I INTRODUCTION
If we had no idea which parts of Greek literature in a certain period were poetry or prose, we would regard it as our first job to find out. How much of the Greek prose of the Imperial period is rhythmic has excited less attention; and yet the question should greatly affect both our reading of specific texts and our understanding of the whole literary scene. By ‘rhythmic’ prose, this article means only prose that follows the Hellenistic system of rhythm started, it is said, by Hegesias, and adopted by Cicero and by many Latin writers of the Imperial period. Estimates of how much Greek Imperial prose is rhythmic have long varied drastically. Some experts suggest that all or much artistic Greek prose in the period is rhythmic, others that what little there is fades out after the first century a.d., as part of the victory of Atticism. There has been fairly little substantial work on rhythmic prose in the first three centuries a.d. for over fifty years (more on accentual prose from the fourth). The object of this article is to investigate a large part of one author's work thoroughly, and to establish that that part is rhythmic. It will also aim to show how that conclusion should greatly affect our whole conception of the author as a writer, and our reading of his every sentence.Footnote 1
Appian may seem an unexpected choice: his style and his manner of writing have not on the whole been greatly appreciated, and the question whether he is rhythmic has never to my knowledge been considered. Recent scholarship has realized how much Appian designs and thinks about his work; the realization needs to be extended to the specifics of his writing, as would seem obvious if we were considering Sallust, Tacitus or Thucydides (Appian's foremost model). The study should in principle be of significance for Appian, and for the period more widely; the detailed discussion will offer some means of establishing what other Imperial prose is rhythmic, and give us pointers to reading that rhythmic prose with stylistic sensibility.Footnote 2
The study of Greek prose-rhythm has been diverted by consideration of rhythm in the times before Hegesias, a subject intrinsically much more obscure. There are more hopes of success in pondering whether or not individual Imperial authors adhere to a system that is essentially known than there are in working out a rhythmic system for each Classical author. The endings that will be taken as rhythmic in this article are the standard ̄ ̆ ̄ ̄ ̆ ̄ / ̄ ̄ ̄ ̄ ̆ ̄ , ̄ ̆ ̄ ̄, ̄ ̆ ̄ ̄ ̄, ̄ ̆ ̄ ̆ ̄. Any long can be resolved except for the last, which can be replaced by a breuis in longo; the last syllable, whether or not thus lengthened by quasi-metrical pause, is simply marked as long in the examples below. Two other endings (their resolved forms included) are not here classed as rhythmic, that is, as part of this rhythmic system: ̄ ̆ ̄ ̄ ̄ ̆ ̄ and ̄ ̄ ̄ ̄. The former is not supported as rhythmic by Latin practice with atque + consonant or by Latin theorizing. The matter is not of crucial importance, since this ending comes much less often than the second. It does, though, form a higher proportion of apparently unrhythmic endings in Appian than in pre-Hellenistic prose, so that regarding it as unrhythmic will make it harder to show that Appian was rhythmic (since there will be a smaller difference between Appian and pre-Hellenistic prose in the number of apparently rhythmic endings). For that very reason, it would be prudent for now to regard it as unrhythmic. The second rhythm, ̄ ̄ ̄ ̄, is in shape completely unlike the single-short sequences here regarded as rhythmic; the resolved forms it would bring with it—and it would be still more anomalous to admit the ending but exclude its resolutions—mark out the double-short character that it would possess. (Note also the worries at Cic. Orat. 224, Quint. Inst. 9.4.97 [in Russell's text], 9.4.101.) The sequence accounts, with its resolutions, for most unrhythmic endings. But if we did regard it as rhythmic, that would not create a problem for the argument to come. The argument could simply be altered from presenting unrhythmic prose as higher in its proportion of unrhythmic endings than rhythmic prose: it would instead present it as higher in its proportion of a particular ending which was on this view rhythmic in rhythmic prose but was less used there than in unrhythmic prose.Footnote 3
In Latin prose, the practised reader can usually decide from a few pages whether an author is rhythmic; it appears to be much harder in Greek prose. The system was devised to suit Greek. In unrhythmic Latin ̄ ̄ ̄ ̄ occurs still more frequently than in Greek, thanks to various linguistic changes in Latin and Proto-Italic; and the standard Latin pattern of a final verb makes all the more conspicuous the efforts required of Romans to sustain rhythmic writing. With Greek prose, it is easy to be unsure; I have myself changed my mind more than once on whether Appian is rhythmic. These difficulties prompt us to see whether quantification will help. Some past use of figures in this area makes one nervous of employing statistics; but we have a simple question, and can produce firmly differentiated masses and apply straightforward methods.Footnote 4
Particular problems of prosody need not be gone into here. Clashing vowels we will negotiate later in our voyage; for the present, we will leave on one side those places where only the supposition of hiatus would produce a rhythmic ending. (Where a rhythmic ending is produced only by elision, or crasis, of kinds not routine in prose, the passage is counted as non-rhythmic.) It does not especially matter how mute and liquid are treated, provided that we approach them in the same way across the texts that we consider. It is a tenable position that Attic in the Imperial period was valued as a spoken, not just as a written, dialect, and that it is proper to apply the rules of shortening that we see in Attic speech of the fifth and fourth centuries b.c.; at any rate, they will be applied here. A few of the non-rhythmic Classical texts we will consider are admittedly in dialects other than Attic; but we are interested only in comparing with Appian the data a non-rhythmical text would produce, not in the rhythm these texts would have, if they had one, in their own dialect. Finally, there is the matter of Latin names. For some names, scansion in Greek is obvious; but when it is not, closes that include them will be left aside.Footnote 5
II SENTENCES
For the purposes of this article, I have scanned the whole of the Bella Ciuilia, a subset of Appian's larger creation. This is the most extended continuous part of his work. It is unlikely that one book of it would be rhythmic, one not; it is unlikely too that these books would be rhythmic and the rest of his Roman History not. I have read all the rest of his work with rhythm in mind, and have also taken random samples for numerical comparison; on this basis, it seems clear that what is true of the Bella Ciuilia is true of the whole Roman History. The Civil Wars occupy 645 Teubner pages; over 20,000 endings to phrases or sentences have been examined.
It will eventually become apparent that rhythm in Appian informs the whole sentence, not just the end of it; but, since it is more difficult to determine where an internal phrase ends, the discussion will begin by looking at the ends of sentences. It will also at some points bring in half-sentences, that is, endings followed by a high point; there seems to be no difference as regards Appian's practice, but for the sake of perspicuity attention will chiefly be concentrated on endings followed by a modern full stop. It will be borne in mind that no rhythmic prose is rhythmic throughout; to give a first rough example, the 108 sentences of Cicero's certainly rhythmic Pro Rabirio Perduellionis Reo, fragments excluded, give 95 rhythmic, 13 not (88.0% and 12.0%); 100 sentences of his certainly rhythmic Academica Priora (from 9 quibus de rebus to 32 relinquamus), in Plasberg's text, offer 93 rhythmic (with iudici at 19) and 7 non-rhythmic. If we took paragraph endings, a point where we might a priori especially expect a rhythmic close, in Thucydides Book 8 (OCT or Alberti) we find, documents and the very end aside, 41 closes that would be rhythmic if Thucydides had followed the Hellenistic system, 24 that would be non-rhythmic (63.1% and 36.9% respectively of the total rhythmic + non-rhythmic). (One other close would need hiatus to be rhythmic and is left out of account; three of the non-rhythmic endings would become rhythmic with elision, in one case with elision of ταῦτα, elided five times, it seems, in Thucydides’ MSS.)
In Book 1 of Appian, paragraph endings in Mendelssohn-Viereck's edition are 70 rhythmic, 11 non-rhythmic (86.4% and 13.6% of rhythmic + non-rhythmic). (Four further paragraph endings would be rhythmic but would require hiatus and are omitted; one non-rhythmic paragraph ending [1.31] would be rhythmic if we elided ἀντί, as apparently 14 times in Appian's MSS, one [393] if we elided ταῦτα, as apparently 63 times in Appian's MSS.) In Book 1, if we exclude endings which require hiatus to be rhythmic, 606 sentences end rhythmically, 81 not, that is 88.2% rhythmic, 11.8% non-rhythmic; the sentences of Book 2 give 694 rhythmic, 83 non-rhythmic, that is 89.3% and 10.7%. In Book 3, the sentences are 464 rhythmic, 63 non-rhythmic, that is 88.0% and 12.0%. (One of the non-rhythmic endings in Book 1, four in Book 2 and one in Book 3 would become rhythmic with hiatus but the hiatus would need to replace a manuscript elision; they are left as non-rhythmic, not excluded.) We could also look in Book 3 at colons as well as full points; together they produce 631 rhythmic endings, 79 non-rhythmic, that is 88.9% and 11.1%. The same two exercises for Books 4 and 5 yield: Book 4: rhythmic sentence-endings 575, non-rhythmic 99 (85.3% and 14.7%); Book 5: rhythmic sentence-endings 627, non-rhythmic 88 (87.7% and 12.3%); Book 4: endings at colons and full points rhythmic 786, non-rhythmic 136 (85.2%, 14.8%); Book 5: endings at colons and full points 816 rhythmic, 120 non-rhythmic (87.2% and 12.8%).Footnote 6
We should try to form some kind of comparison between these figures and the figures for an author unlikely to use the Hellenistic system of rhythm. We may first take the sentence-endings in Book 5 and compare them with the sentence-endings in Lysias speeches 1-7 and 9-10; these are 429 rhythmic, 289 non-rhythmic (59.7% and 40.3%; the same exclusion for hiatus as with Appian). To get the total exactly the same as in Appian Book 5, we may leave off the last three sentences of Lysias and so get 427 and 288 (same percentages). We could now employ a chi-squared test to see how likely it is that the difference between the figures is due to chance. It would be prudent to use the method that will produce a higher probability of chance: that is, we are not yet supposing we can predict an average result, but are merely comparing two sets of data. The procedure is to calculate the sum of the two values of
$${\left( {n1 - n2} \right)^2}/\left( {n1 + n2} \right)$$
where n1 and n2 are the numbers observed in a given category in the two sets of data; in our case the categories will be first rhythmic endings, and then unrhythmic endings, n1 in Appian and n2 in Lysias. We square the difference between 627 and 427 (rhythmic endings) and divide it by the sum of 627 and 427, and so with the difference between 88 and 288 (unrhythmic endings); we then add the two figures together. So χ2 = (627−427)2 / (627 + 427) + (88−288)2 / (88 + 288) = 144.33.
There is only one degree of freedom (our four figures in a table would give two rows and two columns; if we subtract one from each two and multiply, we obtain one). With χ2 = 144.33 we would pass far beyond the usual tables for probability, and arrive at a probability of purely random difference so small—with such a string of noughts after the decimal point—that it can be regarded as equivalent to zero. The same would apply if we took from Appian a sample with a lower proportion of rhythmic endings; so the sentences of Book 4 (575 rhythmic, 99 non-rhythmic; 85.3% and 14.7%) may be set against the first 674 sentences from the sample in Lysias (400 rhythmic, 274 non-rhythmic; 59.3% and 40.7%). χ2 = 113.51. For the sake of comparison, if χ2 were 10.83, the probability would be 0.001 (1 in 1,000), if 15.14, 0.0001 (1 in 10,000).Footnote 7
This seems encouraging. The problem with our proceedings so far is that the sentences form a single stretch of text or a set of texts (some of Lysias’ speeches); they are either consecutive (Appian) or usually so (Lysias). Yet the assumption of such a test is that the instances are independent; the close cohesion of the sentences conflicts with this assumption. The result of the calculation is still so remarkable that it looks likely that the two sets of data are different kinds of thing; but we must try some further tests to give our impression more solidity. First we may try a smaller sample of randomly chosen sentences within the two bodies of data used in our original test. The smallness of the sample already means that the result will be less striking. One hundred sentences were chosen randomly from the same portion of Lysias and from Appian, Bella Ciuilia 5; any that were unusable for the familiar reasons (hiatus and Latin names) were replaced by other randomly chosen sentences. The result was: Appian: rhythmic 82, non-rhythmic 18; Lysias: rhythmic 56, non-rhythmic 44. Here the same χ2 test produced 15.80; the probability of chance is, as we have seen, <0.0001 (less than 1 in 10,000). In a second and larger supplementary test, twenty sentences were chosen at random from twenty authors unlikely to use the Hellenistic system of rhythm; these were compared with 400 sentences chosen at random from the whole Bella Ciuilia. (We are not yet in a position to provide a matching twenty authors of rhythmic prose, since we are only in the course of showing that even one particular Imperial Greek author wrote rhythmically.) Again in both cases unusable sentences were replaced with usable ones randomly selected. The 400 sentences of Appian gave 328 rhythmic closes, 72 non-rhythmic (82% and 18%). The twenty authors produced 249 rhythmic closes, 151 non-rhythmic closes (62.25% and 37.75%). (328−249)2 / (328 + 249) + (72−151)2 / (72 + 151) gives χ2 as 38.80283825. This figure again is too high for the tables, and results in a probability of chance that may be regarded as in practice equivalent to zero. We already have, then, a decisive proof.Footnote 8
These were the specific numbers for rhythmic sentences, which show a range limited to between 9 and 14 out of 20: Heraclitus (already fragmented; of course non-consecutive sentences only) 14, Herodotus 9, Gorgias 13, Antiphon 13, Andocides (1–3) 12, Thucydides 12, Hippocratic corpus (vols. 1-2 Kühlewein) 13, Lysias 14, Plato (first three volumes of OCT) 14, Xenophon 13, Isaeus 11, Aeneas Tacticus 14, Demosthenes 11, Aeschines 14, Apollodorus ([Demosthenes] 46, 49, 50, 52, 53 and 59) 9, Hyperides 13, Dinarchus 13, Lycurgus 13, Aristotle 11, Theophrastus (Historia Plantarum and De Causis Plantarum) 13.Footnote 9
III HIATUS
Two further strong arguments may be brought in, partly to add still more assurance, but also to show the practical effect of the rhythmic system, and to take us deeper into the nature of Appian's rhythmic prose. First, hiatus.
Cicero is well known to avoid, almost entirely, the use of atque followed by a consonant except in rhythmic closes, though these sometimes extend further back into the sentence than in the standard forms of them adopted here. This clear connection between an unusual feature and rhythm helps to confirm that Cicero is writing rhythmically. Many Greek authors of the Imperial period are well known to avoid hiatus, though not entirely. Suppose that Appian employs hiatus, save for the commonest words, preponderantly at closes that would be rhythmic; that would show a strong connection between the two phenomena, and help to confirm that rhythm is real, that he is writing rhythmically. It would also provide anticipatory confirmation of a point we will come to presently, that rhythm is not confined to the end of sentences or half-sentences.Footnote 10
We would do best to exclude from consideration (a) places where there is a significant possibility that the first of the two vowels could be elided (or the second prodelided, or both merged in crasis): these are not adequately clear cases of hiatus. We would also do best to exclude (b) very common words, notably καί and ὁ (any part), where hiatus seems not greatly to concern authors of the period: here the presence of hiatus is not striking, and so does not provide a salient departure from a usual avoidance. The two classes (a) and (b) come together, since MSS and papyri seem to indicate that in Greek prose it is usually common words, not words in general, which are liable to elision. It would be perilous to use the authority of MSS and papyri to assert that there is elision at any particular point; but the general diffusion of this approach in them is a strong reason for caution in supposing hiatus with such words. After all, literary papyri on the whole come from Egypt, and the greatest number from the second century a.d.: precisely the milieu of Appian. The presence of Appian himself on papyri is exiguous; but we may take an example from a recently published second-century papyrus of Plutarch, close to the author's lifetime. It displays αλλωσπερ at Plut. De prof. virt. 75C (P.Oxy. 5153 col. ii 4); the MSS, according to Gärtner's edition of 1993, have ἀλλ᾿ ὥσπερ. ἀλλ᾿ is far commoner than ἀλλά followed by vowel as the two forms are reported for the MSS of Appian (290 to 47; for the Civil Wars, B is Mendelssohn-Viereck's principal guide). There is at the least a considerable possibility that Appian most often wished the word to be elided. With other common words, the MSS of Appian sometimes elide, sometimes not: thus 2480 instances of δ᾿, as against 1688 of δέ + vowel. In general, elision must be considered a serious possibility for such words, whether universally or sporadically.Footnote 11
Many of the words where elision, crasis, etc., occur often or from time to time in the MSS of Appian and other authors: δέ, τε (much more rarely elided in Appian than δέ), ἀλλά, οὐδέ, μηδέ, μήτε, οὔτε, εἴτε, εἶτα, παρά, ἀντί, ἀπό, ἐπί, κατά, ὑπό, διά, μετά, ἵνα, τάχα, κἀκεῖνος (κἀκεῖθεν, κἀκεῖ, etc. also found), κἀμέ, κἀν, κἄν, οὗτος, τοιοῦτος, ὅδε, πᾶς, ἄλλος, ἕτερος, τις, ποτε, μάλιστα; (only in MSS of other prose authors) ἄρα (but ἆρ᾿ once in Appian), ἔνθα, ἔτι, ὅσος, τοιόσδε, πρῶτος, τότε, ὅτε. Frequency of hiatus in Appian and elsewhere suggests it would also be reasonable to exclude hiatus after the following: ἤ, πάνυ, περί, possessive adjectives (τοῦ ἐμοῦ οἰκέτου, etc.); and before οὖν, parts of εἶναι, and unemphatic personal pronouns (including αὐτόν). Proper names too are sometimes treated with licence; numerals are often in hiatus, especially the lower ones. We should further omit places where the hiatus is accompanied by punctuation, since pause in the sentence would sufficiently justify it; the only exceptions are where the commas of editors seem purely syntactic and pause not inevitable.Footnote 12
It may be most enlightening to take a limited portion of text, and consider the occurrences of hiatus in some detail. (Italics are used for the hiatus which concerns us in each example.) These are the majority of instances from Book 5.1-200 before colon or full point where elision is unlikely and hiatus—in most cases only hiatus—would produce a rhythmic ending: 1 (first sentence) | καὶ εὐθὺς ὀφθεῖ̄σα ̆ ἐ̆κρά̆τεῑ. | (cf. for example Mac. fr. 9.3 μαθοῦ̄σα ̆ ἐ̆πε̆κύ̄ρω̄σε̄ν, | Syr. 362 ἐπιτοξεύσᾱσα ̆ ἔ̄κτεῑνε̄ν, |), 39 | ὥς μοι περὶ Ϲύρων λέγο̄ντι ̆ εἴ̄ρη̄ταῑ· | (very similar phrases, again rhythmic, 2.61, Ill. 11), 99 ὡς δικαιό̆τε̆ρα ̆ αἱ̄ρού̄με̆νο̄ν. |, 123 | καὶ τάδε μὲν ὡς προσεδόκη̄σε̆ν ὁ̆ Ἀ̆γρί̄ππᾱς | ἐγί̄γνε̆το ̆ ἅ̆πᾱντᾱ· | (cf. 2.189 ἐγί̄γνε̆το ̆ ὅ̆μοῑᾱ. |; commas after μέν and Ἀγρίππας in Mendelssohn-Viereck), 151 καὶ ἐς τὸ τεῖχος ἐξήλᾱντο ̆ ὀ̆λί̆γοῑ, | καὶ αὐτοῖς εἵπο̄ντο̆ ἕ̆τε̆ροῑ·, 154 | ἐκάλει τῆι σάλπῑγγι ̆ ἀ̆νᾰχω̄ρεῖ̄ν. | (cf. for example 2.366 μείζο̆νι ̆ ἐ̆γί̄γνο̄ντο̄. |, 412 ἠρεμοῦ̄ντι ̆ ἀ̆πέ̆κλεῑσε̄ν· |, 3.28 ἐν θαύ̄μᾰτι ̆ ἔ̆χο̄ντε̄ς, |), 162 | καὶ τὴν διαβολὴ̄ν ἐ̆γὼ̄ τή̄νδε̄ [note word-order] | ἠγνόη̄σα ̆ ἐ̆πὶ̆ πλεῖ̄στο̄ν. | (cf. 1.300 | καὶ ἔκεῑτο ̆ ἐ̆πὶ̆ πλεῖ̄στο̄ν, |; ἐπὶ πλεῖστον is often the last part of a rhythmic ending without hiatus, cf. for example [with full stop or colon to follow] 1.9, 86, Sam. fr. 4.18, Pun. 618), 165 | οὓς τῆς ἐμῆς προτίθη̄μι ̆ εὐ̄κλεί̄ᾱς. | (note word-order), 168 | τούτοις ὅμοῑᾰ ἔ̆λε̆γο̄ν. | (cf. 63 | ψεύδεσθαι διδαχθέ̄ντα ̆ ἔ̆λε̆γο̄ν |; comma after ὅμοια in Mendelssohn-Viereck).
Within the sentence or part-sentence, elision is unlikely and hiatus would produce a rhythmic ending in instances like the following: 5.3 | ἕτερα ἐμφύ̄λῐᾰ ὅ̆μοῑᾱ, | στρατηγοῦ μέν κτλ. (cf. Hisp. 70 ἐσεμνύ̄νε̆το ̆ ὅ̆μοῑᾱ· |), 6 ὁπλίταις τε ἐξ ὧν ἐδύνᾱντο ̆ ἀ̆νε̆πλή̄ροῡν | καὶ ἐρέταις κτλ., 25 ὅτε ἐν Μουτίνηι συνηλλάσσο̄ντο ̆ ἀ̄λλή̄λοῑς | (cf. 81 συνέθε̄ντο ̆ ἀ̄λλή̄λοῑς | [same phrase at Pun. 136, followed by comma], 2.415 διεχρήσᾱντο ̆ ἀ̄λλή̄λοῡς· |, 4.108 ἐπειρῶ̄ντο ̆ ἀ̄λλή̄λω̄ν. | ; nine instances in Appian of an ending in -ντο followed by ἀλλήλ- and a comma), or the parallel participial phrases at 75 ὅθεν αὐτὸν ὁ στρατὸς ὁ Ἀντωνίου κατεμέμφετο καὶ̄ ὁ̆ Καῖ̄σᾱρ | ὡς ἀντιπράσσο̄ντᾰ Ἀ̄ντω̄νί̆ω̄ι, | καὶ Φουλβία ὡς πολεμοποιοῦ̄ντα ̆ ἐ̆ν ἀ̆καί̄ρω̄ι, | κτλ. (cf. for example 3.58 | τὸν δό̄ντα ̆ ἀ̆νε̆λό̄ντε̄ς. |, Pun. 72 τὸν δό̄ντα ̆ ἤ̄λε̄γξε̄. |, Syr. 198 | ἀναφέρο̄ντα ̆ ἐ̄ς Ῥώ̄μη̄ν· |, Mith. 201 | καὶ τὸν ὑπαντήσᾱντα ̆ ἔ̄κτεῑνε̄. |).
It already looks as if there is a relation between hiatus and rhythm. Now we need to take a more general look at hiatus within the portion of text 5.1-200. The broad picture, which applies to the Civil Wars as a whole, is that general rhythmical principle allows hiatus, like breuis in longo, after a rhythmic close, so at the end of a phrase and in ‘metrical’ pause; in these cases the first element in the hiatus may be a short vowel or a long vowel or a diphthong. Hiatus is also permitted where a phrase ends unrhythmically, as a sentence can; the same point about the first element applies. Within a phrase, if the common words above are excluded, hiatus is at the least rare except as part of a rhythmic close; in hiatus within a rhythmic close, the first element is normally a short vowel. This picture suggests a special connection between hiatus and rhythm. It is arguable that there are no actual exceptions in this portion; if there are a few exceptions, the preponderance that remains suffices to establish a significant pattern, and thus the presence of the rhythmic system in the text. It would be reasonable to maintain that in two apparent exceptions we are again dealing with very common words or word-groups after which hiatus matters less. At 60 εἰ μὴ τὰ πρότερα ἐπινί̄κῐᾰ λά̆βοῑε̆ν. |, where the hiatus would not be part of a rhythmic close, πρότερος can be thought similar to πρῶτος above. So cf. for example Cass. Dio 39.17.1 and six more times τῶι (…) προτέρωι ἔτει, and Joseph. AJ 18.344 and Plut. Cic. 41.7 τοῦ προτέρου ἀνδρός. At 88 τῶι ὄντι is followed by hiatus, as at 2.112 and in 14 out of 64 occurrences of the phrase in Dio of Prusa (excluding from the 64 two places where the phrase is followed by punctuation and hiatus). The remaining exception is 19 πολὺ πλείονα αἰτούντων. It would be easy to replace πλείονα with πλέονα: the form is very common in Appian (cf. for example 5.51, and note the similarly tempting close of a sentence in the MS at Pun. 490 πολὺ πλείονας ὄντας., and in the MSS at Sic. fr. 2.2 ἐπὶ πλείοσι συνθέσθαι., cf. fr. 2.3 ἐπὶ πλέοσῐ σῠνέ̆θε̄ντο̄. |).Footnote 13
We have not yet discussed the nature of phrases; ‘phrase’ is only a shorthand, since a sizable single word can form a separate entity for these purposes, and a brief beginning to a sentence or part of it can be followed by a break and the main start. In all the instances of hiatus from B Civ. 5.1-200 which could be inside phrases but not as part of a rhythmic close, actually the word or group of words, or those following, or both, are marked out by emphasis in the meaning of the sentence, and thus detachment and the end of a phrase are apt. Here, then, the phrases close unrhythmically. Particles, word-order, or a long first element in the hiatus provide further support for the articulation: 6 | καὶ αὐτὸ ὁπλίταις τε ἐξ ὧν ἐδύνᾱντο̆ ἀ̆νε̆πλή̄ροῡν | καὶ ἐρέταις ἐκ θεραπόντων ἢ δεσμωτῶν, ; 38 | καὶ τὰ ἀναγκαῖα ἐ ς τὸ πέραν τοῦ ποταμοῦ μετενεγκάντων τε καὶ ἐπὶ τῆς ὄχθης, εἴ τῐς ἐ̆πῐχεῑροί̄η̄, | σκευᾰσᾰμέ̆νω̄ν τό̄ξοῑς, |; 73 καὶ ἡ δόξα ἐς τοὺς στρατιώτᾱς ἐ̆φέ̆ρε̆το̄. | (cf. previous sentence | ὃ δὲ καὶ̄ γέ̆νοῑτο̄ | ἐς τοὺς στρατοὺ̄ς ἐ̆δᾰπᾰνᾶ̄το̄. |); 74 | προκρούσματα ἐς τὸν Καί̄σᾰρᾰ ἐ̆γί̄γνε̆το̄ | καὶ δῐᾰφο̆ραὶ̄ μεί̄ζοῡς· |; 92 | χω̄ρί̆ο̄ν τε̄ | ὥρῑστο̆ τῆ̄ι δί̆κη̄ι | Γάβιοι … καὶ̄ σῠνέ̆δρῐο̄ν | τοῖς κρί̄νοῡσῐν ἐ̆γί̄νε̆το̄ | καὶ βήματα ἐ̄ν μέ̆σω̄ι δύ̆ο̄ | τοῖ̄ς ἐ̆ροῦ̄σῐν |̮ ὡ̄ς ἐ̄ν δί̆κη̄ι. |; 108 | καὶ̄ μά̆λῑστᾱ | τοῦδε τοῦ ἐμφυλίου οὐκ ἐκ Μακεδονίας ὑμῖν ἢ Θράικης κτλ.; 175 | ὡς δὲ καὶ σπεύδοντα εἶ̄δε̆ τὸ̄ν Λεύ̄κῐο̄ν |̮ πᾰρε̄λθεῖ̄ν | ἐς τὸ Καίσαρος χαράκωμα, ; 177 | καὶ τάδε λέγω οὐ̄ πᾰραῑτού̄με̆νο̄ς | παθεῖ̄ν ὅ̆ τῐ θέ̆λοῑς |; 197 οὕτως ἀεί μοι προσενήνεχθε ὡ ς κτλ.Footnote 14
Similar principles apply to the demarcation of phrases that are rhythmic, when there is no punctuation; we are thus not trapped in a circular dependence on rhythm alone. Here are some examples, either for hiatus which forms part of a rhythmic close or for hiatus the first element of which occurs at the end of a rhythmic close. It will be noted that only in the latter type can the first element be a long vowel or a diphthong (cf. for example 3.284 ἐς τὸ χαράχω̄μα ̆ αὐ̄το ὶ̄ | ὑ̆πὸ̆ ἀ̆δο̄ξί̆ᾱς, | ἀλλὰ παρ᾿ αὐ̄τὸ̆ ἔ̄στη̄σᾱν· |). 21 οὐχ ὡ̄ς ἐ̆βού̄λε̄σθε ̄ | ἀ̄ λλ᾿ ὡ̄ς ἦ̄ν ἄ̄ξῐο̄ν, |; 36 ὡ̄ς βᾰσῐλί̆δα ̄ | ἀχθῆναῑ μὲ̆ν ἐ̆κέ̆λεῡσε̄ν, |; 37 | ὅτι Ῥωμαίων καὶ Πᾱρθῠαί̄ω̄ν | ὄ̄ντε̆ς ἐ̆φό̆ρῐοῑ | ἐς ἑκατέρους ἐπιδε̄ξί̆ω̄ς εἶ̄χο̄ν |; 65 | ἐλοιδόρουν τε καὶ̄ ἔ̆βᾱλλο̄ν | καὶ φεύγο̄ντα ̆ ἐ̆δί̆ω̄κο̄ν | ἔς τε κτλ.; 97 | ὅθεν ἐκ τῶν ἱ̆ε̆ρῶ̄ν ἐ̆δᾰνεί̄ζε̆το̄ | σὺν χάριτι ̆ ἀ̆ πο̆δώ̄σεῑν |̮ ὑ̆πῑσχνού̄με̆νο̄ς, | ἀ̆πό̆ τε̆ Ῥώ̄μη̄ς | ἐ̄κ τοῦ̄ Κᾰπῐτω̄λί̆οῡ | κτλ.; 99 οἱ γὰρ περὶ σφῶ̄ν δε̆δῐό̆τε̄ς | ἢ τὰ ὄ̄ντα ̆ ἀ̆φαῑρού̄με̆νοι ̄ | ἢ τὴ̄ν πο̆λῑτεί̄ᾱν | ὅλως ἀπο̄στρε̆φό̆με̆νοι ̄ | ἐς αὐτὸν ἐχώροῡν μά̆λῑστᾱ· |; 104 | ἐπόρθει τὰ τοῖς τρισὶν ἀ̄νδρά̆σῐν ὑ̆πή̄κο̆ᾱ | ἔς τε κτλ.; 152 τοῖς δὲ τὰ μὲν ὅπλα καὶ̄ τὰ̆ σώ̄μᾰτα ̄ | ὅλα σῠνε̆κέ̆κο̄πτο̄, |; 174 | ἐπιδεικνὺς ἅμα τὴν γνώμη̄ν ἀ̆πὸ̆ τοῦ̄ σχή̄μᾰτο̄ς· | καὶ̄ ὁ̆ Καῖ̄σᾱρ σῠνεὶ̄ς | ἀ̄ντε̆μῑμεῖ̄το ̄ | ἐς δεῖ̄γμᾰ καὶ̄ ὅ̆δε̄ | τῆς εἰς τὸν Λεύκιον εὐνοίᾱς ἐ̆σο̆μέ̆νη̄ς. |; 176 | εἰ μὲν ξένος ὢν ἐπολέμη̄σᾰ, ὦ̄ Καῖ̄σᾱρ, | αἰσχρὰν ἂν τὴν τοιαύτην ἧ̄σσᾰν ἡ̄γού̄μη̄ν | καὶ̄ αἰ̄σχί̄ο̆να ̆ ἔ̆ τῑ | τὴ̄ν πᾰρά̆δο̆σῑν· |; 195 | προυκάθη̄το ̆ ἐ̆πὶ̆ βή̄μᾰτο̄ς | καὶ προσέτᾱξε̆ μὲ̆ν ἅ̆πᾱσῑ | κτλ.
It is clear enough that the use of hiatus follows a system. This system is bound up with rhythm, and hence the existence of this system helps to confirm the existence of rhythm. In other words, the lesser system implies the larger one.
IV PHRASES
The second confirmatory argument is for the reader particularly strong. Closes that appear rhythmic occur in Appian not just at the end of sentences or half-sentences but throughout the sentence. It would be rare to find 15 consecutive words, as words are conventionally counted, where there was not an apparent rhythmic close (a few seeming exceptions, like 1.451 from οἳ καὶ πάντα, disappear in the light of the discussion on hiatus above). In say Thucydides, Appian's chief model, it is not rare to find a stretch of 15 words with no close that would be rhythmic (closes that would be rhythmic are again marked with ‘|’): so 6.10.5 ὥστε χρὴ σκοπεῖν τινὰ αὐτὰ καὶ μὴ με̆τε̆ώ̄ρω̄ι τῆ̄ι πό̆λεῑ | [τῆι Krüger: τε codd.: τε τῆι Jones] ἀξιοῦν κινδυνεύειν καὶ ἀρχῆς ἄλλης ὀρέγεσθαι πρὶν ἣν ἔχομεν βεβαιωσώμεθα, εἰ Χαλκιδῆς γε οἱ ἐπὶ Θράικης ἔτη τοσαῦτα ἀφεστῶτες ἀφ᾿ ἡμῶν [ἀφ᾿ om. CMG] ἔτι ἀχείρωτοί εἰσι καὶ ἄλλοι τινὲς κατὰ τὰς ἠπείρους ἐνδοιαστῶς ἀκροῶνται. ἡμεῖς δὲ Ἐγεσταίοις δὴ̄ οὖ̄σῑ ξῡμμά̆χοῑς |̮ ὡ̄ς ἀ̆δῐκοῡμέ̆νοῑς | κτλ. (35 words between ‘rhythmic’ closes; 18 if we read ἀφεστῶ̄τε̆ς ἡ̄μῶ̄ν |), 17.7-8 οἱ γὰ̄ρ πᾰτέ̆ρε̆ς ἡ̄μῶ̄ν | τοὺς αὐτοὺς τούτους οὕσπερ νῦν φασὶ πολεμίους ὑπολείποντας ἂν ἡμᾶς πλεῖν καὶ προσέτι τὸν Μῆδον ἐχθρὸν ἔχοντες τὴν ἀρχὴν ἐκτήσαντο, οὐκ ἄλλωι τινὶ ἢ τῆι περιουσίαι τοῦ ναυτικοῦ ἰσχύ̄οντες. καὶ νῦν οὔτε ἀνέλπιστοί πω μᾶλλον Πελοποννήσιοι ἐς ἡμᾶς ἐγένοντο, εἴ τι καὶ πάνυ ἔρρωνται, τὸ μὲν ἐς τὴν γῆν ἡμῶν ἐσβάλλειν, κἂν μὴ ἐκπλεύσωμεν, ἱκανοί εἰσι, τῶ̄ι δὲ̆ ναῡτῐκῶ̄ι | οὐκ ἂν δύναιντο βλάπτειν· (56 words between, or, with a possible close at μᾶλλο̄ν Πε̆λο̆πο̄ννή̄σῐοῑ |, 34 and 20). Nor would say Lysias be unforthcoming; cf. for example 1.7 | πασῶν ἦν βελτίστη· καὶ γὰρ οἰκονόμος δεινὴ καὶ φειδωλὸς [ἀγαθὴ] [ἀγ. del. Dobree] καὶ ἀκριβῶς πάντα διοικοῦσα. ἐπειδὴ δέ μοι ἡ μήτηρ ἐτελεύτησεν, πάντων [Xpc (one letter [not ἣ?] deleted before π.): ἣ πάντων η] τῶν κακῶν ἀποθανοῦσα αἰτία μοι γεγένηται [for perfect cf., for example, 14.35 and note ἀποθ.]. ἐπ᾿ ἐκφορὰν γὰρ αὐτῆι ἀκολουθήσασα ἡ ἐμὴ γυνὴ ὑπὸ τούτου τοῦ ἀνθρώπου ὀφθεῖσα χρόνω̄ι δῐᾱφθεί̄ρε̆ταῑ· | (39 words between, or, with ἀγαθή, 8 and 30; perhaps not 19 and 17 with ἣ: ἣ might in sense make ἣ̄ πά̄ντω̄ν τῶ̄ν κᾰκῶ̄ν | an unsatisfactory rhythmic phrase).
If we take a short passage at random from Thucydides 7 (38.2), and treat it as if rhythmic, we find in the central part a looseness that would be surprising in Appian. | τῆι δ᾿ ὑ̄στε̆ραί̄αῑ | οἱ μὲν Ϲυρακόσιοι ἡ̄σύ̆χᾱζο̄ν, | οὐδὲν δηλοῦντες ὁποῖόν τι τὸ μέλλον ποιήσουσιν· ὁ δὲ Νικίας ἰδὼν ἀντίπαλα τὰ τῆς ναυμαχίᾱς γε̆νό̆με̆νᾱ | καὶ ἐλπίζων αὐτοὺς αὖθις ἐπιχειρήσειν τούς τε τριηράρχους ἠνάγκαζεν ἐπισκευάζειν τὰς ναῦς, εἴ τίς τι ἐ̆πε̆πο̆νή̄κεῑ, | καὶ ὁλκάδᾱς προ̆ώ̄ρμῐσε̄ | πρὸ τοῦ σφε̆τέ̆ροῡ σταῡρώ̄μᾰτο̄ς, | ὃ αὐτοῖς πρὸ τῶν νεῶν ἀντὶ λιμένος κληιστοῦ ἐν τῆι θαλάσσηι ἐπεπήγει. [-τοῖ̄ς πρὸ̆ τῶ̄ν νε̆ῶ̄ν | and -τὶ̆ λῐμέ̆νο̄ς κλη̄ιστοῦ̄ | possible]. In Appian, if we ignore the existence of phrases which are self-contained but non-rhythmic, that is, if we treat such phrases, except when full stop or colon follows, as part of the next phrase with a rhythmic close, then phrases containing rhythmic closes seem to have an average of six or seven words. (This average treats overlapping rhythmic closes [see p. 14 below] as two phrases.) If we take the most famously affecting passage in Thucydides Book 7—and in Appian rhythmic density is notably found at intense moments—we find sequences of at least 31 and 21 words between closes that would be rhythmic (75.2 δεινὸν … ἀπολείψει, 75.2-3 ξυνέβαινε … τινὰ τῶν; if no would-be rhythmic close at τοῦ στρατοπέδου, 54 words); there are other sequences of 14, 12, 11, 12, 9, 13 and 10 words (75.4 εἴ … ἀπιόντων and οὐκ … τό, 75.5 καὶ κατάμεμψις … πόλει and καὶ ταύτηι … τεσσάρων, 75.7 οἷς … τοῦτο, ἀντὶ δ’ … ἐναντίοις and ὅμως … ταῦτα). There is no remotely similar passage in the Civil Wars.
So far, the assumptions made on where a rhythmic close is to be found have not impeded the argument: less generosity on both sides of the equation would still leave a substantial gap between Appian and non-rhythmic prose. But working our assumptions out more explicitly will take us further into the rhythmic quality of Appian's writing. Here it will be necessary to show what a reading of the prose as rhythmic would look like, and so to posit such reading as a hypothesis; the neatness with which it fits the text will then confirm the hypothesis. But the emphasis of the discussion will not be solely on further establishing Appian's rhythmic status: the literary roles of rhythm will also start to emerge.
A good place to begin, as was glimpsed in the previous section, is particles. Some particles tend to indicate that the phrase they belong in is substantial enough to be considered rhythmic—provided, of course, that it is long enough to contain a rhythmic close. The same applies to what precedes a phrase that is marked as beginning by a particle, whether pre-positive or post-positive; so does the proviso on length. Part of what the proviso ultimately requires is that the reader (or listener) should be capable of noticing rhythmic phrases and should identify them as such if they emerge. That seems a reasonable model to work with, and will be reinforced when we come to overlapping rhythms.
Some examples, starting with two to illustrate what the reader does: 2.270 … τὸ δὲ πεζὸν αὐτὸν Πομπήιον ἄγειν κατὰ σπουδὴν ἐς τὴ̄ν Ἰ̆τᾰλί̆ᾱν | εὔνουν τε πρὸς αὐ̄τὸ̆ν οὖ̄σᾱν | καὶ πολεμίω̄ν ἔ̆ρη̄μο̄ν, | κτλ. Here τε marks the beginning of a new phrase at the word before, and, because that phrase is long enough for a rhythmic close, καί is seen to mark a new phrase too; these two phrases with rhythmic close balance each other. By contrast, in what follows, κρατυνάμενον δ᾿ αὐτήν τε καὶ Γαλατίαν καὶ̄ Ἰ̆βη̄ρί̆ᾱν |, the brevity of αὐτήν means that the reader does not attempt to make αὐτήν τε a rhythmic phrase on its own, but rather part of the single phrase marked by τε … καὶ … καί (or else of a phrase ending καὶ̄ Γᾰλᾰτί̆ᾱν |). 1.421 | ἐπὶ ἐστίᾱσῐν ἐ̆κά̆λεῑ | Νωρβανόν τε καὶ τοὺς συνόντας αὐτῶ̄ι στρᾰτη̄γού̄ς, | κτλ. The τε and καί show that the object of ἐπὶ ἐστίασιν ἐκάλει will be substantial (it continues for three more rhythmic phrases); the reader can then take that initial verbal phrase as a self-contained and rhythmic unit. Its potential rhythmic close had already opened that as an option.
4.210 Μᾶρκος δὲ Βρούτω̄ι στρᾰτη̄γῶ̄ν | προεγέγραπτο μὲν καὶ ὅ̆δε̆ δῐὰ̆ τό̆δε̄, | ἡττωμένοῡ [ἡττημένου Mendelssohn] δὲ̆ τοῦ̄ Βρού̄τοῡ | συλλᾱμβᾰνό̆με̆νο̄ς | ὑπεκρίνατο εἶναι θεράπων κτλ. Because of μέν, Βρούτωι στρατηγῶν is highlighted as part of a distinct phrase (Μᾶρκος δὲ Βρούτωι στρατηγῶν), a part taken up in διὰ τόδε as Μᾶρκος is in ὅδε. The μέν, δέ and pluperfect set up connections between the third and fourth rhythmic phrases and the first and second, and στρατηγῶν contrasts with θεράπων (cf. 185 [disguise as slave]; 194 [disguise as praetor]). The rhythm thus helps to articulate point for the phrases within the sentence.Footnote 15
1.442 | οὗτος γὰ̄ρ δο̆κεῖ̄ πρῶ̄το̄ς | οὓς ἐκό̆λᾰσε̆ θᾰνά̆τω̄ι |̮ προ̆γρά̄ψαῑ | καὶ γέρα τοῖ̄ς ἀ̆ναῑροῦ̄σῑ | καὶ μήνυτρα τοῖ̄ς ἐ̆λέ̄γχοῡσῑ | καὶ κολάσεις τοῖς κρύπτοῡσῐν ἐ̆πῐγρά̄ψαῑ. | (Cf. Mith. 86: ἐπεκήρυξε δὲ καὶ ζημίαν τοῖς καταθάπτοῡσῐν αὐ̄τοὺ̄ς | ἢ ἐπικρύπτουσι καὶ μήνυτρα τοῖ̄ς ἐ̆λέ̄γχοῡσῑν | ἢ τοὺς κρυπτομένοῡς ἀ̆ναῑροῦ̄σῑ, |.) The καί before μήνυτρα brings out the parallel between the phrase with rhythmic close that it introduces and the phrase with rhythmic close that precedes. The καί before κολάσεις brings out another parallel group of words; but this phrase is ended with the infinitive ἐπιγράψαι, which governs the preceding accusatives and itself matches the infinitive προγράψαι, which rhythmically ended the first half of the sentence.Footnote 16
1.447 … ἐπὶ τὰς πόλεις ὁ Ϲύλλᾱς με̆τή̄ιεῑ | καὶ ἐκόλᾱζε̆ καὶ̄ τά̄σδε̄, | τῶν μὲν ἀκροπόλεῑς κᾰτᾱσκά̄πτω̄ν | ἢ τείχη̄ κᾰθαῑρῶ̄ν | ἢ κοινὰς ζημίᾱς ἐ̆πῐτῐθεὶ̄ς | ἢ εἰσφοραῖς ἐκτρύχω̄ν βᾰρῠτά̆ταῑς· |. The μέν marks the start of a new development in the sentence; the three ἤs, because the units are long enough, bring out the parallel between the four participial phrases, which all end rhythmically. The first two match each other exactly, with accusative and participle; the third adds an epithet (to an accusative of a different sort); the fourth uses a dative and puts its superlative epithet at the end. The rhythm underlines and sharpens the structure. For a similar growth cf., for example, 5.327 | χαλεπαίνων δέ, ὡς ὁ Καῖ̄σᾰρ ἔ̆λε̆γε̄ν, | εἴ̄τε̆ ἐ̆πὶ̆ τού̄τοῑς | εἴτε κατὰ γνώμη̄ν ἄ̆πῑστο̄ν | εἴθ᾿ ὑπὸ ζήλου τῶν ἑτέρων μεγάλους στρατοὺ̄ς ἐ̆χό̄ντω̄ν | εἴτε Μηνοδώρου διερεθίζο̄ντο̆ς αὐ̄τὸ̄ν | ἀνοχὰς μᾶλλον ἢ βεβαίους σπονδὰς εἶναῑ νο̆μί̄ζεῑν, | κτλ. (in this case the last member begun by εἴτε turns into two phrases with rhythmic close).
3.194 (193 … στρατὸν δὲ οὐδένα πω ἔχο̄ντε̆ς ἴ̆δῐο̄ν | … ἐς τὰ̄ς νέ̆ᾱς ἀ̄ρχὰ̄ς | πάντα ἀ̆νε̆τί̆θε̄ντο̄.) | τῶ̄ι Καί̄σᾰρῐ δ᾿ ὁ̄ στρᾰτό̄ς | πελέκεάς τε καὶ ῥαβδοφόροῡς ἐ̄σκεῡᾱσμέ̆νοῡς | προ̆σᾰγᾰγό̄ντε̄ς, | ἠξίουν ἑαυτὸν ἀντιστράτη̄γο̆ν ἀ̆πο̆φῆ̄ναῑ, | πολέμου τε ἡγεμονεύοντα καὶ σφῶν αἰεὶ ὑπ᾿ ἄρχοῡσῐ τᾱχθέ̄ντω̄ν. | τε and καί indicate a substantial object; this throws emphasis on to the opening words (| τῶ̄ι Καί̄σᾰρῐ δ᾿ ὁ̄ στρᾰτό̄ς |) as a separate entity. Both Octavian and his army are emphasized by the context. The army's treatment of Octavian is opposed to the Senate's, and they try to take his magistracy into their own hands; Octavian and army are matched in the last part of the sentence (πολέμου τε κτλ.). Rhythm alerts the reader to the organization of the prose.Footnote 17
4.480 (Μοῦρκος δ᾿ αὐτοῖς καὶ Ἀη̄νό̆βᾱρβο̄ς | ἑκατὸν καὶ τριάκοντα μακραῖ̄ς ὑ̆πή̄ντω̄ν. |) καὶ αὐ̄τοὺ̄ς αἱ̄ ὁ̄λκά̆δε̄ς | ἱστίω̄ι μὲ̆ν αἱ̄ πρῶ̄ταῑ | διέ̆φῠγο̆ν ὀ̆λί̆γαῑ, | αἱ λοιπαὶ δέ, χαλάσαντο̆ς ἄ̆φνω̄ τοῦ̄ πνεύ̄μᾰτο̄ς, | ἐν γαλήνηι σταθερᾶι κατὰ τὸ πέ̆λᾰγο̆ς ἠ̄λῶ̄ντο̄, | ὑπό του θεῶν ἐκδεδομέναι τοῖ̄ς πο̆λε̆μί̆οῑς. | The μέν makes the ὁλκάδες stand out (for the order at the start καὶ αὐ̄τοὺ̄ς αἱ̄ ὁ̄λκά̆δε̄ς | cf. for example 323–4 … ἐ̆πο̆λῐο̄ρκοῦ̄ντο̄. | καὶ αὐτοῖ̄ς ὁ̆ Βροῦ̄το̄ς |). The subject αἱ ὁλκάδες is then re-expressed as two subsets. The ὁλκάδες are an unmilitary sort of ship, frighteningly confronted with a large number of military craft (cf. for μακραί and for ὁλκάδες 2.175, 204, 222, 243-5, 4.347, 368, 5.406, 526; slowness of μακραί Hann. 584). The μέν also separates αἱ πρῶται as predicate; the two rhythmic closes in swift succession for the ships that get away (… πρῶ̄ταῑ | … ὀ̆λί̆γαῑ |) contrast with the longer phrases to convey the lot of the hapless ships that are becalmed. Rhythm is used expressively.Footnote 18
From the examples already considered it will be apparent that what look like rhythmic closes occur frequently in each sentence. How tightly rhythmic closes can be packed in is illustrated by the phenomenon of overlapping rhythmic closes. Overlap has been suggested in Cicero; it operates through the reader or listener following one rhythmic close until that close is in turn absorbed into the start of the next. Some strong examples follow. 1.431 | Λαμπώνιός τε … ὅσοι τε ἄλλοῑ στρᾰτη̄γοὶ̄ | τῆς Καρβωνείου στάσεως αὐτοῖ̄ς σῠνῆ̄σᾱν, |̮ δῐέ̆φῠγο̄ν. |, cf. 413 καὶ Καρρίνᾱς δὲ̆ νῡκτό̄ς, | ὕδατός τε ὄντος ἐξ οὐ̄ρᾰνοῦ̄ πο̄λλοῦ̄ |̮ καὶ̄ σκό̆τοῡς, | αἰσθομένων μέν τι τῶ̄ν πε̆ρῐκᾰθη̄μέ̆νω̄ν, | διὰ δὲ τὸν ὄ̄μβρο̆ν ἀ̆με̆λού̄ντω̄ν, |̮ δῐέ̆φῠγε̄. |Footnote 19
1.371 τά τε ψηφίσματα ἃ ἐπεκήρῡξᾰν αὐ̄τῶ̄ι |̮ λο̆γῑζό̆με̆νοῑ |. 1.360 Ϲύλλας δὲ τοῖς πρὸς αὐτὸν ἥκουσῐν ἀ̆πὸ̆ τῆ̄ς βοῡλῆ̄ς |̮ [a lesser example] ἀ̆πε̆κρί̄νᾰτο̄ | αὐτὸ̄ς μὲ̆ν οὔ̄πο̆τε̄ | ἀνδράσι τοιάδε ἐ̄ργᾰσᾰμέ̆νοῑς |̮ ἔ̆σε̄σθαῑ φί̆λο̄ς, | τῆι πόλει δ’ οὐ̄ φθο̆νή̄σεῑν | χαριζομένηι τὴν σωτη̄ρί̆ᾱν αὐ̄τοῖ̄ς· |. 1.396 … | ἔστε Μάριον τὸν ἕ̆τε̆ρο̆ν ὕ̆πᾰτο̄ν |̮ [a lesser example] με̆γά̆λη̄ι μά̆χη̄ι | περὶ Πραινε̄στὸ̆ν ἡ̄ττῆ̄σθαῑ |̮ πῠθό̆με̆νο̄ς | κτλ.
2.360 (death of Pompey) … | ἐπιστραφεὶς ἐς αὐ̄τὸ̆ν εἶ̄πε̄ν· | ῾ἆρά σε γινώσκω, συστρατιῶτα;᾿ καὶ ὃς αὐτίκα μὲ̆ν ἐ̆πέ̆νεῡσε̄ν, | ἀποστραφέντα δ᾿ εὐθὺς ἐπάτᾱξε̆ πρῶ̄το̆ς, |̮ εἶ̄θ᾿ ἕ̆τε̆ροῑ. | In this example the reader is not waiting, say, for a main verb; the sentence moves its perspective when it had seemed complete. In 1.413 above, by contrast, the rhythm intensifies the movement back to the subject of the sentence in the final verb. In general, in these examples rhythm underlines and heightens the structure of sentences.Footnote 20
The density of rhythm varies in Appian, within the kind of limits indicated at the start of this section. This is not a sign that rhythm does not exist; rather, both density and variation can serve an expressive function. We have already seen an example (4.480, on the ships). Such features play an important role in the verbal realization of Appian's large design. It is not simply that we can accumulate enough instances to make expressiveness a plausible possibility, as we might do, say, with examples of expressive enjambement in Lucan or Statius. We can survey those places in the Civil Wars where rhythmic closes are densely packed and where at the same time unrhythmic closes are scarce. The passages, it seems, mostly contribute to the total sequence of the Civil Wars by heightening moments important to its themes or climactic in its structure. Books 2 and 3 will serve as examples. One of the two large groups of passages marks points where the civil strife of the Romans reaches peaks of action or reaction: so 2.75-7 on the killing of Clodius and the response, 141 (second part) on terrified flight as Caesar invades, 257 (last part) on panic at Caesar's defeat, 495 (first part) on killing in the tumult at Caesar's death, 629 on Caesar's victories over the Pompeians, 3.32 (first part) emotive report of Caesar's death to Octavian. The other large group helps to highlight the movement towards monarchy: so 2.446 on Caesar's apparently kingly behaviour, 3.39-40 enthusiastic response to the new Caesar, 3.378 Octavian's entry into Rome, and various significant passages on the relation of Antony, Octavian, the soldiers and the Senate (so 3.185, where Octavian is winning legions over). Passages in speeches contribute too; they not only intensify the rhetoric but underline the issues of the work: so the matching passages 2.570 (circumstances, and beginning, of Brutus’ speech) and 3.53-4 (near beginning of Octavian's speech to Antony) on the interpretation of the conspirators’ deeds.Footnote 21
A few examples of density (and variation) may be considered in more detail. The first is 3.65: τοιαῦτα τοῦ Καίσαρος εἰπόντος ὁ Ἀντώνιο̄ς κᾰτε̆πλά̆γη̄, | τῆ̄ς τε̆ πᾱρρη̄σί̆ᾱς | καὶ̄ τῆ̄ς εὐ̄το̄λμί̆ᾱς | παρὰ δό̄ξᾰν οἱ̄ πο̄λλῆ̄ς | καὶ̄ πᾰρ᾿ ἡ̄λῐκί̆ᾱν |̮ φᾰνεί̄ση̄ς. | χαλεψάμενος δὲ τοῖς τε λόγοις οὐχ ὅσον ἔδει τὸ πρέπον ἐς αὐ̄τὸ̆ν ἐ̄σχη̄κό̆σῑ | καὶ μάλιστα τῶν χρημάτων τῆ̄ι ἀ̆παῑτή̄σεῑ, | αὐστηρότερον αὐτὸν ὧ̄δε̆ ἠ̄μεί̄ψᾰτο̄· | (related to the opening here, but less dense, is Syr. 206, on Scipio's awe-inspiring appearance). The short rhythmic phrases, marked out by prepositives and postpositives, show the impression which Octavian makes on Antony; the awed intensity fits the whole place of the future monarch in Appian's grand architecture. Longer phrases follow, on the most probable scansion; they portray the less noble reasons for Antony's chilly response.Footnote 22
At 2.286, density highlights Pompey's imminent fall and his uneasy consciousness of it. The passage is leading up to Pharsalus. ἅπερ ὁ Πομπήιος οἷα πολέμων ἔμπειρος ἀπεστρέφετο καὶ νεμεσῶ̄ν ἐ̆π᾿ αὐ̄τοῖ̄ς | ἐ̆νε̆κᾰλύ̄πτε̆το̄, | κᾰτε̆σῐώ̄πᾱ δ᾿ ὅ̆μω̄ς | ὑ̆πὸ̆ ὄ̆κνοῡ καὶ̄ δέ̆οῡς, | ὥσπερ οὐ̄ στρᾰτη̄γῶ̄ν ἔ̆τῑ, | ἀλλὰ̄ στρᾰτη̄γού̄με̆νο̄ς | καὶ πά̄ντᾰ πρά̄σσω̄ν | ὑ̆πὸ̆ ἀ̆νά̄γκη̄ς |̮ πᾰρὰ̄ γνώ̄μη̄ν. | τοσοῦτον ἀ̄νδρὶ̆ με̆γᾰλοῡργῶ̄ι | καὶ παρὰ πᾶν ἔργον ἐς ἐκεί̄νη̄ν τὴ̄ν ἡ̄μέ̆ρᾱν | εὐτυχεστάτω̄ι γε̆νο̆μέ̆νω̄ι | τὸ δύσθυμο̆ν ἐ̆νε̆πε̄πτώ̄κεῑ, | εἴτε ὅτι τὰ συμφέρο̄ντᾰ κρί̄νω̄ν | οὐ̄κ ἔ̆πεῑθε̄ν, | ἀλλ᾿ ἐπὶ κύ̆βο̆ν ἐ̆χώ̄ρεῑ | πλήθους ἀνδρῶν τοσῶ̄νδε̆ σω̄τη̄ρί̆ᾱς | καὶ τῆς ἑαυτοῦ δόξης ἐς τό̆τε̆ ἀ̆η̄ττή̄τοῡ· | εἴτε τι καὶ μαντικώτερον αὐτὸν πλησιάζοντος ἤδη τοῦ κακοῦ̄ σῠνε̆τά̆ρᾱσσε̄, | μέλλοντα τῆς ἡμέρᾱς ἐ̆κεί̄νη̄ς | ἐκ δυναστείᾱς το̆σῆ̄σδε̄ | ἀθρόως ἐ̄κπε̆σεῖ̄σθαῑ. | This passage, like the preceding one, shows perceptible changes; its overall density remains notable. The relatively long unrhythmic start on Pompey's disdain of his imperceptive fellows is set against a more inward passage (καὶ νεμεσῶν … γνώμην), in which almost every syllable is part of a rhythmic close. For the pair ὑπὸ ὄκνου καὶ δέους as a rhythmic phrase, cf. for example Sam. fr. 12.4 | ὠμῶ̄ς δ᾿ αὐ̄τοὺ̄ς καὶ̄ πῐκρῶ̄ς | κτείνω̄ν τε̆ καὶ̄ σῡλῶ̄ν |, Hisp. 88 | ὁ̆ δὲ̆ τό̄λμη̄ι καὶ̄ τύ̆χη̄ι |, Pun. 552 | ὁ ἐμὸ̄ς τρό̆πο̄ς καὶ̄ πό̆νο̄ς |. The neat ὥσπερ οὐ̄ στρᾰτη̄γῶ̄ν ἔ̆τῑ, | ἀλλὰ̄ στρᾰτη̄γού̄με̆νο̄ς | finds significant echoes after this passage, but the echoes have a less compactly rhythmic form: 299 στρατηγεῖ̄τε̆ τοῦ̄ πό̆νοῡ | μᾶλλον ἢ̄ στρᾰτη̄γεῖ̄σθε̄· |, 307 | καὶ οὐδὲ στρατηγοῦντα ἔτι μᾶλλον ἢ̄ στρᾰτη̄γού̄με̆νο̄ν. | , 4.520 ῾ἐοίκαμεν ὡς Πομπήιος Μάγνος πολεμήσειν, οὐ στρατηγοῦντες ἔτι μᾶλλον ἢ̄ στρᾰτη̄γού̄με̆νοῑ.᾿ | (cf. also Syr. 20 ἐλπὶς δὲ καὶ ἔχο̄ντᾰ [sc. πάντα] βᾰσῐλεύ̄σεῑν | βασιλευόμενο̆ν ὑ̆π᾿ ἐ̆κεί̄νοῡ. | and, for example, Hann. 129 ὑποπτευόμενό̄ν τε̆ ὑ̆π᾿ ἐ̆κεί̄νοῡ | καὶ ὑποπτεύο̄ντε̆ς αὐ̄τό̄ν. |). The phrases broaden a little with the narrator's perspective (τοσοῦτον … ἐνεπεπτώκει); tight phrases conveying Pompey's possible thoughts (τὰ συμφέροντα κτλ.) enlarge somewhat as he considers the momentousness of the occasion for his side and himself (πλήθους … ἀηττήτου). A long unrhythmic sequence stands back to consider a more mysterious possibility (εἴτε τι κτλ.); a set of rhythmic phrases (μέλλοντα κτλ.) then trenchantly presents the coming heavy and sudden fall, again a part of Appian's grand scheme (note Proem 59 on the division of the Civil Wars by generals). Such a detailed reading of the fluctuations appears to be invited by Appian's overall use of rhythm in the work; that use certainly indicates a connection between the density in this passage and its power within the big design.
Density can mark out vital passages that have a lower emotional temperature. So 3.370: ἀθρόα δὴ πάντων ἐς πάντα ἦ̄ν με̆τά̆θε̆σῑς, | ἀντὶ μὲν δισχιλίων καὶ πε̄ντᾰκο̆σί̆ω̄ν δρᾰχμῶ̄ν | τὰς πε̄ντᾰκῑσχῑλί̆ᾱς, | ἀ̄ντὶ̆ δὲ̆ τῶ̄ν δύ̆ο̆ τε̆λῶ̄ν | τοῖς ὀκτὼ̄ δο̆θῆ̄ναῑ | Καί̄σᾰρά̆ τε̆ αὐ̄τοῖ̄ς | ἀντὶ τῶν δέκα ἀνδρῶ̄ν δῐᾰνέ̆μεῑν | καὶ ἐς τὴ̄ν ὕ̆πᾰτο̆ν ἀ̄ρχὴ̄ν | παραγγέλλεῑν ἀ̆πό̄ντᾱ. |. The sentence begins with a sweeping phrase on the sudden alteration in the Senate's scheme (ἀθρόα … μετάθεσις), and, as first item, a sum of money (ἀντὶ … δραχμῶν); the rhythmic pace then quickens. Indignity attends the extraordinary and soon repented changes of mind which have been produced by Octavian's advance towards Rome (cf. 349, 355, 357, 364). The final clause and word provide a remarkable culmination. This passage too shows the rise of Octavian and the progress towards monarchy (cf. 371).
The three passages exemplify the connection between Appian's shaping on a massive scale and on a small one. Scholars are now, as we glimpsed earlier, more aware of Appian's large and pervasive structuring; we should endeavour to consider how that structuring is implemented on the level of style.Footnote 23
As this section has started to indicate, the realization that Appian is rhythmic, and that rhythm runs right through his sentences, should transform our view of him as a writer. His writing now reveals itself as expressive and pointed, in the service both of local impact and argument and of an immense edifice. We will need to engage more responsively with the specifics of his writing: they are a fundamental part of his literary creation.
V IMPERIAL GREEK PROSE: PROSPECTS
If the significance of the subject is allowed for Appian, investigation and study of more authors would seem desirable (I hope to carry some out). But which authors? It would need much more work to attain a full knowledge of how far rhythmic prose extends in the Imperial period; but the discussion so far will enable us to form some preliminary ideas. We will return to the set of twenty sentences each from authors who are unlikely to be rhythmic. They could provide a point of comparison for samples of twenty sentences from Imperial authors, each of whom might or might not be rhythmic. If the pre-Imperial set were an instance of ‘normal’ distribution, that would add precision to our comparison. We could then work out for this set a sample standard deviation (s) from the mean (μ): μ would be the mean number of sentences out of 20 which would, if the author were following the later system, close rhythmically. This pre-Imperial s could then be used to measure the Imperial writers. To produce it, we would find μ, add together the square of the differences of each example (each 20) from μ, divide by the total of examples – 1, and obtain the square root. Initially, we could get a better idea of the distribution with more examples; so we could break the 400 random sentences up into 100 groups of 4. Two of these groups show 0 sentences that would close rhythmically, 15 show 1, 29 show 2, 40 show 3, 14 show 4. μ is 2.49, s is 0.9795381307. Two groups would fall below μ – 2s (i.e. μ – [2 × s]), 15 within μ – 2s, 29 would be within μ – s, 40 within μ + s, 14 within μ + 2s. The general structure and the percentages in the categories (μ + or –) s, 2s, 3s roughly resemble those of a normal distribution (within s: 69%; within 2s: 98%; within 3s: 100%; expected: about 68%, 95%, 99.7%). The 29 and 40 depart in a limited way from the symmetry of the Gaussian curve which depicts normal distribution; that departure might easily be rectified with a sample statistically more adequate than 100 (29 is 72.5% of 40). At all events, we could now hypothesize s for the twenty pre-Imperial groups of 20, not to prove anything, but as a means of articulating the divergence between the pre-Imperial list and samples of 20 from some Imperial writers.Footnote 24
Samples of twenty sentences were assembled from Achilles Tatius, Aelius Aristides, Alexander of Aphrodisias (De Anima, De Fato, De Mixtione), Appian (Bella Ciuilia), Arrian, Cassius Dio, Chariton, Clement, Dio of Prusa, Epictetus/Arrian, Galen, Hebrews, Heliodorus, Herodian the historian, Josephus, Longinus, Longus, Lucian, Maximus of Tyre, Pausanias, Philo, Philostratus (narrative parts of Vitae Sophistarum; separate sample for quotations from speeches), Plotinus, Plutarch, Polemon, Theon, Xenophon of Ephesus. In some cases, separate samples of 20 were taken from different works or groups of works. It may be remembered that no author in the pre-Imperial set showed more than 14 rhythmical closes out of 20, a point which would lie just beyond (μ = 12.45+) hypothesized s, that is 13.98811231. Of the Imperial authors, Cassius Dio, Herodian, and some works of Galen showed 16, beyond the hypothesized pre-Imperial 2s (15.5622462); Achilles Tatius, Alexander of Aphrodisias, Chariton, some Galen, Longinus, Philo, Polemon: 17, just within 3s (17.06433693); Appian, Josephus, Bellum Judaicum and Contra Apionem (not the Antiquitates), Longus: 18, beyond 3s—a significant boundary, if we have to do with a normal distribution for the pre-Imperial authors; Plutarch's declamatory De Fortuna Romanorum: 19, beyond 4s (18.60244924).
For each of these authors singly (Herodian, etc.), the result is notable: in a normal distribution, about 95% would be within 2s, so results of 16 already prompt more investigation. (A result below 15 should not preclude it for other authors.) In those cases where there has been some earlier research, it confirms the idea that the author is rhythmic. Together the results for all these authors (Herodian, etc.) suggest much rhythmic writing in the period, with intriguing generic and other connections. The prominent place of historiography forms a contrast with Roman historiography, where Sallust, Livy and Tacitus (Histories, Annals, Agricola, Germania) stand out from their times in not writing rhythmically. But there also exists Greek historiography in the Imperial period which is unlikely to be rhythmic. So Arrian's Anabasis contrasts with Herodian, etc.: a sample showed 9 out of 20, strikingly low within the whole group of Imperial authors sampled, and below 2s. Narrative obviously connects historiography with the novel. Achilles Tatius, Chariton and Longus look rhythmic on the present basis; Xenophon of Ephesus may have reached us in an abridged form; the late Heliodorus is the other author who presents us with 9. The position with rhetoric appears complicated; the Hellenistic material would have suggested it was the first place to look. It was indeed where some earlier scholars looked.Footnote 25
One would get further with an s which was smaller in proportion to the numbers in each example, that is with a larger number of sentences in each; to see the possibilities, one may experiment with 50 sentences each (one could get a hypothetical s for eight groups of 50 in the pre-Imperial writers and add a further two groups from Lysias). Of two samples of 50 sentences from Plutarch's Lives, one is beyond the pre-Imperial 2s, one well beyond 3s; of two samples of 50 from the Moralia, both are beyond s and within 2s. So far the Lives look to be rhythmic, and to have a higher proportion of rhythm than the Moralia. A sample of 50 from Galen's less technical writing takes us well beyond 4s; a sample of 50 from his more technical writing is within s. There look to be important distinctions within writers’ œuvres. Across authors, the use of rhythm would serve different functions; thus the use in Plutarch's Lives would differ from that in Appian. Plutarch's use would, for instance, highlight the imaginative richness of his vocabulary, so unlike Appian's forcefully restricted and recurring words and phrases. These different stylistic functions would relate to differences in thought and large structure. It looks as if rhythmic writing is a feature which should affect our overall vision of the literary landscape in this period; it has also emerged as a means of expression and articulation which should influence our reading of rhythmic texts. Both aspects are important incitements to investigating further.Footnote 26
