On the analogy of the colloquial register in some modern languages, where narrative and argument may be punctuated by oaths and exclamations (sometimes obscene or blasphemous) in order to maintain a high affective level and compel the hearer's attention, it is reasonable to postulate that Attic conversation also was punctuated by oaths, that this ingredient in comic language was drawn from life, and that the comparative frequency of ║ (|) M M (M) Δ in comedy is sufficiently explained thereby. There are obvious affinities between some passages of comedy, relaxed conversation in Plato and Xenophon, and the forceful, man-to-man tone which Demosthenes sometimes adopts to such good effect (e.g. xxi 209). Compare, for instance, Ar. V. 133 f. ἔστιν δ' ⋯νομα ⋮ τῷ μ⋯ν γέροντι ⋮ Φιλοκλέων / να⋯ μ⋯ Δία, τῷ δ' υἱεῖ κτλ., where the oath is a response to imagined incredulity, and X. Smp. 4.27 αὐτ⋯ν δ⋯ σέ, ἔɸη, ⋯γὼ εἶδον να⋯ μ⋯ τ⋯ν Ἀπόλλω, ὅτε κτλ. (‘Oh, yes, I did!’).
It is also important that the commonest oaths fit, in most of their forms, the end of an iambic trimeter: (να⋯) μ⋯ τ⋯ν Δία, ν⋯ (τ⋯ν) Δία,ν⋯ τούς θεούς, μ⋯ τοὺς θεούς. Add that in Aristophanic dialogue (by contrast with Menander) over half the iambic trimeters end with major pause, and half the remainder with minor pause, and we can see why Δ / established itself early as a distinctive comic pattern. Out of 105 examples of M M (M) Δ cited from comedy in Section II above, 59 have the oath at verse-end.
In the case of πάνυ, which was almost exclusively Attic and — to judge from its great rarity in tragedy — felt by Athenian poets to be prosaic, we lack evidence on its functions in the colloquial register; it may or may not have served as affective punctuation. In prose, we have to reckon with the fact that π Mπ and Mπ π constituted a genuine stylistic choice (cf. n. 32) as far back as the evidence will take us, since the two earliest instances in prose are [X.] Ath. 2.3 πάνυ δι⋯ χρείαν and ibid. 3.5 πολλ⋯ ἔτι πάνυ. The oath, as treated by the comic poets on the basis of colloquial usage, is bound to have served as a model for πάνυ, exerting an influence which pulled πάνυ to the end of the verse, but there was also a powerful metrical constraint. As a dibrach ending in a vowel which could not be elided or enter into crasis, πάνυ was especially appropriate for verse-end. That in itself was enough both to establish Mπ π as the dominant pattern in comedy and to promote Mπ … π. Out of the total of 104 examples of Mπ (…) π in comedy, 93 have πάνυ at verse-end, which makes Mπ (…) π / one of the hallmarks of comic style. Mπ … π does not occur in prose in association with any other feature identified as colloquial, but it should be noted that Aiskhines and Demosthenes are much fonder of Mπ π than other prose authors. In some cases one can see that the order Mπ π avoids a succession of short syllables (e.g. D. xviii 130, liv 1) or hiatus (e.g. D. xxx 36) or both (e.g. D. xliii 10), but there are other cases in which it has the opposite effect (e.g. D. xxiv 140, xliii 53). The possibility of comic influence on oratorical language cannot be dismissed out of hand. It is also possible that someone will find positive determinants which will explain all the cases of Mπ π in prose.
σɸόδρα, which, like πάνυ, is peculiarly Attic, is metrically more tractable than πάνυ, since it can be elided; even so, out of the 80 comic examples of Mσ (…) σ no less than 58 have σɸόδρα at verse-end, and of those 58 there are 22 at major pause, 8 at presumed major pause and 9 at minor pause. The comic treatment of σɸόδρα is thus comparable with the treatment of πάνυ, and Timokles (CGFP) 222(b).4 τηρεῖν…σɸόδρα is in fact the closest analogy we have to Ar. Pl. 234 f. ἄχθομαι…πάνυ.
δέ and γάρ are a different matter, and in some significant respects different from each other. Postponement of δέ is especially prominent in Aeschylus (45 examples, including a few in which the text is suspect) and then abundant in fourth-century comedy. It is much less common in Euripides (18 examples), rare in Sophocles (6) and Aristophanes (6), and virtually limited in prose to the categories which I labelled (l)–(3). There is as yet no evidence to associate postponement of δέ with colloquial language; on the contrary, it seems to have begun as a feature of poetic language and to have been taken up and exploited by fourth-century comedy. If, in addition to being Aeschylean, it was colloquial in the fourth century, what happened to it afterwards? Except for such an isolated and inexplicable case as Diod. xx85.1 (v.l.!) — in a military narrative — it is not a feature of the Koine at literary, documentary or subliterate level.
Postponement of γάρ was no doubt encouraged by postponement of δέ, but it is not itself notably poetic (20 examples in tragedy, of which only three come in my class (5)). One can see how it could possibly have developed in the spoken language of the fourth century, extending the function of γάρ as an explanatory particle (rather on the lines of γε) in a way which makes it comparable with the English ‘you see’ in (e.g.) ‘He didn't dare pick it up. He hurt his back last year, you see’. For an extension of this kind we may compare the current extension of the English genitive affix in (I heard both examples a year or two ago) ‘Then the girl whose place she was taking's mother turned up’ and ‘The man that Christopher liked's Introduction is much better’. Moreover, postponed γάρ appears in a segment of conversation constructed in indirect speech by Theophrastus in Char. 8.9 τ⋯ πρ⋯γμα βο⋯σθαι γάρ (p C N γάρ). Again we must ask: what happened to it afterwards? A couple of cases in Theophrastus' botanical works (CP iii 11.3 and HP iv 6.1) could be a reflex of the influence of comedy on literary language at Athens. The influence was plainly short-lived, since it did not affect the Koine.
It is not hard to see why serious poetry in the fifth century and earlier should have experimented occasionally with the postponement of δέ and γάρ: treatment of M M q as a valid alternative to M q M is metrically very convenient. No poet, however, could afford to use common words in a bizarre, un-Greek way merely to save himself time and trouble in constructing a verse. Linguistic innovation is normally analogical, proceeding by extension from a starting-point already there, and the most obvious starting-point for postponement of δέ and γάρ is constituted by my class (3). This consideration provides comic postponement with a pedigree, but does not deny it individuality. The remarkable scale and frequency with which comedy exploited a phenomenon which tragedy used with restraint and prose hardly at all gives comic postponement the right to be regarded as a quite distinctive artificial feature of comedy.