Jonathan Barnes examines the Ai Khanoum papyrus

The ruins of the Greek city of Ai Khanum lie at the confluence of the Oxus and the Kokcha, on the frontier of the province of Bactria, some 125 miles north of Kabul — and 2,500 miles east of Athens as the owl flies. Ai Khanum was doubtless one of the eight cities in Bactria and Sogdiana which, so Strabo reports (XI xi 4), were founded under the aegis of Alexander the Great — presumably Ai Khanum was Alexandria-on-Oxus, a city mentioned by Ptolemy in his Geography (VI xii 6). Its history was short; for a century and a half after its foundation it was pillaged, sacked, and destroyed. It had been a garrison town, but also a place of some culture and wealth. The French archaeologists who investigated the site between 1963 and 1978 uncovered a spacious city: broad streets, elegant houses (several of them with bathrooms), gymnasia, temples and shrines, a theatre with seating for 5,000, and a vast palace, some 350 by 250 metres square, in which resided Eucratides, the last king of Bactria.

The main sanctuary and temple of Ai Khanoum (Wikipedia)








On the 18th of September 1973, on the floor of one of the rooms of the palace (thenceforth known as the royal library), the excavators turned up traces of two books. One of them had been written on papyrus, the other on parchment. When the palace was sacked, the scrolls will have been chucked away, or burned, or torn up and used for hygienic purposes; but some scraps chanced to fall on compacted and absorbent clay, and although the parchment and the papyrus gradually rotted away, the ink was transferred to the clay — and, mirabile dictu, the Greek letters were still legible.

More than mirabile. You might well imagine that the clay acted like blotting paper, so that the transferred letters were mirror-images of the text (the papyrus had ‘Zeus’, say, and the clay ‘Suez’). But that was not so: the letters on the clay are the right way round, and Zeus is Zeus. How on earth can that have come about? Or can it be — but perish the thought — that the two books are a Piltdown phenomenon?

The clay on which the parchment left its mark had broken into two small pieces, each containing traces of part of a column of text. On one of the pieces a few letters could be made out (but no words), on the other a dozen more or less complete words (but no sentences). Next to nothing, then. But the longer sequences of letters all show an iambic rhythm —


for example, or


— and that is enough to show that the text was in iambic pentameters. Part of a play? Its world première staged in Alexandria-on-Oxus with King Eucratides in the royal box?

The text on papyrus is very different, and there is more of it. When the archaeologists discovered the lump of clay onto which the papyrus had fallen it was broken into eleven pieces. Reassembled, the pieces formed a rough equilateral triangle, the sides some six or seven inches long. On it the remains of four columns of writing were visible, running downward from one of the sides to the opposite apex. Each column consisted of 30 odd lines (none survives in its full height). Each line contained some fifteen to twenty letters. The Greek text is written (as was customary) without any distinction between upper and lower case letters, without accents and without breathings, without punctuation, without blank spaces to separate one word from the next. Of the first of the four columns nothing remains, not even a legible letter. In the second column 23 lines present legible letters, and a number of them present legible words. The third column is the least ill preserved: 25 legible lines, 15 of them virtually complete. Column four offers no more than parts of ten lines.

The first two surviving lines of column II are now blank. At the end of the third line there are traces of one letter — or perhaps of two. In his transcription of the text the editor, Claude Rapin, followed the normal papyrological conventions, and at the end of II 3 he printed three raised full stops (to indicate indeterminate letter-traces) followed by a lower case gamma with a dot underneath it (to indicate that only part of the letter was visible). A heterodox convention makes for greater visibility: large black dots rather than raised full stops, grey rather than under-dots, and upper-case, thus:


That is what the editor thought he saw; but the palaeographical notes to the edition say: ‘EΓ or EN/OΓ or ON’. That is to say, two letters are partially visible, the first of which is either epsilon or omicron and the second either gamma or nu.

For the fourth line the transcription is


And the palaeographical notes say: ‘AICΘHTA is probable’. Later scholars have seen ‘AICΘHTA’, and then ‘AICΘHTA’. But no later scholar has seen the lump of clay, which was deposited in a museum in Kabul: the new readings depend on a scrutiny of photographs taken by Rapin. The letters and the partial letters are very often hard to make out: different eyes will see different shapes, and the same eye will see one shape today and another tomorrow. He thought he saw a tabby cat glide smoothly into view: he looked again and saw it was the bottom of a mu: if things go on like this, he said, we shall be in a stew.

In the case of II 4 we are not in a stew; for it is pretty clear that the copyist wrote ‘AICΘHTA’, whatever can now be seen of his letters. But things are not always so; and it would be a mistake to imagine that a later scrutiny is always more accurate than an earlier. (And it should be added, to increase the uncertainty of things, that the scribe who copied the text will surely have made some mistakes, small or great — all copyists do, even photocopyists.)

Next, lines 5-7, which are transcribed like this:





The short horizontal stroke between lines 6 and 7 is a paragraphus — a sign used in ancient texts to mark a change of speaker in a play or in a dialogue. There will prove to be a further five paragraphoi later in the text (III 13/14, 17/18; IV 6/7, 7/8, 14/15). The changes are usually also marked by gaps between words; and here there is a blank space before ΦΑΜΕΓ in line 6 and before ΟΥΚΟΥΝ in line 7.

The words marked off by the two blank spaces are well preserved and readily understood. For


can only be

φαμὲν γὰρ εἶπεν

Yes, that’s what we say, he said.

That is enough to show that the text, or at least this part of the text, is a prose dialogue. Moreover, it is what ancient scholars called a ‘narrative’ dialogue — διηγηματικόν rather than δραματικόν — in which the discussion is presented as the report of a primary narrator (Diogenes Laertius, II 50). That is shown by the occurrence of εἶπεν here (and again at III 14 and IV 7). In such dialogues the speakers are generally named; and although there is no name in the surviving text (nor any room for one), but it may well be that the characters were named at the start of the discussion, and that names were later dropped.

‘That’s what we say’: who are ‘we’? and what do we say? The three dots in line 5 ask to be replaced by KAI, so that lines 5-6 must be read like this:

ἀλλὰ καὶ τὰς ἰδέας αὐτὰς ἀλλήλων

but also the ideas themselves one another

The ‘but also’ strongly suggests an earlier ‘not only’, and the ‘that’s what we say’ of the reply strongly suggests an earlier ‘Don’t we say ...?’. Thus the previous speaker will have said:

We say, don’t we, that not only ... but also the ideas themselves ...’

The ‘that’ clause must have been an accusative + infinitive, and the infinitive will have governed the genitive ἀλλήλων. What was the verb? Line 9 contains μετίσχειν, and line 11 probably μετέχειν: ‘share in’ or ‘have a share of’ — the verb, which takes a genitive, was used frequently by Plato and his followers to denote a relation between particular items and Ideas or Forms. The ideas in line 5, which recur in II 10 and III 20 and are replaced by forms or εἴδη in III 9, are surely Platonic Forms. And the particular things are ‘the perceptibles’ or τὰ αἰσθητὰ at the end of II 4 — the expression was often (and misleadingly) employed by Platonists and others to refer to particular items as opposed to the universal and imperceptible Ideas. In any event, it is plain that what ‘we say’ is this:

Not only do perceptible things have a share of Ideas but in addition Ideas have a share of one another.

For ‘share in’ scholars usually write ‘participate in’ — and Liddell and Scott give ‘participate in’ as a special sense of μετέχειν. But Platonists do not give the verb a special sense; and ‘participate’ oddly suggests that Ideas are like dances or discussions rather than like cakes and the blame. Not that ‘share’ is particularly clear: what can it mean to say that Aristides shares in or has a share of justice? Justice isn’t a pork pie, or a joint stock company.

‘We’, then — whether it means ‘you and I’ or ‘us lot’ — are Platonists, and the text found in the palace library of Alexandria-on-Oxus was a philosophical dialogue, a discussion between a couple of Platonically inclined philosophers about a central — and a controversial — part of Plato’s philosophy.

Aristides is just, and so is Alcmaeon: they’ve got something in common, namely justice. Leonora is a llama, and so is Laura: they’ve got something in common, namely the genus llama. Aristides and Leonora are ‘perceptibles’ — they are particular, individual items. Justice and the genus are not perceptibles — they are universal and imperceptible, and (according to the Platonists) they are eternal and unchanging items, first-rate Forms, ideal Ideas. What makes it true that Aristides and Alcmaeon are just is that each has a share of the Idea of justice. What makes it true that Leonora and Laura are llamas is that each has a share in the Idea of the llama.

But just as you can say things about Aristides and about Leonora, so you can say things about justice and about the llama. Justice is a virtue, and a disposition, and many other things besides. The llama is a quadruped, and a ruminant, and many other things besides. But if Aristides is just inasmuch as he has a share of the Idea of justice, then surely justice is a virtue inasmuch as it has a share of the Idea of virtue? And if Leonora is a llama insofar as she shares in the Idea of the llama, then surely the llama is a quadruped insofar as it shares in the Idea of quadrupedality? And so, as ‘we say’ and as Plato said, Ideas have shares in Ideas.

That thesis has consequences, some of them puzzling, and it is reasonable to think that the papyrus dialogue will touch on one or more of them. The next remarks begin with


οὔκουν or οὐκοῦν? (And what’s the difference?) I guess that the sense is ‘Well then ...’ and that the interlocutor asks another question ‘expecting the answer ‘Yes’’). But what question?

The transcription of II 8 reads:


Then the papyrus degenerates: but line 9 ends in


line 11 in


and line 12 in


So the discussion is still concerned with sharing, and now turns to causes or explanations; and plainly to the causes or explanations of sharing: Aristides was just inasmuch as he stood in a certain relation to the Idea of justice — but what caused him to stand in that relation to that Idea? The llama is a quadruped insofar as it shares in the Idea of quadrupedality — but why does it share in that Idea rather than in the Idea of bipedality? A hypothetical reconstruction of II 7-10 has the interlocutor suggest that it is the same case in the two cases.

The hypothetical reconstruction is rash, and even the claim that the discussion turns to the causes of sharing may seem audacious — it deduces a lot from a little (from ten legible letters, and two dozen letter-traces). But the deduction is confirmed, or at least supported, by the text of columns III and IV where sharing and causes reappear. In particular, consider III 6-7:




ἀναγκαῖον εἶναι τὸ τῆς μεθέξεως αἴτιον

necessarily the cause of the sharing is

Then ‘cause’ is plain at III 16. ‘share’ is probably present at IV 9 and 10.

As for ‘necessarily’, lines III 14-15 are transcribed like this:




So III 14 reads

ἀναγκαῖον εἶπεν

Necessarily, he said.

And, as the paragraphus and the blank space show, the second speaker agrees with what his interlocutor had said, and ‘necessarily’ picks up ‘necessarily’ in III 6 just as ‘we say’ in II 6 picked up ‘we say’ in II 3.

The text at III 5-21 is in fact the best preserved part of the papyrus. It may be translated, with some little confidence, as follows:

Necessarily, the cause of the sharing is unchanging [ἀκίνητον]; for each of the Forms is unchanging both for these reasons and because the coming-into-being and perishing of the perceptible items is eternal [τὸ τὴν γένεσιν εἶναι καὶ τὴν φθορὰν ἀίδιον].

Necessarily, he said.

And this must be deemed to be the most authoritative [κυριώτατον] and first of the causes.

The cause of the sharing — whatever it is which causes Aristides to have a share of justice and Leonora to have a share of the Ideal llama — is unchanging, just as the Ideas are; but it is not itself: it is some third and most authoritative kind of thing.

If the general drift of III 5-21 is clear, the details are not. What are ‘these reasons’? The fact that Ideas as well as perceptibles share in Ideas? That might perhaps explain why the cause is most authoritative (since it is authoritative over the Ideas); but does it presuppose that the cause is changeless? Again, if coming-into-being goes on forever, then perhaps its cause must be eternal. But why should its cause be changeless?

And there is a more pressing question: why suppose that the cause of sharing is a third kind of thing? Doesn’t Plato hold that the Ideas are themselves causes? In the Phaedo Socrates asks why a certain body is hot: the safe answer is ‘Because it has a share of the Idea of the hot’, a more adventurous answer is ‘Because it has a share in the Idea of fever’ — and any true and satisfactory answer will invoke an Idea. True; but the question in the papyrus dialogue is not — or not only — ‘Why is this item thus-and-so?’ (‘Why is this body hot?’). It is — or it is also — about how things become what they are: ‘Why does this item become thus-and-so?’ (‘What made this body become hot?’). If the body is hot inasmuch as it shares in the Idea of heat, how did it come to be hot? how did it acquire the Idea of heat? Suppose it acquired that Idea thanks to an Idea, inasmuch as it had a share in some Idea: what could that further Idea be? The Idea of acquisition of the Idea of heat? Then, in general things come to be thus-and-so insofar as they share in the Idea of coming to be thus-and-so ... That way madness lies — or at any rate an infinite regress. Whatever the cause of sharing may be, it is not an Idea.

Those lines of thought echo, or at any rate are vaguely similar to, certain passages in Aristotle’s Metaphysics. For example, Book Eta raises the question of how it is that substances are unities: llamas, say, are ruminant animals — but how is it that those two elements, ruminant and animal, bond together to form a unity? According to Aristotle,

because of that puzzle, those who speak of sharing are puzzled as to what the cause of sharing [αἴτιον τῆς μεθέξεως] is and as to what it is to share (1045b7-9).

The Platonists wondered how rumination and animalhood hung together in such a way that llamas (say) were unified substances and not simply bundles of qualities; and their puzzlement led them to wonder what causes sharing — and indeed to wonder what it is for one thing to share in another.

A second pertinent text comes in Book Lambda:

No-one says why there will always be coming-into-being or what the cause of coming-into-being is. Those who posit two principles must say that there is another more authoritative [κυριώτερον] principle, and those who posit forms must do so too: for why did it or does it share? (1075b12-20)

Earlier philosophers who have attempted to explain the nature of the universe have failed to solve a number of problems, one of them being why or for what cause there is an eternal sequence of generation and corruption, of coming-into-being and perishing. The Platonists, who appeal to sharing, must posit a further and more authoritative principle to explain why such participation comes about, world without end. That is what the Platonists in the papyrus dialogue are doing. Not only that: the most authoritative principle of the dialogue is a changeless changer — and what is that but Aristotle’s own first cause of everything, his Unmoved Mover or (as the Greek is better translated) his Changeless Changer?

All that is heady stuff: some philosophers call it metaphysics (a word unknown to the Greeks): liberal shepherds will find a grosser name. For isn’t the whole edifice constructed on mud? After all, what does it mean to say that Aristides shares in justice, or that llamas share in rumination? Isn’t it simply to say, in a grandiose jargon, that Aristides is just and that llamas are ruminants? In that case, to ask why or for what cause Aristides shares in justice and llamas share in rumination is to ask why Aristides is just and what causes llamas to ruminate. Those particular questions may perhaps receive particular answers — and very different particular answers. The Platonic question is not particular but general: it asks why things share in Ideas, or why it is that things are thus-and-so. But there is no general answer to that general question. An Irish philosopher remarked of some of his contemporaries that they raise a dust and then complain they cannot see. Liberal shepherds may same the same of the Platonists.

There is, to be sure, a great deal more to be said about those dusty issues. And the papyrus also excites more down-to-earth questions: who wrote it? and when? and how on earth did it find its way to the banks of the Oxus?

As to the date, the papyrus was certainly inscribed before 145 BC, when Alexandria-on-Oxus was largely destroyed. On the basis of the style and form of the Greek letters papyrologists opine that it dates from the middle of the third century BC. However that may be, the date of the papyrus copy gives not a date but only a terminus ante quem for the composition of the text, and the content of the text suggests a period some decades before 250 BC. For although the esoteric questions raised by Plato’s theory of Ideas greatly exercised the minds of Plato’s early followers (and also the mind of Aristotle), the evidence, such as it is, strongly suggests that the questions had lost their fascination by the first decades of the third century, and that deep philosophical minds had turned to other deep philosophical matters. In any event, nothing in the text is formally inconsistent with a date as early as, say, 367 BC — the year in which Aristotle left the wastes of Macedonia and came to Athens and to Plato’s Academy.

And in fact most scholars who have dared an opinion have implicitly opted for a date before 323, the year in which Aristotle died. For they have attributed the dialogue to Aristotle: it is a fragment from his Sophist, or from the critical essay On Ideas, or from On Philosophy, or (of course) from an otherwise unknown work. The suggestions are beguiling — Aristotle in Afghanistan ...

Yet there seems to be a devastating objection: the participants in the dialogue are discussing what ‘we say’, and what they say proves that they were speaking as committed Platonists. Aristotle, on the other hand, was a merciless critic of the Theory of Ideas and all who sail in her — he can hardly have written a dialogue in which he assumes the part of a questing Platonist.

It might be replied that the author of a dialogue in which the participants are Platonists need not himself be a Platonist; but Cicero refers to ‘the Aristotelian custom’ of writing dialogues ‘in which the remarks of the others yield the leading part to himself’ (ad Att XIII xix 4): in Aristotle’s dialogues the chief speaker was Aristotle. Other replies to the objection may be concocted; but the favoured way of dealing with it appeals to a familiar and striking fact about Book Alpha of the Metaphysics. There, where he discusses various aspects of Plato’s philosophy, among them his notion of Ideas, Aristotle speaks in the first person plural: of various Platonist opinions he say ‘we show’ this (990b9), ‘we think’ that (b11), ‘we say’ the other (b16) ... He writes as a Platonist; and the common and simple explanation of that practice proposes that when he wrote the passage Aristotle was a Platonist. There, then, is a parallel to the papyrus dialogue: it was written by Aristotle when he was still a follower of Plato, and as Metaphysics A shows, keenly engaged with Plato’s Ideas.

That disposes of the ‘devastating objection’ to Aristotelian authorship of the dialogue. But it does not constitute evidence in favour of Aristotle. The favouring evidence consists entirely in the similarities between the dialogue and some passages in Aristotle’s Metaphysics. Even if the similarities are not accidental, they show at most that the author of the dialogue was acquainted with some parts of Aristotle’s work. To be sure, it is consistent with everything in the papyrus that the dialogue was written by Aristotle; but a prudent scholar will affirm nothing more rousing than that banality. After all, what do we know about the dialogue? Not its title, not its length, not the position or the function within it of the surviving fragment, not its overall subject (if it had one), not its general attitude to Platonism and the Platonic Ideas (ditto), not its purpose. Was it — to take one extreme hypothesis — a long and serious discussion of some of the perceived problems with Platonism? Or was it — to swing to the opposite extreme — part of the commonplace book of a Macedonian veteran who relieved the tedium of life in the Bactrian backwoods by recalling the intellectual excitements of his lost youth?

So much, or so little, for the question of authorship. Next, how did the dialogue get to the banks of the Oxus (if it was not written there)? But that is another story.

Jonathan Barnes FBA was a Fellow of Oriel College Oxford from 1968 to 1978 and a Fellow of Balliol from 1978 to 1994. From 1994 to 2002 he was Professor of Philosophy at the University of Geneva, and from 2002 to 2006 he held a chair in philosophy at the Sorbonne. His many publications on ancient philosophy include: Aristotle: a Very Short Introduction (Oxford 2000), Coffee with Aristotle (Duncan Baird, 2008), Early Greek Philosophy (Penguin, 1987), The Presocratic Philosophers (Routledge 1982), and the revised Oxford translation of the Complete Works of Aristotle (2 vols, 1998).

Front cover of four books: Coffee with Aristotle; Aristotle: a Very Short Introduction; The Presocratic Philosophers; the revised Oxford translation of the Complete Works of Aristotle