Textual Comment on Aristotle, Poetics 1447b7–9, Part 2

2016.02.02 | By Coleman Connelly

From the Arabic translation of Aristotle’s Poetics preserved in Parisinus Arabus 2346. Image via Wikimedia Commons.

§1. In Part 1 of this comment, I discussed Poetics 1447b7 to 1447b9 in its intellectual context, pointing out that the text as printed in Kassel’s 1966 edition is at odds with what must be Aristotle’s meaning. I demonstrated how the text printed in Gutas and Tarán’s 2012 editio maior—the first to make expert and accurate use of the Syriac and Arabic traditions of the Poetics—in fact jibes much more closely with Aristotle’s meaning. In this second part, I will explain how Tarán, working in conjunction with Gutas, arrived at this reading, basing himself on earlier conjectures and the crucial evidence provided by the Syro-Arabic translation of Aristotle’s work. Many students of the Poetics still make casual use of Kassel’s edition, especially since it is the text digitized by the TLG, and many translations of the Poetics currently in use simply follow Kassel’s text, without incorporating the evidence provided by the Syriac and Arabic traditions.[1] I hope that, by walking the reader through the reasoning and evidence behind the text of Poetics 1447b7–9 printed in the new 2012 edition, I can urge more people to consult this important work, which is readily accessible at least to those whose institutions subscribe to Brill Online. More broadly, I hope to demonstrate to my fellow Hellenists how important the Graeco-Syriac and Graeco-Arabic traditions can be to our work.

§2. Before I explain the text of Poetics 1447b7–9, however, it will be helpful to review, very briefly and in broad strokes, the transmission history of Aristotle’s work. We possess four primary witnesses. Two are Greek manuscripts, namely a mid- to late-tenth century Greek manuscript (A) and a mid-twelfth century manuscript (B).[2] Since B is missing the opening of the work and begins only at 1448a29, it will not occupy us in our discussion of 1447b7–9 below. In addition to A and B, we also possess the literal Latin translation made in 1278 by William of Moerbeke while he served as the Latin Archbishop of Corinth and surviving in two manuscripts written shortly after the translation’s composition.[3] From William’s Latin, we are able to reconstruct the Greek manuscript on which he relied (Φ). It is clear from common errors that A and Φ go back to a shared lost ancestor, Greek manuscript Π. Together, Π and B can be traced to a hyparchtype Ξ, the common source then for all of what we may term the ‘Graeco-Latin tradition’ of the Poetics.[4]

§3. There is more, however. Thanks to the trans-generational efforts of Syriac and Arabic translators in ninth- and tenth-century Baghdad, we possess a fourth primary witness to the Poetics which greatly supplements our knowledge of the text. This is the literal, at times nearly word-for-word Arabic translation of the East Syrian Christian philosopher ’Abū Bišr Mattā ibn Yūnus, made in the first half of the tenth-century and surviving, in a revised and somewhat lacunose form, in a single eleventh-century manuscript (Parisinus Arabus 2346). ’Abū Bišr, however, did not translate from the Greek, of which we know he was ignorant, but from a Syriac translation of the Poetics which was made in the ninth-century and itself clearly revised at least once. Except for a single fragment preserved in a later Syriac author’s quotation, we have no direct access to this Syriac translation. Nevertheless, so scrupulous and literal is ’Abū Bišr’s Arabic that, with some ingenuity, we can deduce the readings found in the Greek manuscript from which the Syriac translator worked (Σ). We can also deduce something about the nature of Σ. It was clearly a manuscript of great antiquity, for it must have been written in uncials and scriptio continua. This becomes clear from some of the errors made by the Syriac translator and reproduced in the Arabic by ’Abū Bišr, errors that can only be due to Syriac translator’s misconstruing of a majuscule text written without word divisions. By far the most charming of these errors occurs at Poetics 1451b21, where the Syriac translator was confronted with ΟΙΟΝΕΝΤΩΑΓΑΘΩΝΟΣΑΝΘΕΙ, which is correctly read as οἷον ἐν τῷ ᾽Αγάθωνος ᾽Ανθεῖ (‘Just as in the Antheus of Agathon’). Largely ignorant of the cultural context and completely unaware of Agathon and his lost tragedy, the translator was nevertheless well-versed in Greek philosophy. From ’Abū Bišr’s Arabic, it is clear that the Syriac translator parsed the Greek before him as οἷον ἓν τὸ ἀγαθὸν ὃς ἂν θῇ (‘Just as whoever posits the good to be unitary’). The Greek manuscript Σ from which the ninth-century Syriac translator worked was therefore not written in Byzantine minuscule but in uncials and hence considerably older than our earliest Greek witness, A.[5] Much more might be said about the indirect evidence from quotations in later Syriac and Arabic authors that further inform us about the various versions of ’Abū Bišr’s translation or of its Syriac intermediary, that in turn help us reconstruct Σ, and that may even attest to a separate Greek manuscript (Ψ) used in one of the revisions of ’Abū Bišr’s translation.[6] Let us stop here, however, and turn to the text of Poetics 1447b7–9.

§4. Based solely on the evidence afforded by A and the Latin, i.e., by the text that Π must have had, Poetics 1447b7–9 would read something like this:

ἡ δὲ ἐποποιία μόνον τοῖς λόγοις ψιλοῖς ἢ τοῖς μέτροις καὶ τούτοις εἴτε μιγνῦσα μετ’ ἀλλήλων εἴθ’ ἑνί τινι γένει χρωμένη τῶν μέτρων τυγχάνουσα μέχρι τοῦ νῦν . . .

This is clearly nonsensical and over the years many attempts have been made to improve the text.[7] The best we shall probably achieve is the text printed by Tarán in collaboration with Gutas, where the Syro-Arabic tradition provides the only hard evidence for our reconstruction:

ἡ δὲ [ἐποποιία] μόνον τοῖς λόγοις ψιλοῖς ἢ τοῖς μέτροις καὶ τούτοις εἴτε μιγνῦσα μετ’ ἀλλήλων εἴθ’ ἑνί τινι γένει χρωμένη τῶν μέτρων ἀνώνυμος τυγχάνει οὖσα μέχρι τοῦ νῦν . . .

How did Gutas and Tarán arrive at this text?[8] Let us start with ἐποποιία. Even before Margoliouth first made the Arabic translation available to classical scholars working on Aristotle’s text in 1887, Überweg had proposed this word’s deletion (i.e. bracketing). Clearly, Aristotle cannot be saying that epic poetry is written in prose, to name only one objection one might raise to retaining it. It is probably an intrusive and erroneous gloss on ἡ δέ from an ancient or medieval reader. The Arabic confirms that this is an intrusion into the text and a fairly late one at that, for ’Abū Bišr does not translate it and it was evidently not in Σ. We may mention here too an example of a reading preserved by the Arabic but which editors have decided not to adopt. The Arabic does not translate μόνον, and instead has ’akṯara (‘more’). While Tkatsch argued that this represented a mistranslation, it is far more likely that Σ had μᾶλλον for μόνον and that the Arabic is simply reflecting this reading, for the Arabic translation routinely has ’akṯara for μᾶλλον.[9] Nevertheless, the μόνον preserved in the Graeco-Latin tradition is the reading which the editors have adopted.

§5. Let us turn now to the ἀνώνυμος (‘nameless’) printed Gutas and Tarán’s collaboration. As we saw, Π completely omits this word, resulting in nonsense. It is in fact only preserved in the Arabic (bi-lā tasmiyatin, ‘without an appellation’ ), which clearly demonstrates its presence in Σ. Remarkably, Bernays had already conjectured—on the basis of what follows and without access to the Arabic—that ἀνώνυμος was the missing word, but it is only with the Arabic’s witness to Σ that we are able to confirm this reading beyond a doubt. It should be noted that what may be a reference to this passage in John Tzetzes provides circumstantial evidence for the reading ἀνώνυμος as well.[10] What of Gutas’s and Tarán’s τυγχάνει οὖσα for Π’s τυγχάνουσα? This is in fact the nineteenth-century conjecture of Suckow, adopted also by, e.g., Janko in his translation of the Poetics. What light can the Syro-Arabic tradition shed on this point? To answer this question and to give a cautionary example of how a lack of expertise in Syriac and Arabic can mislead an editor of a Greek text who consults the Syriac and Arabic evidence, it will be helpful to look back at the text printed by Kassel, which we saw in Part I:

ἡ δὲ [ἐποποιία] μόνον τοῖς λόγοις ψιλοῖς <καὶ> ἡ τοῖς μέτροις καὶ τούτοις εἴτε μιγνῦσα μετ’ ἀλλήλων εἴθ’ ἑνί τινι γένει χρωμένη τῶν μέτρων ἀνώνυμοι τυγχάνουσι μέχρι τοῦ νῦν . . .

First, we should observe that <καὶ> ἡ is the conjecture of Lobel, adopted by Kassel. In fact, Π has ἢ and the Syriac translator clearly also found Η not ΚΑΙH in his uncial manuscript Σ and read it as ἢ, for the Arabic has ’aw (‘or’). Lobel assumed that two mimetic arts, not one, were being spoken of here, and it was for this reason that he proposed <καὶ> ἡ and made the necessary change of ἀνώνυμος τυγχάνει οὖσα to ἀνώνυμοι τυγχάνουσι οὖσαι, which Kassel simplified to ἀνώνυμοι τυγχάνουσι.

§6. Now, Lobel and Kassel may have been led to this conjecture of a plural by the neo-Latin translation made by Tkatsch in the early twentieth century of the Arabic Poetics. They certainly both cite Tkatsch and Kassel notes in his apparatus Tkatsch’s rendering of the Arabic as quae est (sive sunt) sine appellatione, which indicates correctly that the number of the Arabic—singular or plural—is ambiguous in this instance. Can the Arabic in fact help us ascertain whether Σ had the singular ἀνώνυμος, as Bernays, or the plural ἀνώνυμοι, as Lobel and Kassel? In fact it can, but only when we consider the lost Syriac intermediary upon which is was based, as was only observed and demonstrated in 2012 by Dimitri Gutas.[11] The question hinges on the relative clause with which the Arabic translator chose to render whatever stood in the Syriac for ἀνώνυμος/ἀνώνυμοι: allatī hīya bi-lā tasmiyatin (‘which is/are without an appellation’). Due to a quirk of Arabic grammar, the feminine singular relative pronoun allatī (‘which’) and the feminine singular personal hīya (‘she’, here used in place of the linking verb ‘is/are’) can refer either to a singular feminine antecedent or to a plural antecedent if that antecedent is inanimate. That is, inanimate plurals regularly take feminine singular pronouns, verbs, and adjectives in Arabic. Does this relative clause modify the implied noun τέχνη (‘craft) in ἡ δέ, rendered quite accurately in Arabic as ba‘ḍuhā (‘one of them’ [i.e., one of the crafts]), and thus reflect a reading of ἀνώνυμος? Or does it in fact modify the Arabic equivalent of τῶν μέτρων (‘the meters’), that is al-’awzān, which the relative clause immediately follows? Since al-’awzān is an inanimate plural it is quite possible grammatically for the relative clause to refer to it. If it does, then we might be led to believe that the Syriac translator read something like τῶν μέτρων, ἃ ἀνώνυμα τυγχάνει ὄντα μέχρι τοῦ νῦν (‘the meters, which have hitherto been nameless’) in Σ. Or might this plural somehow reflect, obscurely, a reading of Lobel and Kassel’s plural ἀνώνυμοι in Σ?

§7. In fact, the Arabic confirms beyond a doubt that Σ had ἀνώνυμος, but we can only ascertain this by considering what the Syriac intermediary translation must have said and how the Arabic translator reacted to it, as Gutas was the first to observe. In more natural Arabic, as distinct from the ‘translationese’ of ’Abū Bišr, we would in fact expect al-’awzān allatī hīya bi-lā tasmiyatin ’ilā al-’ān to mean ‘the meters, which have been [lit., ‘are’] nameless until now’ and thus reflect a reading of τῶν μέτρων, ἃ ἀνώνυμα τυγχάνει ὄντα μέχρι τοῦ νῦν in Σ. This reading would, it goes without saying, be in gross violation of Aristotle’s sense here. There is in fact no need to make this extravagant assumption. To do so is to ignore the fact that ’Abū Bišr was translating from a Syriac intermediary, not directly from the Greek. The Syriac translator clearly rendered ἀνώνυμος with a relative clause meaning ‘which is nameless’. Syriac’s relative pronoun (d-) is without gender, but this relative clause must have contained some feminine singular construction, probably a participle, which ensured that the feminine singular τέχνη implied in ἡ δέ, and not the plural for ‘meters’, was the unambiguous antecedent in the Syriac. For in Syriac, unlike in Arabic, a feminine singular pronoun, verb, or adjective can only refer to a feminine singular noun, not to an inanimate plural noun. As we just saw, the Arabic correctly renders ἡ δέ with a feminine singular pronoun ba‘ḍuhā (‘one of them’), which jibes with the Arabic for ‘τέχνη’—the feminine singular noun ṣinā‘aand which undoubtedly reflects a feminine singular pronoun in the lost Syriac. (The usual Syriac for ‘τέχνη’, ’ūmānūṯā, is also feminine singular). ’Abū Bišr literally rendered the Syriac’s feminine singular relative clause ‘which is without a name’ and, with equal scruples, placed it at the end of the sentence where he found it in the Syriac. In this way, the relative clause ended up right after the word for ‘meters’ in the Arabic, thereby unintentionally introducing an ambiguity for the Arabic reader without access to the Syriac. In light of Gutas’s reconstruction of the Syriac intermediary here, we can safely deduce that Σ had ἀνώνυμος. There is no manuscript evidence whatsoever for Lobel and Kassel’s <καὶ> ἡ . . . ἀνώνυμοι and this conjecture, together with the resulting notion that Aristotle is talking about two unnamed genres here, can finally be laid to rest. Explicit feliciter, deo gratias, amen.

§8. But wait! We still haven’t resolved whether Σ had the τυγχάνουσα of Π or whether it might have borne witness to Suckow’s conjectured τυγχάνει οὖσα. In short, we cannot be certain what Σ read here. As Gutas has argued, however, Σ probably had τυγχάνουσα. The Arabic translation is, whenever possible, word-for-word and here ’Abū Bišr has only one Arabic word, the copula hīya, whereas above at 1447a25, where we find the phrase τυγχάνωσιν οὖσαι, he uses two Arabic words. Nevertheless, Suckow’s conjectured τυγχάνει οὖσα is attractive, and Tarán adopts it in his edition.[12]

§9. I hope this excursus on the text of Poetics 1447b7–9 has demonstrated the rewards of using evidence from Arabic translations in the editing of Greek texts. At the same time, I hope it has served as an object lesson for the dangers of ignoring potential Syriac intermediaries when consulting these translations. In this comment, I have concentrated on what light the Syriac and Arabic traditions can shed on our interpretation of a Greek author and his thought. Yet why, in the first place, was the Poetics translated into Syriac in ninth-century Iraq, in a society with no tradition of live theater? And why, in turn, did an East Syrian Christian logician render this Syriac translation into Arabic for his tenth-century contemporaries? What did the Syriac and the Arabic translator make of such alien terms as ‘comedy’ and ‘tragedy’, or ‘epic’ and ‘spectacle’? It is to these questions that I will turn in my upcoming contributions.



Janko, R. 1987. Aristotle: Poetics I, with the Tractatus Coislinianus. A Hypothetical Reconstruction of Poetics II. The Fragments of the On poets. Indianapolis.

Kassel, R. 1965 [1966]. Aristotelis, De arte poetica liber. Oxford.

Lobel, E. 1929. “A Crux in the Poetics,” CQ 23 (76–79).

Margoliouth, D. S. 1887. Analecta Orientalia ad Poeticam Aristoteleam. London.

Margoliouth, D. S. 1911. The Poetics of Aristote: Translated from Greek into English and from Arabic into Latin, with a Revised Text, Introduction, Commentary, Glossary and Onomasticon. London.

Suckow, G.F.W. 1855. Die wissenschaftliche und künstlerische Form der Platonischen Schriften. Berlin.

Tarán, L. and D. Gutas. 2012. Aristotle Poetics: Editio Maior of the Greek Text with Historical Introductions and Philological Commentaries. Leiden/Boston.

Tkatsch, J. 1928. Die arabische Übersetzung der Poetik des Aristoteles und die Grundlage der Kritik des griechischen Textes. Wien.

Überweg, F. 1870. Aristotelis Ars Poetica. Ad fidem potissimum codicis antiquissimi Ac (Parisiensis 1741) edidit. Berlin.



[1] One notable exception is the translation of Janko 1987.

[2] These are, respectively, Parisinus Graecus 1741 and Riccardianus 46.

[3] These are Etonensis 129 and Toletanus, bibl. Capit. 47.10.

[4] I draw here from the extensive study of Gutas and Tarán 2012, especially Chapter 3, and the reader should consult this work for more information.

[5] See Gutas and Tarán 2012:146–147. We should of course keep in mind that both Π and hence the hyparchetype of the Graeco-Latin tradition (Ξ) can be securely said to have been in majuscule as well: see Gutas and Tarán 2012:143–144.

[6] I draw here from the extensive study of Gutas and Tarán 2012, especially Chapters 2 and 3, and the reader should consult this work for more information.

[7] It should be noted that for most of this time, editors did not even have the benefit of the Latin, whose status as a primary witness was not recognized until the second half of the twentieth century. Of course, in this particular instance, the Latin is of no help.

[8] The following draws on and explicates Gutas and Tarán 2012:226–231 (Tarán’s commentary) and 312–314 (Gutas’s commentary).

[9] See the examples cited in Gutas and Tarán 2012:314 (Gutas’s commentary).

[10] Tzetzes, Versus de poematum generibus, line 11. Numerous parallels for unnamed classes of things deploying the word ἀνώνυμος can be found in Plato and Aristotle: see with references Gutas and Tarán 2012:228n8 (Tarán’s commentary).

[11] Gutas and Tarán 2012:312–314 (Gutas’s commentary), which I here explicate.

[12] See Gutas and Tarán 2012:230–231 (Tarán’s commentary).