue. Torstrik had tried to buttress his claim of a double recension of De anima 2 by pointing out other similar cases throughout the corpus Aristotelicum, including (obviously) Physics 7.Footnote ⁵⁶ His most convincing parallel is De partibus animalium 4.11–12, for which a different text is known to have been preserved in manuscript Vaticanus graecus 261 Y (early fourteenth century, already used by Bekker).Footnote ⁵⁷ After Torstrik, the issues of a ‘double recension’ for De anima 2 and for De partibus animalium 4.11–12 have often been discussed together. It is worth keeping in mind that De anima and De partibus animalium are two distinct treatises that may not share the same transmission history. Two factors still tie the two issues together.

(1) The treatise De partibus animalium is also preserved in E, of which Y is probably an apographon. Footnote ⁵⁸ Now, the end of the treatise in E (from 4.5, 680b36 πλῆθος onward, f. 345^r) is from a different and much later hand: the end of the original recension of the treatise, with which the codex originally ended, has been lost and replaced at a later date.Footnote ⁵⁹ The original recension of E is thus an unknown quantity for this section of the treatise for which Y offers a divergent text. One cannot help wondering if the text found there in Y, the alleged ‘recensio altera’ for De partibus animalium, was copied directly from E before the loss occurred.Footnote ⁶⁰ Parisinus E could, once more, turn out to be the main source for a divergent version of Aristotle’s text.Footnote ⁶¹

(2) The way in which the text preserved in Y differs from the vulgate is, on the whole, not dissimilar when compared to the way in which the non-standard version of De anima 2 differs from the vulgate text. Again, the text of Y offers a wide array of small-scale innovations (syntax: διὰ … τὸ κεκλῆσθαι for ἐπεὶ … κέκληται in 691b28; small paraphrastic additions: τούτου δ’ αἴτιον ὅτι … for αἴτιον δ’ ὅτι in 692a1–2 and again in 22, ἡ … αἰτία δήλη for δῆλον in 693b24; lexical substitutions: πρόσθεν for πρότερον in 692a17, οἷον for ὥσπερ in 692a12, ὁμοίως for ὡσαύτως in 692b7, κινοῦντα for ὠθοῦντα in 693b17) which altogether leave the general structure and meaning remarkably unaffected.

A scholarly consensus has gradually emerged around the view that the text of Y for De partibus animalium 4.11–12 derives from the vulgate and should be excluded from the constitutio textus.Footnote ⁶² There is, however, little agreement as to how it was produced on the basis of the vulgate. Thurot pictures students copying some reference material while focussing mostly on its meaning with little regard for textual accuracy (and offers an interesting medieval parallel).Footnote ⁶³ Düring calls the style ‘as un-Aristotelian as possible’ and describes a naughty scribe having some improvisational fun with the text before losing interest.Footnote ⁶⁴ Louis envisages the text as ‘a kind of school assignment’.Footnote ⁶⁵ Torraca thinks of the author of the text as having set out to write a genuine paraphrase in the manner of Themistius or Sophonias, before giving up due to exhaustion or lack of time.Footnote ⁶⁶ Pappa gathers the ample evidence that is still extant for the philological skills of Pachymeres, the main scribe of Y, and claims with some caution that Pachymeres’ involvement in Y ‘is not independent from the production of [its] recension’.Footnote ⁶⁷

Given the similarities between the non-standard version of De anima and the version found in Y of De partibus animalium, the issue is worth a brief discussion. Both share two puzzling features: (a) the evidence for a divergent version is restricted to a section of the text; (b) its innovations usually leave the general structure and meaning untouched without bringing any major improvement to the text.

Feature (a) is especially challenging for De partibus animalium, for which the evidence of a divergent text is limited to a short section that deals with the topic of oviparous blooded animals. This feature is hardly explained by the hypotheses put forward by Torraca and Pappa (why would the author of a paraphrase or a scholar revising the text not start with its beginning or most interesting part instead of so random a piece?). Feature (b) is equally difficult to understand if one thinks of the version of Y as a paraphrase, pace Torraca, or as the product of editorial interventions by a highly competent scholar with special concerns for the state of the text, pace Pappa. The motivations ascribed to the author by Düring (the thrill of manipulating the text?) and Louis (practicing one’s understanding of Aristotle?) remain obscure and do not explain feature (a) either.

Thurot, all in all, might have come closest to the truth. Feature (a) has a ready-made explanation in the well-attested workshop practice of switching between scribes for practical reasons, sometimes without any textual basis. Feature (b) could be accounted for by supposing an unfaithful scribe, who would perhaps have had access to exegetical materials (for instance annotations from an oral lecture) but whose motivations and methods remain obscure. With some imagination, one could think of concrete situations that would result in lower accuracy standards in one’s output, such as working under harsh times constraints, taking notes from oral dictation or setting out to produce an exemplar for one’s strictly personal use.Footnote ⁶⁸ Perhaps faithfulness in the transmission of the corpus Aristotelicum was not always as valued as one would have expected, even if most witnesses of this kind have been subsequently eliminated from the tradition.Footnote ⁶⁹

There is, however, one important dissimilarity between the case of De anima 2 and the case of De partibus animalium 4.11–12: there is no extant ancient exegetical work for the latter (before the commentary of Michael of Ephesus), whereas there are connections between the non-standard version of De anima 2, on the one hand, and between Themistius’ paraphrase and the commentary attributed to ‘Simplicius’ on the other. The connection is multi-staged for at least one clear case of interpolation, so that one need not assume that one single scholar produced the non-standard version by obsessively altering the text on the basis of the ancient exegetical tradition. The explananda in the case of De anima 2 are therefore (a) the existence of a non-standard version for Book 2 only (unless the rest of it has been lost), (b) the amount and nature of its innovations with regard to the vulgate, (c) the connections between these innovations and the ancient exegetical tradition. Here is one hypothetical scenario that could potentially account for all of them with some plausibility. Step 1 would have taken place after the production of Philoponus’ commentary in the sixth century, step 2 would have taken place before the production of the earliest witness to the non-standard version in the tenth century.

1. (1) Production of a scholarly copy of Aristotle’s De anima (vulgate version) with a special layout incorporating materials from the ancient exegetical tradition, including both scholia from the ancient commentaries and excerpts from Themistius’ paraphrase for the passages where it directly matches Aristotle’s text. This wealth of paratext is not always clearly demarcated from the main text.

2. (2) Use of this scholarly copy in a workshop as antigraphon (model for the production of a new copy of the same text). The copying of Book 2 is assigned to an unfaithful scribe who ends up mixing some of the paratext with the main text in her output, resulting in the non-standard version.

This sketch is speculative. What matters is that, with some imagination, there are possible historical paths by which the vulgate text would gradually transform into the non-standard version.