I

A second-century papyrus from Oxyrhynchus, published as P.Oxy. 2456, preserves the final section of a complete list of the plays of Euripides. The text is on the verso (the recto contains a tax register, also from the second century), and may have been a library catalogue; nothing in particular suggests a school exercise.Footnote ¹ The plays are arranged by alphabetical order of first letter, which is how alphabetical order originally worked in antiquity from its earliest Greek uses in the third century BC.Footnote ² Until recently, we could assert that the use of subsequent letters as tie-breakers when words begin with the same letter is not attested before the second century AD, and even then is a sporadic practice until much later.Footnote ³ A recently published lexicon of poetic and dialectal words dating to the third or second century BC has overturned that picture: its organization by complete alphabetical order shows that the practice was known already in Ptolemaic times.Footnote ⁴ Nevertheless, this remains an isolated example, and our papyrus clearly uses alphabetical order in the usual ancient way.

Here is the list of plays, with the Greek title (as it appears in the papyrus):Footnote ⁵

^a Scholars tend to place this play in the first half of Euripides’ career (Collard and Cropp (Reference Collard and Cropp2008) 2.161) or more specifically between 440 and 429 (Moles (Reference Moles, Bruno, Filosa and Marinelli2022) 247–49), on the basis of its subject matter, tenuous grounds for such a conclusion.

^b Collard and Cropp (Reference Collard and Cropp2008) 1.xxxii place the drama among those with unknown dates, though they add ‘437–424?’ in parentheses; then at 2.171 they note that ‘a date in the 430s has been inferred’ on the basis of an emended place name in one of the testimonia to the play advocated, on no secure basis, by Wilamowitz (discussed at Magnani (Reference Magnani2022b) 38–39).

^c These resolution data are much less secure than in the case of other plays, based as they are on the proposal of Van Looy (Reference Looy1964) 132–84 (especially p. 176) that some fragments attributed simply to ‘Phrixus’ belong to this play.

From this table, the following points emerge:

1. 1. That some kind of secondary organizational principle is at work is suggested by the conspicuous separation of the two plays named Phrixus. If there was no such secondary principle, there would be a natural tendency for two homonymous works to appear next to each other, or even for them to be combined into a single entry.

2. 2. The nature of that principle is strongly suggested by the relative positions of three pairs of plays: Stheneboea and Sisyphus; Telephus and Trojan Women; Philoctetes and Phoenician Women. For these six plays we have explicit ancient evidence either recording their exact date, or giving a firm terminus ante quem (Stheneboea) or a narrow range of possible dates (Phoenician Women). For each pair, the chronologically earlier play comes first. The probability of this happening at random is 1 in 8 (½ x ½ x ½).

3. 3. A further play, Phoenix, has a firm terminus ante quem that permits (on a chronological hypothesis) the order Phoenix, Philoctetes, Phoenician Women or Philoctetes, Phoenix, Phoenician Women, but not Philoctetes, Phoenician Women, Phoenix. The order in the papyrus is the first of these sequences, so consistent with the chronological theory, and thus reducing the probability that this order is the product of chance to 1 in 12 (⅛ x ⅔).

4. 4. Two further plays can be assigned date ranges deduced from resolution data; that is, from Euripides’ increasing tendency over time to resolve long syllables in iambic trimeters into two short syllables, and to make greater use of anapaestic substitutions:

   1. a. The date range for Temenidae permits (on a chronological hypothesis) the order Telephus, Temenidae, Trojan Women or Telephus, Trojan Women, Temenidae, but not Temenidae, Telephus, Trojan Women; the order in the papyrus is the second of these sequences, so consistent with the chronological theory, and thus reducing the probability that this order is the product of chance to 1 in 18 .

   2. b. The date range for Phaethon permits (on a chronological hypothesis) the order Philoctetes, Phaethon, Phoenician Women, but not Phaethon, Philoctetes, Phoenician Women or Philoctetes, Phoenician Women, Phaethon; the order in the papyrus is the first of these sequences, so consistent with the chronological theory, and thus reducing the probability that this order is the product of chance to 1 in 54 .

5. 5. The resolution data for Phrixus B are less secure than for the other plays (n.7 below). But if accepted, the ensuing date range permits (on a chronological hypothesis) the order Phrixus B, Philoctetes, Phoenician Women or Philoctetes, Phrixus B, Phoenician Women, but not Philoctetes, Phoenician Women, Phrixus B; the order in the papyrus is the second of these sequences, so consistent with the chronological theory, and thus reducing the probability that this order is the product of chance to 1 in 81 (¹⁄₅₄ x ⅔).

6. 6. We began by noting that some kind of organizational principle beyond alphabetical order of first letter was probably at work here; we have now identified that the probability that the chronological sequence for each letter offered by the papyrus was achieved at random is scarcely more than 1 per cent. There is no room for reasonable doubt here: this arrangement was deliberate.

II

Still more evidence in favour of a deliberate chronological arrangement is provided by the case of Tennes. This play was part of the collection transmitted under Euripides’ name, as we know from the inclusion of an hypothesis (prose summary) of the drama among the standard collection (see Section IV.3 below), as well as a citation of ‘Euripides’ Tennes’ in Stobaeus.Footnote ⁶ But the first Byzantine abridgement of the ancient Life of Euripides asserts that this play, along with Rhadamanthys and Pirithous, was not by him;Footnote ⁷ this will reflect a statement in the original, Hellenistic version of the Life that they were of disputed authenticity.Footnote ⁸ Whether we credit that statement, or rely instead on the indubitable presence of the play in the Euripidean transmission, is not relevant here; for our purposes, the crucial point is that the authenticity of the play was doubted by some. Its placing in the papyrus as the final play beginning with tau thus probably reflects the absence of a didascalic production record;Footnote ⁹ lacking a date, its natural place in a chronological system was after all the plays that did have one. Less probably, Tennes did have a date, but was placed last, probably out of chronological sequence, because its authenticity was otherwise suspect; compare (mutatis mutandis) how the Pindaric epinikia which did not belong to any of the crown games were added by the Alexandrian editor to the very end of that edition, in last place among the Nemeans,Footnote ¹⁰ or how the final poem of Sappho book 2 seems to have been somehow separate from the rest of that volume.Footnote ¹¹ The least likely option is that Tennes received its place simply as the consequence of its date; not only would it be quite a coincidence that the one play on the papyrus whose authenticity was questioned in antiquity just happened to be chronologically last among its grouping, but also it would be strange for a play from so late in Euripides’ career, the period from which plays were most likely to survive and for which records were presumably best, to generate controversy as to its authenticity in the first place.

Sisyphus also needs further consideration here. There is no explicit ancient claim that this play was not by Euripides;Footnote ¹² ancient evidence, as noted in the table above, rather tells us that this was the name of the satyr play of Euripides’ tetralogy of 415. But we possess a long speech from an unnamed drama (whether tragedy or satyr play, we are never told), cited by Sextus Empiricus and said by him to be by Critias; but said by Pseudo-Plutarch (generally thought to be Aëtius), who cites the speech in part, to belong to Euripides, and to have been delivered by Sisyphus. On this fragile basis, some modern scholars have posited that Euripides’ Sisyphus of 415 did not reach Alexandria, and that the play from which this speech comes, which they call Sisyphus, replaced it in the Euripidean corpus, though evidently without securing ancient consensus that it was by Euripides.Footnote ¹³

In this connexion the failure of P.Oxy. 2456 to designate Sisyphus as a satyr play, the only place in the papyrus where that particular indication is missing, attracts attention.Footnote ¹⁴ This could easily be just a slip,Footnote ¹⁵ though if so, it is the only one in the surviving text. If it is not a mistake and the writer of our list knew a non-satyric Sisyphus, that cannot have been the Euripidean play of 415. Occam’s Razor suggests that it was the play credibly called Sisyphus by modern scholars, the one ascribed by ancient sources alternatively to Critias and to Euripides. In that case, and under this hypothesis, it was placed last among the sigma plays because the writer of our list questioned its ascription to Euripides. Yet the content of one of the fragments quoted from Euripides’ Sisyphus is strongly satyric in character,Footnote ¹⁶ something that also tells against the idea that P.Oxy. 2456 omits σατυρικός because Sisyphus was more like Alcestis (a play that came fourth in its tetralogy, but omitted satyrs) than a satyr play proper.Footnote ¹⁷ Positing a mistaken omission in P.Oxy. 2456, then, is the best way of accounting for the evidence as a whole here.

III

We have determined that the plays in P.Oxy. 2456 are ordered, first, by alphabetical order of first letter, and second, by their original performance date, as evidenced by the didascalic record.Footnote ¹⁸ As a consequence, these data provide a relative order, and/or termini ante vel post quos, for 15 plays, something of obvious importance for analyses both of the plays themselves and of Euripides’ dramatic career as a whole. I summarize the new information (and only that) for each play below:

To cite just one example of how these new data may be put to immediate use, it has long been argued that Temenus and/or Temenidae formed part of a connected trilogy with Archelaus, a play in honour of the Macedonian king of that name who came to the throne in 413, which must therefore post-date that year.Footnote ¹⁹ P.Oxy. 2456 disproves this idea, showing as it does that Temenus was first performed no later than 439. But it remains possible that Temenidae, whose first performance is now revealed to be no earlier than 414, was performed at the same festival as Archelaus, whether or not any third or even fourth play was involved.Footnote ²⁰ At the very least, this time frame makes it eminently plausible, indeed probable, that Temenidae was intended for performance in Macedon, whether or not it was also produced in Athens.

It follows that, of the fragmentary papyrus hypotheses previously identified as belonging to either Temenus or Temenidae, those fragments which mention Archelaus as a character must belong to the latter, since there is no reason to think that the myth had an Archelaus figure before Euripides inserted one to flatter the Macedonian king.Footnote ²¹ We also see that Euripides treated the myth of Temenus and the division of the Peloponnese between his sons at two significantly different times: one well before the Peloponnesian War, perhaps at a time of peace between Athens and Sparta, the other when that war was long underway. The whole question of the relationship between these plays needs a fundamental reassessment in the light of this discovery.Footnote ²²

IV

This is not the only list of Euripidean plays which has come down to us. There are five others, which all use alphabetical order of first letter as their primary organizational principle:

1. 1. An inscription from the Piraeus, dating to the late second or early first century BC, contains several authors’ names (mainly but not exclusively dramatic poets of the fifth and fourth centuries) in the genitive, each followed by a list of titles.Footnote ²³ The authors’ names are given in no particular order overall, though some beginning with the same letter are adjacent. As for the titles of their works, some are organized according to their first letter, whereas others are not. The unusual arrangement of the inscription as a whole makes it unlikely that it presents the order of a library catalogue; it may rather reflect a list of books given by ephebes to the library of their gymnasium, the order of the works corresponding, at least in part, to the order in which the books were given.Footnote ²⁴

   The plays of Euripides are organized by their first letter, yet in the order sigma, theta, delta, pi (though among the pi plays occurs a mysterious sequence ΑΛΑΙ), phi, alpha, epsilon. In the sole case in the list of the same title given to two Euripidean dramas, Phrixus, the name may have occurred twice, there being a gap at the end of the phi-sequence where a second instance of the name would fit. As Wilamowitz saw, no secondary principle of chronological arrangement is at work here, since Euripides’ first play, Daughters of Pelias, is placed second in the list of plays beginning with pi.Footnote ²⁵ Wilamowitz additionally claimed that tragedies and satyr plays are not listed separately, though the evidence for that is mixed:Footnote ²⁶ for sigma we find two tragedies followed by three satyr plays, yet for epsilon, where only the first name survives, that name is Eurystheus, a satyr play. For the plays beginning with sigma and phi, the only ones with which we can compare the order of P.Oxy. 2456, the two sequences are clearly distinct.

2. 2. The Marmor Albanum, a second-century AD statue of Euripides from Rome accompanied by a list of his plays.Footnote ²⁷ We have the list from its beginning as far as eta, after which some names are lost, and then from kappa as far as omicron, where the stone-cutter ceased copying; hence there is no overlap with the order of P.Oxy. 2456. (The ordering does overlap with that of the Piraeus inscription, and is clearly different from it.) In the case of homonymous plays, the title is given only once, with no addition to indicate that it stands for two dramas. A secondary principle of organization can be observed: for each letter, tragedies are placed first, followed by satyr plays. No further principle is evident, certainly not a chronological one: Bacchae appears ahead of Bellerophon; Andromeda ahead of Alexander ahead of Andromache; Helen ahead of Hecuba.

3. 3. The main series of Euripidean hypotheses is attested across a series of papyri which offer a consistent ordering distinct from those of both the Piraeus inscription and the Marmor Albanum.Footnote ²⁸ A secondary organizational principle of chronology was definitely not used, since we find Alexandros immediately preceding Andromache.

   The ordering of the hypotheses is distinct from that of P.Oxy. 2456, too, with Syleus moving straight to Temenus without Sisyphus intervening, and no Phaethon between Philoctetes and Phrixus B.Footnote ²⁹ Yet similarities can also be discerned. Both sources put Scyrians first among the sigma plays, both contain the sequence Hypsipyle, Phrixus A, Phoenix and the sequence Phrixus B, Phoenician Women, Chrysippus, and both begin the tau plays with Temenus. Overall, the two orderings are closer than any other two mentioned in this section.

4. 4. A different set of hypotheses, found in only one papyrus, contains six plays: Busiris (probably), Bacchae, Dictys, Danaë, Helen, Heracles.Footnote ³⁰ The first four are contained in the four columns of a single piece of papyrus. The last two are found in two separate fragments, but probably come from the two columns immediately after the first four; if so, this papyrus is not from a complete collection of Euripides’ plays.Footnote ³¹ This might have chronology as a secondary ordering principle, but too few hypotheses survive for that possibility to be verified. Satyr plays were not placed after tragedies, since Busiris precedes Bacchae.

   There is no overlap with P.Oxy. 2456. The order Dictys followed by Danaë occurs in both the Piraeus inscription (via a safe restoration) and the Marmor Albanum, but the latter has Bacchae ahead of Busiris. The relative position of these pairs of plays in the main series of Euripidean hypotheses is unknown.

5. 5. The so-called ‘alphabetic’ plays of Euripides which survive thanks to a single 14th-century manuscript, Laurentianus plut. 32.2 (Diktyon §16268), had their order disturbed in the course of transmission, and so do not appear there in alphabetical order of first letter.Footnote ³² But the four iota plays (of which three are adjacent) are in chronological order,Footnote ³³ as are Children of Heracles and Heracles (adjacent); so, too, from the so-called ‘select’ plays, are Alcestis and Andromache (adjacent).Footnote ³⁴

The diversity of arrangements reflects the realities of transmission; as Chiara Meccariello says, the tendency of plays to fill an entire roll means that it would be harder for any one ordering to prevail over others.Footnote ³⁵ Yet there is sufficient similarity between the ordering of P.Oxy. 2456 and that of the main series of hypotheses to make a common origin plausible. Since P.Oxy. 2456 is the one with a clear secondary organizational principle, it would preserve the older order in its pristine form, from which the main series of hypotheses would have diverged slightly. The alternative, perhaps incomplete set of hypotheses, if it had a secondary organizational principle of chronology, would also presumably derive from the ordering which P.Oxy. 2456 attests.

V

What about lists for other Greek dramatists? In the list that follows, all the papyri are from the second century unless otherwise specified:Footnote ³⁶

1. 1. A list of comedies by Aristophanes in the 14th-century manuscript Ambrosianus L 39 sup. (Diktyon §42949) fol. 90v, ordered first by alphabetical order of first letter, then by chronology, an arrangement identified by Wilamowitz (who noted that the chronological ordering had been disturbed during transmission) immediately after its publication by Novati; hence its subsequent designation as the ‘Index Novati’.Footnote ³⁷ This provides a definite parallel to the arrangement in P.Oxy. 2456. Here is the list:Footnote ³⁸

   

   The alpha and beta plays are clearly chronological. From gamma to epsilon the entries are hopelessly mixed up; but the lambda plays could be chronological, as could the tau. The omicron plays are chronological, as are four of the five pi plays, with the exception of Proagon. For nu and pi, the final plays are those which are among the four plays said by the ancient Life of Aristophanes not to be by him, with some (according to the Life) saying that they were by Archippus. That they are found in last place each time, just as Tennes was in the list of Euripidean plays, provides strong support for the idea that Tennes was placed in final position because of its suspect authenticity; for all three plays we may infer the absence of a didascalic record.

2. 2. A list of comedies by various authors written on the verso of a Greek–Latin glossary (itself from the late first or early second century). In each case the list is in alphabetical order of first letter; the list for Aristophanes is not chronological for each letter, since the alpha dramas are in a different order from what is found in the Ambrosianus.Footnote ³⁹

3. 3. A list of dramas by Cratinus, of which six survive, is not ordered alphabetically, though it might be chronologically.Footnote ⁴⁰

4. 4. A versified list of dramas by Epicharmus contains six plays, beginning pi–pi–omicron–omicron–mu–pi:Footnote ⁴¹ so not in alphabetical order of first letter, but perhaps derived from such a list, since all six plays begin with a letter from a span of only four.

5. 5. An alphabetical list of Menandrean comedies, written on the back of a land register, is in alphabetical order of first letter.Footnote ⁴² We lack sufficient knowledge of the chronology of Menander’s plays to check whether a secondary, chronological principle of organization was in operation, but one indication tells strongly against it: the two plays called Brothers are included on a single line, as Ἀδϵλφοί ᾱβ̅.Footnote ⁴³ It would be unusual if Menander’s two plays beginning with alpha were chronologically consecutive among his plays whose titles began with this letter; the opposite is more likely, with the playwright allowing some time to elapse before tackling a similar topic again.

6. 6. Another list of Menandrean comedies written as part of a shorthand manual contains no obvious organizing principle.Footnote ⁴⁴

7. 7. A list in full alphabetical order (so not just first letter) of the plays of Aeschylus is transmitted in the mediaeval manuscripts.Footnote ⁴⁵ Its order, which contains only a few mistakes, must have been put together at a late stage in the transmission; it cannot reflect Callimachus’ Pinakes (on which see below),Footnote ⁴⁶ which were composed centuries before full alphabetical order was first used.

VI

Novati proposed that his Index had its origins in Callimachus’ Tables of Those Distinguished in Every Form of Learning, and of Their Writings, in 120 Books Footnote ⁴⁷ ∼ Πίνακϵς τῶν ἐν πάσῃ παιδϵίᾳ διαλαμψάντων καὶ ὧν συνέγραψαν ἐν βιβλίοις κ´ καὶ ρ´ – the alphabetical list, today usually known simply as Callimachus’ Pinakes, which catalogued the riches of the Library of Alexandria.Footnote ⁴⁸ That was a plausible suggestion, given the Library’s unique cultural impact. But it became particularly attractive when Wilamowitz discovered the Index’s secondary principle of arrangement. Callimachus was working with the Aristotelian Didascaliae, which recorded plays in chronological order.Footnote ⁴⁹ As he arranged them by the new method of alphabetical order of first letter, it would have been the natural, frictionless choice to retain the chronological order in the case of plays which began with the same letter when transferring them to his new system. (Callimachus’ interest in the chronology of drama is attested by his separate work, the Table and Register of Dramatic Poets in Chronological Order from the Beginning Footnote ⁵⁰ ∼ Πίναξ καὶ ἀναγραφὴ τῶν κατὰ χρόνους καὶ ἀπ’ ἀρχῆς γϵνομένων διδασκάλων.) By contrast, a subsequent rearrangement of plays on a secondarily chronological basis by someone else, perhaps long after they had been alphabetized, would have necessitated specific consultation of their performance dates to achieve that end, something that could have been done, I suppose, but what would have been the point?

More recently, Dirk Obbink has made a more expansive assertion regarding Callimachus’ practice:

The ‘Index Novati’ is the only evidence that Obbink cites for this claim. Nevertheless, we now have Euripidean evidence; and the existence of such a distinctive organizational system for two dramatists (one tragic poet, one comic) points strongly towards a single source, who must be a significant enough figure in the history of transmission to have had such an influence. The case for Callimachus is thus considerably strengthened.

The existence of other lists which employ only alphabetization by first letter, or that plus a division between tragedies and satyr plays, is no argument to the contrary. There would have been a tendency to simplify Callimachus’ system by retaining its readily comprehensible alphabetical part, but discarding the chronological aspect, something much harder to identify when used as a secondary, rather than a primary, organizational principle. After all, it has taken modern scholars until 2023 to identify that principle in the case of a papyrus published in the year of the Cuban Missile Crisis; there is no reason to expect ancient copyists, most of whom will have had no special interest in the chronology of classical drama, to do better.

Further support for the idea of a Callimachean origin comes from the main series of hypotheses. Its arrangement by alphabetical order of first letter ‘clearly looks back to a complete and ordered edition of Euripides’, and their introductions through title and opening line ‘are themselves derived from a definitive edition or catalogue … This is how works were entered in Callimachus’ Pinakes’.Footnote ⁵² Chronologically consistent with such a derivation is the fact that the hypotheses post-date Hegesias of Magnesia, who flourished ca. 300 BC and invented the Asianic prose on whose rhythms their author drew.Footnote ⁵³ We have seen that the ordering of the hypotheses is similar enough to that of P.Oxy. 2456 to posit a common origin for them both. It would be economical, to say the least, if that common origin was Callimachus.