ElAnt v1n8 - Chronology and Narrative Apparatus in Plato's Dialogues

Volume 1, Number 8
April 1994


Harold Tarrant,
Department of Classics,
University of Newcastle,
N.S.W. 2308,
e-mail: CLHAST@cc.newcastle.edu.au

To confine a study of Platonic chronology to features of the dialogue which can only be applied to narrative dialogues may at first sight seem bizarre. The works concerned are (alphabetically) Charmides , Erastae (if genuine), Euthydemus , Lysis , Parmenides (to 137), Phaedo , Protagoras , Republic , and Symposium . There is little that is controversial about the chronology of these particular works, and even less that would have much bearing on our picture of Plato's development. Stylistic features that seem best to have separated the late dialogues from the early and middle dialogues, namely the avoidance of hiatus and the development of conscious clausulae-preferences, put all the narrative dialogues in the early-middle group, so that the most trusted tests don't give great separation between them. There is, however, one excellent indicator of the relative lateness within the group of most books of the Republic , namely the frequent appearance of certain response-formulae, both positive and negative ( 1 ), which are otherwise held to be common only in post- Republic works. In fact it is virtually non-controversial that the Republic reached its final form later than other narrative dialogues apart from the Parmenides (of which the majority, from 137c4, suddenly leaves narrative form never to return). I shall be taking the relative lateness of the final Republic for granted in the present study.

I shall be tackling this group in part because I believe (i) that we should in some cases distinguish between the dates of a work's conception ( 2 ), basic execution, and final revision ( 3 ), and (ii) that the narrative apparatus dates regularly (those not invariably) from the final revision of the work, preparatory to its circulating more widely on the Athenian intellectual scene. As long as works were intended for the author to read at private gatherings or within his school there was no need to cast it in the form of a narrative. Narrative form provides guidelines for the presentation of the discussion, and thus could only be important when authorial control is relinguished. I do not believe that all narrative dialogues were written before dramatic dialogues ( 4 ), but rather that narrative dialogues were published before any (except the speech-like Apology ) were allowed to circulate freely in dramatic form ( 5 ). It is a narrative dialogue ( Protagoras ) with which Isocrates ( Helen , 1) seems to have been familiar with at a fairly early date. The true 'dramatic' dialogue seems to have been Plato's own invention ( 6 ), and it is possible to infer from Theaetetus 143b5- c5 that this had not previously been a form familiar to Plato's readership. One may also infer that, at least in the case of the Theaetetus , the dramatic form had been the one in which it had originally been written, and that this form had been retained because of the author's disenchantment with the apparatus of narrative: he did not want the bother of hai metaxu ton logon diegeseis . The validity of the current exercise is not dependent upon acceptance of my theories of composition.

We shall see here that there had been a progression from a time when a rich narrative atmosphere had prevailed ( Protagoras , Euthydemus , Symposium ) to books two to ten of the Republic when the narrative was almost redundant. The use of criteria of this sort will strongly suggest that two works in the corpus are traditionally dated rather too late, while two others are usually thought to have been published rather too early. My order will differ from that of G.R. Ledger ( 7 ), in part because his tests on letters obviously apply, if anything, to the date of the bulk of the writing, not to that of the final revision. For reference his findings as applied to the narrative dialogues would give the following order: Lysis , Charmides , Phaedo , [ Erastae ], Protagoras , Euthydemus , Symposium , Republic , Parmenides . However, p. 214 makes it quite clear that the distances between the last five of these works are very small, and that the least biased way of assessing Ledger's data would have given the order Parmenides , Protagoras , Symposium , Euthydemus , Republic .

The first aspect of these works which will concern us here is the presence of what one might call 'narrative response-formulae'. As an alternative to Plato's usual words for 'yes' a narrative dialogue has a number of single words meaning, in effect, 'he said yes'. The most common of these is synephe, but also found are the imperfects ephe, homologei, synhomologei, synechorei , and the occasional aorist including syneneuse and kateneuse , which seem only marginally more specialized in meaning. Furthermore there are alternatives to weaker positive response formulae, such as edokei and synedokei . As an alternative to 'no' there is ouk ephe . Two works use these devices over twenty times, Protagoras and Euthydemus , while the Erastae has quite a lot for its limited size. The Lysis and Phaedo are much more sparing, and a synephe and an ouk ephe appear in the first book of the Republic . Otherwise the Republic , like the Charmides and Parmenides , has none. In the case of the Parmenides , one might suppose that the fact that the narrative is in indirect discourse precludes these formulae, but the Symposium , in spite of its being very short of question-and- answer dialogue, offers an example of such a formula in the infinitive of indirect speech, homologein (199e5), as well as a case of homologei in the imperfect (201a10). The absence of these forms from Republic books 2 to 10 suggests a decline in their use. The frequency of these 'answers' is best examined by comparing the rate of positive ones with that of nai . They far exceed the rate of nai in Protagoras (50%-19%) and Euthydemus (37%-13%), fall slightly short of it in the Symposium (11%-16%) and Erastae (15%-35%), and occur only sparingly in Lysis (4%-18%), Phaedo (3%-12%), and Republic I (1%-15%). They are absent from Charmides (the surprise), Republic 2-10, and Parmenides . It is also to be noted that ouk ephe replaces a 'no' four times in Protagoras , three times in Euthydemus , and once each in Erastae and Republic I. This suggests that the same factors led Plato to avoid or favour narrative negative responses as narrative positive ones, and also establishes that the single case of synephe in Republic I was not an aberration.

It would be extraordinarily naive to demand that the completion- dates for these dialogues should in all cases correspond with the ratio of narrative to non-narrative responses. Clearly Plato may be struggling to capture a genuine narrative flavour in one work and not in another, and that will affect his choice. However, the authenticity of this narrative flavour may have worried Plato less as time progressed. For what it is worth, when one adds narrative responses to the colourless responses nai and ouk , the percentage of narrative responses in the total are as follows: Euthydemus 72%, Protagoras 69%, Erastae 50%, Symposium 40%, Lysis 17%, Republic I 12%, Phaedo 11%, Charmides & Parmenides & Republic II-X 0%. Scholars would be concerned about the early position of Symposium in the list, and the late placing of Charmides . The few who would see Erastae as genuine would probably prefer it earlier than this too. There is one factor, however, which cannot be ignored. The Protagoras uses Ephe as its regular narrative response, nine times, and Synephe only twice. The Euthydemus uses Synephe instead, again nine times, and never Ephe . Erastae , Lysis , and Charmides also use Synephe and never Ephe . It looks very much as if a deliberate decision has been made to drop the latter in favour of the former. One can easily imagine that in a text with virtually no punctuation it would have been extremely difficult for the outside reader to distinguish this EPHE signalling the interlocutor's agreement from the EPHE which introduces direct speech. Protagoras should therefore be assumed to precede Euthydemus and probably those other narrative dialogues which employ Synephe too.

Let us then see what another test will yield. The most obvious feature of the narrative dialogues is the presence of a regular 'he said' or 'I said' or (occasionally) 'she said'. The are three main verbs in use, giving us in the first person eipon , en d' ego , and ephen , and in the third person eipe , e d' hos , and ephe . From my initial investigation I excluded Parmenides because of its paucity of indicative forms (as opposed mainly to phanai ), examined Symposium in respect of the introductory dialogue and narratives within the speeches for similar reasons, and selected only books I, V, and IX of Republic . Figures were hand-collected, which accepts the likelihood of minor human error in return for the opportunity to be sure that each example is indeed being used conventionally to introduce direct speech. Most dialogues were divided into some two to four random chunks in order to monitor internal consistency, which was in general high. Certain trends became visible in the way Plato selects the alternatives, and in general we may see a hardening preference for en d' ego in the first person, and ephe in the third. There was some prejudice against eipe resulting in its absence from Republic V and IX, and Lysis , and generally low figures for this form, except in Protagoras (4.92%) and Erastae (3.33%). These latter dialogues likewise had the lowest figures for e d' hos (around 7%), which increases markedly in Euthydemus (18%), Phaedo (21%), and Charmides (20%). Only Books II and III of Republic again reach such frequency. Of first person forms it was noticeable that no section of Republic I, V, and IX or Lysis had a rate of ephen as high as 20%, while no section of other dialogues had a rate as low as that. The following order was apparent: Erastae (43%), Protagoras (41%), Euthydemus (38%), Symposium (31%), Phaedo (27%), and Charmides (23%). The rate of eipon seemed to follow a curve, being high in dialogues with extreme tendencies elsewhere, Erastae (38%), Symposium (23%), Protagoras (15%), lower in Euthydemus (11%), very low in Phaedo and Charmides (4%), lowish again in Republic I (9%) and Lysis (13%), climbing steeply to Republic V (20%) and IX (29%). A check on other books of the Republic showed books II-IV to be close to the Lysis figure, while books VI-VIII were not far from that for V. Book X stood out here, its 4% unexpectedly matching the Charmides .

These results gave one reason to believe that chronological factors were partly responsible for the various observable trends. I shall present elsewhere charts stemming from a more sophisticated computer-analysis of the results, but at this stage, but I can include the results of some basic distance-measurements obtained by averaging the distances between works for each of the 2 x 3 variables. These are based on numbers of the forms concerned counted from L. Brandwood, A Word Index to Plato (Leeds, 1976). Most significant for sorting out the apparently earlier dialogues are the distances from the Republic in toto: Erastae 21%, Protagoras 14%, Symposium 11%, Euthydemus 11%, Phaedo 8%, Charmides 7%. At this stage come books VIII and IX of Republic 6% and book III 5%, before Lysis 4% only fractionally higher than books II and V 4%, Republic I and VII 3%, book VI 2% and books IV and VIII less than 1%. Some dialogues can be grouped together quite closely on the basis of low distance-measurements between them. On my original hand-collected figures the most obvious affinity, other than that between books of the Republic , is between Phaedo and Charmides (2%); Republic I and Lysis (4%); Protagoras and Symposium (4%); Republic V and Lysis (5%); Republic V and Charmides (6%); Protagoras and Euthydemus (6%); and Euthydemus and Symposium (7%). If we base ourselves on our Brandwood-count, the numbers change somewhat, but the picture does not. I give in descending order of proximity the works that come closest to each work investigated:

1. Erastae: Prt., Smp., Eud. 2. Protagoras: Smp., Eud., Erast. 3. Symposium: Prt., Eud., R V 4. Euthydemus: Smp., Prt., Chrm. 5. Charmides: Phd., R II, R I 6. Phaedo: Chrm., R III, R II 7. Lysis: R X, R I, R VI 8. Republic I: R IV, R VIII, R X 9. Republic II: R III, R IV, R VI 10. Republic III: R II, R VI, Lys. 11. Republic IV: R VIII, R VI, R I 12. Republic V: R VII, R VIII, R IV 13. Republic VI: R VIII, R IV, R VII 14. Republic VII: R V, R VIII, R IV 15. Republic VIII: R IV, R VI, R VII 16. Republic IX: R VIII, R IV, R VI 17. Republic X: R I, Lys, R VII

Numbers 2-4 clearly constitute a group, with or without 1 depending on its authenticity. Nos. 5 and 6 are a group, and much closer to the remaining works than Nos. 2-4. Lysis clearly belongs somewhere in the Republic group, and differences from Republic X, I, VI, IV, III, and VIII all fall between 3.98% and 4.22%. Within this group books IV, VII, VIII, and to a lesser extent VI are so typical that even less typical works seem to be close to them. Much the biggest differences are between IX and X (10.7), IX and Lys. (9.6), and IX and III (9.5). Since book IX is so atypical there is a strong possibility that it was in fact last to be completed, but that book X, though an appendix to what precedes, was completed early, prior to the final polishing of II-IX. That would fit the hypothesis that Book X was written with narrative apparatus complete, while the other books with the probable exception of Book I, where 'narrative response formulae' still occur, were written without narrative apparatus.

It is my belief than that narrative machinery can be used with caution as a chronological indicator, that works 1-4 were probably completed in that order, that works 5 and 6 were completed at roughly the same time, and that thereafter the order of completion R X, Lys., RI, RII to R IX seems not improbable. The exact position of Lysis is negotiable, but its completion belongs to the period of the completion of the earliest books of the Republic . I suspect, as with Book I, that it is a substantial revision of an earlier work perhaps already composed as a narrative. In defence of such an order I offer the ascending order of percentage of eipon in these works: R X, R I, R III, Lys., R II, R IV, R VIII, R VI, R VII, R V, R IX; or the descending order of percentage of en d' ego : R X, Lys., R III, R I, R VI, R II, R IV, R VII, R VIII, R V, R IX.

Other brief response formulae might also tend to change somewhat in preparation for publication. Distance measurements calculated on the basis of 26 variables (25 common ways of saying little more than 'yes', plus a miscellaneous group) are not uninteresting. I excluded all 'narrative response formulae' as I wanted a criterion independent of these, and for other reasons books IV and VI were divided into two sections. On this test there is an amazing closeness between Lysis and Charmides , 1.12, closer than between the two closest parts of the Republic (III and IVB, 1.37). Both Lysis and Charmides are very close to Rep. I (1.47, 1.54), and fairly close to Phaedo (2.02, 2.10), though not as close to it as Rep. I is (1.57). Rep. II is also closer to the Phaedo (2.13) than anything but Rep. X (1.92), whose affinities are with Republic books II, V, III, VII, and IVB (1.92-2.49) rather than with VIII and IX (2.80, 2.82). An order Charm. , Lysis , Rep. I, Phaedo , Rep. II, Rep. X, Rep. III through to IX, suggests itself on this test . Such an order is only remarkable for the position of Rep. X, and that is well explained by our earlier hypothesis: that X was originally written as the conclusion of a work about to be published, while other books were only revised once this conclusion had been written. Perhaps Charmides and Lysis still seem rather late for those who are trying to date the bulk of the philosophic material, but then we have been dealing with a factor stemming partly from the completion-date, partly from the date of the bulk of the writing. There is in fact no difficulty for those who would wish to see the bulk of the Lysis as the earliest material in the corpus. ( 8 )

Much greater distances from the Republic are to encountered in Erastae , Prot. , Parm. (narrative portion), Euthd. and Symp. Of these the first three are relatively close to one another, with Prot. and Parm. both being closest to Erastae (2.68, 2.58) and the latter being closer to them than to anything else. Parm. , however, is closer to Lysis than to Prot. The position of Parm. will be something of a shock both on this test and on a lot of other plausible criteria. For instance even the narrative part (to 137c) displays a higher hiatus-rate than any other Platonic work, Lysis coming next. The answer, I suggest, is that Parm. , though commenced in narrative form, has never been polished for publication in the way that other works have, and hence it displays some features belonging to the completion- dates of Prot. and Lysis . The Erastae , with its largely unnamed cast, may well be another sketch that was abandoned before the final stage. Even more than Erastae , the introduction to Parm. favours eipon for 'I said', likewise preferring eipe to e d' hos as a variant to ephe for 'he said'. ( 9 ) As stated Erastae , Prot. , and Parm. are remoter from the Republic than the Phaedo group.

The position of Euthd. and Symp. are more difficult. Both involve very high distances from Rep. III-IX. The latter involves only a small sample which compares best with Euthd. , Charm. , Phd. , and Lysis (3.14-3.20), while the former is again closest to Charm. , Phd. , and Lysis (2.37-2.67). In both cases I suspect no great interval between the bulk of the writing and final publication. That would explain why an examination of narrative apparatus tends to put them early.

I thus conclude that nothing in my examination of positive response formulae precludes completion-dates suggested above, and that the following sequence would not be irrational: Prot. , [ Erastae , Parm . pt. 1 sketched], Euthd. , Symp. , Charm. , Phd. , Rep. X, Lys. , Rep. I , Rep. II-IX in sequence. That is where narrative dialogue ends for Plato. The onus of providing a constant 'I said' and 'he said' throughout this long work provoked the final abandonment of narrative form, and prompted the great Platonic experiment of publishing in 'dramatic' dialogue-form.


(1) ti men , orthos/orthotata , pantapasi men oun , alethestata (legeis) , pos gar an , adynaton , ou gar oun .

(2) Some works got no further than the 'conception'-stage, e.g. the Philosopher and Hermogenes which are foreshadowed in the Sophist-Politicus and Timaeus , respectively, and the Critias according to the majority view breaks off in mid-execution.

(3) It is presumably this final revision that Philip of Opus is supposed to have given to the Laws . Other works not actually published may have lacked such a stage, and I am not arguing that the final revision is in all cases far separated from the execution.

(4) For this thesis see H. Thesleff, Studies in Platonic Chronology, Commentationes Humanarum Litterarum (Helsinki, 1982).

(5) For the connexion between publication and narrative see Thesleff, Studies in Platonic Chronology , p. 63.

(6) The evidence is a papyrus presented by M. Haslam, BICS 19 (1972), pp. 17-38; P.Oxy. 45 (1977), 29-39. See further my Thrasyllan Platonism (Ithaca, 1993).

(7) Recounting Plato (Oxford, 1989), p. 224.

(8) Ledger, Recounting Plato , pp. 218-223.

(9) Prot. is comparable (with Erastae ) re the 3rd person, Symp. re the 1st person.

Harold Tarrant,
e-mail: CLHAST@cc.newcastle.edu.au

COPYRIGHT NOTE: Copyright remains with authors, but due 
reference should be made to this journal if any part of the above is 
later published elsewhere.

Electronic Antiquity Vol. 1 Issue 8 - April 1994
edited by Peter Toohey and Ian Worthington
ISSN 1320-3606