We start from a passage in Plato's Cratylus, 393b7–c7. For reasons which will become clear in due course, I give Burnet'sFootnote ¹ text (with the addition of underlinings and apparatus of my own), which differs from that of the more recent OCT editionFootnote ² on an important detail.

This passage is very interesting in various respects. For one thing, it contains an ancestor of the famous Aristotelian slogan, stated and discussed for example in Metaphysics Z 7–9, according to which ‘a human begets a human’ (ἄνθρωπος ἄνθρωπον γεννᾷ). It is a problematic ancestor; for Socrates' argument goes on (393c–d, 394a) to treat such cases as a horse begetting a horse as being on a par with such cases as a king begetting a king or a good man begetting a good man, which are in fact quite different. Indeed, Socrates does something even worse: he appears to assume (394a–e) that another analogous case is someone called ‘King’ begetting someone called ‘King’ – or someone called Ἕκτωρ, ‘Holder’, begetting someone called Ἀστυάναξ, ‘Town-lord’.

The relations between what Socrates says and the Aristotelian principle, on the one hand, and between what Socrates says and what he really means, on the other, are a most interesting subject matter; but I will not pursue it on the present occasion.Footnote ⁴ Now I rather wish to turn to a completely different aspect of the passage, namely to its text and to what it can teach us about the early editorial history of Plato's dialogues. And in order to do so it will be useful to recall a few trivial facts about the textual transmission of the Cratylus.Footnote ⁵

The MSS of the dialogue can be grouped into three main families. The first is the β family, namely the family deriving from a lost common ancestor β. Members of this family are B = Bodl. MS E.D. Clarke 39 (written for bishop Arethas by John Calligraphus in a.d. 895)Footnote ⁶ and D = Ven. gr. 185 (twelfth century); here we shall concern ourselves also with Y = Vind. Phil. gr. 21 (fourteenth century), a learned edition, put together by several scribes under the direction of the Constantinopolitan scholars Nicephorus Moschopoulos and Maximus Planudes (who is actually the very scribe of our lines) and alternating different sources for different textual sections. In Cra. Y derives from B and its corrector B² up to 424a5, hence also in our passage; then it switches to another source which rather belongs to the second family.Footnote ⁷ This second family is the most numerous and derives from an extant MS, T = Ven. app. cl. 4.1 (tenth century). Third comes the δ family, namely the family deriving from a lost common ancestor δ. The most important member of this family is W = Vind. suppl. gr. 7 (eleventh century); but the above apparatus also mentions Q = Par. gr. 1813 (thirteenth century).

The exact relation between these three families is hard to determine and seems to change from one dialogue to another: sometimes they seem to form a tripartite stemma, sometimes a bipartite one. But this issue has no bearing on our present concerns, for which it is instead important to stress that βTδ depend, whether directly or indirectly, on a common source or archetype written in majuscule. The existence of this common source is proved, in Cra. as in other dialogues, by a number of textual errors which are common to all MSS and cannot be the result of polygenesis or contamination.Footnote ⁸

After this short introduction let us focus on line 393c1, where ἀλλ᾿ οὗ ἂν (underlined in the above quotation) is the reading of βT, substantially confirmed by δ's ἀλλ᾿ οὗ ἐὰν. This is alternative to the variant ἀλλ᾿ ὃ ἂν, which is ascribed by Duke et al. to the MS Venetus gr. 590 (fourteenth century) but in fact is already to be found, as MéridierFootnote ⁹ claims in his apparatus, in its coeval antigraph Y (fol. 39^v). More precisely, Y's scribe first wrote ἀλλ᾿ οὗ ἂν and then corrected οὗ into ὃ. I have expressed this in my apparatus by ascribing the latter reading to ‘Y^pc’; the fact that the former reading can be found in Y^ac did not have to be explicitly stated there, and should be regarded as implied by the reference to βT, because in this section of the text Y belongs to the β family, as we saw above.

Now, several editors (Ast, Bekker, Stallbaum and Méridier)Footnote ¹⁰ adopt the reading of Y^pc, and many interpreters translate as if they were adopting it: Fowler ‘I am not speaking of prodigies … but of the natural offspring of each species after its kind’, Minio-Paluello ‘Né intendo parlare di nascite mostruose … bensì di ciò solo che nasca dal suo proprio genere secondo natura’, Reeve ‘I'm not talking about some monster … but one that is a natural offspring of its kind’, Dalimier ‘je parle du rejeton naturel de l'espèce’.Footnote ¹¹

The former reading, by contrast, is printed by Hirschig,Footnote ¹² Burnet and Duke et al. What does it mean? The words ἀλλ᾿ οὗ ἂν ᾖ τοῦ γένους are equivalent to ἀλλὰ τὸ γένος οὗ ἂν ᾖ, ‘but the kind of which it is (by its nature an offspring)’: τὸ γένος, the relative pronoun's antecedent, has been attracted into the pronoun's genitive case and displaced inside the relative clause. This was well understood by Ficino,Footnote ¹³ who mirrored it in his Latin: ‘cuius generis secundum naturam est quod nascitur, hoc dico’. Ficino, like the other editors who side with him, doubtless knew that the attraction and displacement of the relative's antecedent which this reading presupposes is a fairly common phenomenon. We can see it at work in Resp. 402c: οὔτε αὐτοὶ οὔτε οὕς φαμεν ἡμῖν παιδευτέον εἶναι τοὺς φύλακας (= οἱ φύλακες οὕς κτλ.: literally ‘neither we ourselves nor the guardians whom we say we have to educate’) or in Sophocles, OC 907–8: οὕσπερ αὐτὸς τοὺς νόμους εἰσῆλθ᾿ ἔχων, | τούτοισι … ἁρμοσθήσεται (= τοῖς νόμοις οὕσπερ κτλ.: literally ‘with the laws which he brought with him as he came, with these … he will be ruled’).Footnote ¹⁴

So the reading with οὗ is grammatically possible and is certainly difficilior than that with ὃ. Indeed, it is also superior as far as meaning is concerned. If we read ὃ, Socrates' argument in the passage would essentially start like this:

That is to say, Socrates would apparently be saying that given a certain kind, the principle he is stating (‘The offspring of an X is to be called an X’) is not concerned with the possibility of a monstrous offspring. He would be considering a possible exception to the principle (e.g. a horse begetting a calf) and putting it aside as irrelevant. But this is not how Socrates really goes on. He rather goes on to explain that even such apparent counterexamples to the principle are merely apparent, once the principle is understood correctly. Let me quote again lines 393c1–5:

The birth of a calf from a horse is no real counterexample, because the calf must be regarded as the (ideal) offspring of the kind it naturally belongs to, namely the ox. Therefore it turns out that in the principle he is discussing, ‘The offspring of an X is to be called an X’, Socrates' clarification should not concern the second ‘an X’ but the first – not the part about the offspring but the part about the parent. He is explaining how we should identify the right parent, as it were, in order to make the principle hold true in all cases. And we obtain this result if at c1 we read οὗ instead of ὃ. This is confirmed by a passage at the end of the same argument (394d7–9), where Socrates recalls the present passage with the words

So far so good. Now that we have become clear about line c1, let us move on to c4. Here after γένηται the MSS add some words (cf. the second underlined phrase in my initial quotation): ἀλλ᾿ ὃ ἐὰν βW, ἄλλο ἐὰν T, ἀλλ᾿ ἐὰν Q. Some editors try to make something of these words. Thus Stallbaum (followed on this score by Hirschig) prints the following text:

which he translates thus: ‘neque si ex homine, opinor, fetus non humanus gignatur, sed si fetus naturalis suscipiatur, homo nominandus erit’. That is to say, ‘nor, I think, if something which is not the [natural] offspring of a human is born, but rather if the [natural] offspring [of a human is born], is he to be called a human’. Another attempt is Fowler's:

which he translates as ‘nor if any offspring that is not human should be born from a human being, should that other offspring be called a human being’.

Fowler's text is hardly grammatical and hence out of the question; Stallbaum's, though grammatically correct, is quite clumsy. Both attempts to preserve the MSS text are trumped by Peipers' simple and elegant deletion, which gives line c4 the shape in which it can be read in the editions of Méridier, Burnet and Duke et al., and in the quotation which opened this article.

Indeed, we can strengthen the case for the deletion, and at the same time make progress towards a deeper understanding of our passage's transmission, if we reason as follows.

1. (1) At c4 we do not really have various readings before us. The difference between ἀλλ᾿ ὃ ἐὰν (βW) and ἄλλο ἐὰν (T) is a matter of word-division, accents and breathings – all matters over which our MSS have no real authority. Both families actually offer us one and the same reading, the majuscule ΑΛΛΟΕΑΝ, which different scribes interpreted differently and presented in different shapes. Q's scribe merely went one step further by adding elision.Footnote ¹⁵

2. (2) This ΑΛΛΟΕΑΝ bears a striking resemblance to the disputed words at c1, in particular to the Y^pc variant ἀλλ᾿ ὃ ἂν, which in majuscule script was written ΑΛΛΟΑΝ. What is the meaning of this resemblance? I venture a hypothesis: ΑΛΛΟΕΑΝ (c4) is the result of a well-known kind of scribal mistake, i.e. the insertion of a marginal variant at the wrong place. In fact, this seems to have been already Burnet's hypothesis. For Burnet prints the deleted words not as they are attested in any MS but in an emended and interpreted form, ἀλλ᾿ ὃ ἂν, which emphasizes (rightly, in my view) the connection with c1. Indeed, in his apparatus ad c4 he explicitly adds a brief reference to the earlier line: ‘(cf. c1)’.

This reconstruction is plausible enough to constitute a further reason for endorsing Peipers' deletion at c4. Moreover, as I announced above, it has interesting implications regarding the passage's transmission. To see what those implications are, let us continue our reasoning.

1. (3) Recall that the three families of our MSS depend on a common source or archetype. Let us dub it ‘ω’.

2. (4) Generally speaking, when several MSS independent of each other share the same textual error, this must have already been present in their nearest common source, from which it was copied into them.Footnote ¹⁶

3. (5) Therefore the intrusive ΑΛΛΟΕΑΝ was already present at c4 in the text of ω.

The story might have gone like this. At some stage of the transmission of the Cratylus, a common ancestor of βTδprior to ω – let us dub it ‘ω₁’ – had ΑΛΛΟϒΑΝ (= ἀλλ᾿ οὗ ἂν) in the text at c1 and a variant ΑΛΛΟΑΝ (= ἀλλ᾿ ὃ ἂν) written in the margin. At some further stage of the transmission of Plato's text, which may be either intermediate between ω₁ and ω or (a more economical hypothesis) identical with ω, the marginal variant was corrupted into ΑΛΛΟΕΑΝ and mistakenly incorporated into the text at the wrong place, namely in line c4. In any case, the variant was in ω's text at c4, and thence it was copied into all three families of MSS. Only Y^pc, and of course its apograph Ven. 590, offer it (also) in its intended position, namely in the main text at c1. This ought not to disconcert us: Y is the kind of MS which is not unlikely to preserve ancient readings.Footnote ¹⁷

Here is a sketchy stemma, where the merely possible intermediate stage between ω₁ and ω is referred to as ‘(ω₂)’:Footnote ¹⁸

If at c1 δ reads ἐάν, this may be accidental polygenesis, corruption of ἄν into ἐάν being fairly common in our MSS (cf. apparatus of Duke et al. ad 385a2); or it may be the effect of contamination.

Now, ω has been tentatively identified by Antonio Carlini as the official text of Plato preserved in the Imperial Library at Constantinople (which was founded by the emperor Constantius II in a.d. 356).Footnote ¹⁹ Thus it is, if Carlini is right, quite an ancient source. How much more ancient does ω₁ have to be? Could it be one of the ancient editions of Plato's dialogues which have been hypothesized by modern scholarship, like the Alexandrian one by Aristophanes of Byzantium or even (but less probably) the Academic one? This is not impossible: scholars are in any case ready to ascribe to those editions some of the errors which are common to all our MSS, including the tetralogical ordering of the dialogues, which includes spurious materials and hence cannot stem from Plato himself.Footnote ²⁰ Indeed, an especially interesting parallel case is worth consideration. In a ground-breaking study of the ninth century MS O = Vaticanus gr. 1 (which contains Laws, Epinomis, Letters, Definitions and Spuria), Maria Jagoda Luzzatto has recently argued that the marginal notes written on it by a later hand, O⁴ (eleventh century),Footnote ²¹ derive from a lost sixth century MS which in turn collected scholia and variant readings stemming from yet more ancient sources.Footnote ²²

We have been proceeding backwards from the readings of the medieval MSS to the ancient editions of Plato. Now, continuing along the same path, I wish to make a final point about the ultimate origin of our textual variant; and in order to do so I need to recall a few basic facts about ancient Attic orthography.

To judge from the inscriptions, during most of the fifth century b.c. it was standard in Attica to use one single sign, Ο, for o (and another single sign, Ε, for e), whether short or long, whether open or close. Thus, for example, the words which we nowadays print as ἀνθρώπου and ἐπιστήμη would have been written respectively as ΑΝΘΡΟΠΟ and ΕΠΙΣΤΕΜΕ. But then this state of affairs underwent a twofold change. (i) In 403 b.c. Athens officially adopted the Ionic alphabet, which involved a distinction between Ε/Ο as the signs for close e/o, whether long or short, and Η/Ω as the signs for long open e/o. (ii) Between the end of the fifth century b.c. and the midst of the fourth a separate and gradual process brought about a distinction between Ε/Ο as the signs for short close e/o and ΕΙ/Οϒ as the signs for long close e/o.Footnote ²³

Now, there is every reason to believe that Plato adhered to the first change but not to the second. That he adhered to the first change (which took place in 403 b.c., hence probably before he started his career as a writer) is proved by the Cratylus' various references to Η (393e, 407b, 414c, 427c) and Ω (393d) as standard letters of the alphabet, as well as by some passages of the dialogue in which Socrates points out that the ‘ancient’ alphabet used Ε in place of Η (οὐ γὰρ ἦτα ἐχρώμεθα ἀλλὰ εἶ τὸ παλαιόν, 426c; cf. 411e, 418c) or claims that an etymology becomes clear if you consider the word as written ‘in the Attic language’ (ἐν τῇ Ἀττικῇ φωνῇ, 398d: presumably as contrasted with the Ionic one) or ‘in the ancient Attic way’ (Ἀττικιστὶ ὡς τὸ παλαιόν, 410c) – thereby meaning the pre-403 Attic alphabet, which featured Ε/Ο instead of Η/Ω as signs for long open e/o.

By contrast, there is no such positive evidence that Plato adhered to the second change, and indeed there is much positive evidence to the contrary. For example, in the Cratylus the whole argument at 416b–d, where Socrates sets forth the etymology of καλόν (‘beautiful’), is intelligible only if we assume that Plato is trading on the ambiguity of ΚΑΛΟΝ as between καλόν and καλοῦν (neuter participle of καλέω, ‘call’); indeed, at 416c7 and d4 βTδ actually read as καλόν (which might represent an original ΚΑΛΟΝ) what Duke et al. print as καλοῦν, accepting corrections by Badham and Burnet. Another nice example occurs at Resp. 509d, where Socrates explicitly hints at a wordplay between ὁρατοῦ (genitive of ὁρατόν, ‘visible’) and οὐρανοῦ (genitive of οὐρανός, ‘heaven’) which becomes much clearer if we write both words respectively as ΟΡΑΤΟ and ΟΡΑΝΟ, as Slings boldly but correctly does in his edition.Footnote ²⁴ And a third-century b.c. papyrus of the Laches (PPetrie II 50 = Plato 23 CPF) writes twice ΤΟΣ (iv.17, vi.12) for τούς.Footnote ²⁵

These things being so, at Cra. 393c1 ἀλλ᾿ οὗ ἂν is undoubtedly the correct reading to print in a modern edition; but the letters which Plato actually wrote, or had written by someone, were probably ΑΛΛΟΑΝ, because he wrote both ὅ and οὗ as Ο.Footnote ²⁶ Therefore it is at least possible that both variants at Cra. 393c1 arose as attempts to interpret his original words. To see a real instance of a comparable scholarly attempt we only have to look up the Scholia vetera in Pindarum, III p. 16.5–6 Drachmann, where we find Aristarchus arguing that what in the text of Nem. 1.24 available to him is still written, ‘in the ancient script’ (τῇ ἀρχαίᾳ σημασίᾳ), as ΕΣΛΟΣ actually has to be interpreted as ἐσλούς for metrical reasons.Footnote ²⁷