The Clock in the Verse

Nineteen of Euripides' plays survive, and only eight have a year fixed by evidence outside the plays themselves. The rest are dated by counting syllables. Here is that count, run in your browser over every spoken line, and the test a clock ought to face: hide a date it knows, and see whether it can find it again.

Euripides was credited in antiquity with about ninety plays. The Suda gives two traditions: According to some, he wrote 75 plays; others say 92, but 77 survive. Nineteen reach us more or less whole, and one of those, Rhesus, may not be his. For a few the date of the first performance is on record, because the ancient summary attached to the play (the hypothesis) copies the official production list, or because another writer mentions it. For the others, the date printed in every edition and on every reading list (Electra, c. 420) comes mainly from the verse itself.

The reason is a habit. The spoken lines of Greek tragedy are iambic trimeters: twelve positions, alternating short and long, and a poet may split a long position into two short syllables. That split is called a resolution. Euripides allowed himself very few of them early in his career, and more and more as it went on. J. Gottfried Hermann noticed the trend in 1807 and drew the obvious conclusion: count the resolutions in a play, and you know roughly when it was written.

A resolution adds exactly one syllable to the line. So once you know that a line is a trimeter, counting its resolutions is simple arithmetic: the number of syllables minus twelve. The hard part is deciding which lines are trimeters, and how many syllables each one has.

Fetching the text of nineteen plays…

Here is what the scanner does, and all it does. It marks η, ω, diphthongs, and any vowel with a circumflex or iota subscript as long by nature. It marks ε and ο as short, and a vowel before two consonants as long by position. α, ι and υ it leaves undecided, because their length is not shown in the spelling. A short vowel before a stop followed by λ, ρ, μ or ν is also left undecided, because Attic verse can treat that syllable either way. It then asks whether the syllables can fill the pattern x – ⏑ – | x – ⏑ – | x – ⏑ x, where any long or x position may be two shorts. By default it merges two vowel pairs that tragedy regularly runs together (synizesis): the genitive ending -εως, and μή, ἤ, ἐπεί or ἐγώ before οὐ. It tries any further merge only when a line will not scan otherwise. A line counts as a spoken trimeter only if it scans and at least three of the four lines around it scan too: the dialogue comes in long runs, and a sung line that happens to fit the pattern usually stands alone. Lines Murray printed in square brackets, as later insertions, are left out.

The clock, and the test it has to pass

Eight plays have a date from outside the verse. These are the anchors. The evidence for each one, from the Oxford Archive of Performances of Greek and Roman Drama:

Before fitting anything, I wrote down the test (the pre-registration, unchanged since). Fit a straight line of year against rate on seven anchors. Ask it to date the eighth. Repeat for each anchor. If the typical miss (the root mean square) was five years or less, the clock would count as working to five years. If not, the page would say so.

It failed, narrowly. The typical miss is 5.3 years, not five. Helen and Medea are placed within half a year. Alcestis comes out almost ten years too late. And the clock does not tick steadily at either end. Hippolytus (428) has fewer resolutions than Alcestis (438) or Medea (431), so the early plays sit on a flat stretch where the count cannot tell a decade apart. At the other end, Orestes (408) has more resolutions than either of the plays staged after the poet's death. Across the middle of the career, where the count climbs about one and a half resolutions per hundred lines a year, it does what Hermann said it would. What it cannot do is order two plays a few years apart, or separate the plays of the 430s from one another.

There is one free check the fit never saw. Phoenician Women was kept out of the anchors because its outside evidence gives a range, not a year: the APGRD record puts it in 410 or 409, and Seaford and Murray give 411 to 409. The clock, not told this, says 413, give or take five: two to four years early, inside its own error.

Is the problem the counter?

A crude scanner run over an OCR'd text might make the clock look worse than it is. So here are two counts made by hand. E. B. Ceadel published one in 1941, and it is the count Seaford and the Wikipedia table both use; eighteen of his figures are reproduced in the Wikipedia table on Euripides, and ten of them are quoted by Richard Seaford in 1982. C. F. W. Müller published a count for all nineteen plays in 1866, in a footnote to a Latin monograph that is now public on the Internet Archive. Both leave out resolutions inside proper names, which this scanner cannot recognise, so their figures run lower than mine. The question is whether they run in the same order.

They do. Three counts made in 1866, 1941 and today, two by hand and one by machine, put the nineteen plays in nearly the same order. And the test gives the same verdict on all three: an error a little over five years, and the same four anchor pairs out of order. You can switch the clock above to either hand count and watch. The clock's coarseness belongs to Euripides, not to the counter.

What the clock says about the other eleven

Pick a name on the right-hand strip of the clock to read its date. The clock puts Children of Heracles about 431, Andromache and Suppliants about 425, Hecuba 423, Electra 422, Heracles 421, Ion 417, Iphigenia among the Taurians 416. That is close to the dates in the handbooks (Kovacs gives c. 430, 425, 423, 424, 420, 416, 413, 414). This agreement is not evidence: those handbook dates were reached largely by this same count, so the clock is just agreeing with itself. Only the held-out test above says how far to trust it.

Two plays need a warning. Rhesus has the resolution rate of early Euripides, and the clock dates it about 429. But the clock assumes Euripides wrote it, and Kovacs judges it probably a fourth-century play by a now nameless poet. A low rate is just as consistent with a different, more conservative poet, and the count cannot tell the two apart. Cyclops is a satyr play, the comic piece that followed a set of three tragedies. Seaford, who counted it himself at 35.0 by Ceadel's rules, put the difficulty exactly:

In Cyclops it is 35.0, which would entitle us to place the play after 412 BC, were it not that the figure may express no more than the greater metrical freedom that satyric drama has over tragedy.Richard Seaford, “The date of Euripides’ Cyclops”, Journal of Hellenic Studies 102 (1982), p. 161

Why these nineteen: two shelves

The nineteen did not survive as one collection. They came down by two different routes, and the second route was luck.

Sometime after ca. A.D. 250 a group of ten plays, which we call “the select plays,” were chosen, perhaps for school use, and increasingly only these were copied. By fortunate chance, however, one or two volumes of the Collected Works, containing nine plays beginning with the letters epsilon, eta, iota, and kappa, survived into the Byzantine period and were copied onto a single manuscript now in the Laurentian Library in Florence.David Kovacs, “Euripides”, Perseus Encyclopedia

The select plays came with ancient commentary (scholia) in the margins, and many medieval manuscripts preserve them. The nine alphabetical plays have no commentary, and they survive in only two manuscripts, known as L and P. Sort all nineteen titles in Greek alphabetical order and you can see the volume:

The nine plays from Ἑλένη to Κύκλωψ form one continuous stretch of the alphabet. The only break is Ἱππόλυτος, whose title falls in the same stretch but which was also one of the select plays. Any other plays of his whose titles began with those letters are lost. What the count adds is that the lucky volume was not a slice of one period. Its nine plays run from the flat early end of the clock (Children of Heracles, 6.9) to the steep late end (Iphigenia at Aulis, 43.6), much as the select ten do. Without that one volume we would not have Helen or Iphigenia at Aulis, two of the eight anchors, nor most of the middle of the clock: Electra, Heracles, Ion and Iphigenia among the Taurians.