^{page 257 note 2} The evidence is presented throughout in a tripartite classification: (i) State documents: i.e. all inscriptions set up at the instigation of the (ii) ‘Sub-units’ documents: i.e. inscriptions set up by (a) state ‘sub-units’ themselves (tribes, demes, phratries), (b) clubs and priestly brotherhoods (e.g. ) (c) individuals or groups of individuals in consequence of having held some official position (iii) Private documents: i.e. all inscriptions set up by individuals in a private capacity. ‘State’ documents have no sign added to the numbering to denote them; ‘Sub-unit’ documents are denoted by ‘U’ written between line number and date; ‘Private’ documents are denoted by ‘P’ in the same position.

^{page 257 note 3} This is the only instance of this verb which I have found in the period in question. However, in IG i². 4. 12 (485/4) we read and in i². 114. 42 (409/8) (˜ noun ). In all other cases is used for fining and penalizing.

^{page 257 note 4} See also p. 258.

^{page 258 note 1} For see p. 268 ( before ov).

^{page 258 note 2} But it is worth noting how often Private inscriptions present us with unusual forms, the majority of which probably do no more than betray the ignorance or carelessness of the stonecutter or his patron.

^{page 258 note 3} For see p. 268 ( before o).

^{page 259 note 1} The Greek Dialects, p. 369.Google Scholar

^{page 260 note 1} See Meisterhans, , Grammatik der attischen Inschriften³, p. 67, n. 586.Google Scholar

^{page 260 note 2} Op. cit., p. 168.

^{page 261 note 1} Op. cit., p. 33.

^{page 261 note 2} In each section any examples of the omission of the intervocalic iota are quoted first, followed by any of the same word(s) with the iota retained. Then the total number of examples of other words with the iota retained is given, and a list (without references) of the words (or a sample of the words) which compose this total. (Note that, since each section is divided into State, Sub-unit, and Private, ‘U’ and ‘P’ are not used in the numbering.)

^{page 262 note 1} Note that the intervocalic ι in is never omitted in any case. is therefore here excluded from all the relevant combinations.

^{page 262 note 2} Cf. (7) below.

^{page 263 note 1} Figures in brackets after examples in dicate the number of examples of omission.

^{page 263 note 2} Note that, unlike demoticsx in those in generally tend to drop the iota in this period. But in the fourth century before 323 B.C. the vast majority of demotics both in and retain the iota. Cf. demotics in p. 265.

^{page 263 note 3} Cf. (b) (5) below.

^{page 263 note 4} I have excluded instances of here because I have found only one example of omission, out of scores of occurrences of the word in decrees, viz. 373. 17 (322/1). From 323 B.C. to the end of the fourth century alone there are 68 examples of Cf. see (3) below.

^{page 264 note 1} For and see (6) below.

^{page 264 note 2} never loses the iota in any case and is therefore excluded from the lists.

^{page 264 note 3} Cf. (a) (5) above.

^{page 265 note 1} As with demotics in tends to drop the iota during the period under consideration, Before 323 B.C. the earliest instances have the iota, although omission occurs as early as IG i². 202. 36 (443/2). Cf. p. 263 n. 2.

^{page 266 note 1} For this and later Tables see CQ, N.S. xiv (1964), 241 n. 4, and note further that in the present article the evidence of a third Dating Category—which comprises inscriptions of a very inexact or doubtful date—is introduced. This was necessitated by the fact that otherwise the important evidence of private documents would be virtually completely neglected.Google Scholar

^{page 266 note 2} This is true also of the period before 323 B.C., and after 146 B.C. the preponderance of is even greater.

^{page 266 note 3} Note that the iota in and is never omitted. This is characteristic of all periods.

^{page 267 note 1} Note that 2322. 325 (183/2) and 11681. P (post fin. s. iv) are the only examples with the iota retained. This is in accord with the periods before 323 B.C. and after 146 B.C.

^{page 267 note 2} The latest example of is otherwise 116. 12 (361/0).

^{page 268 note 1} Perhaps mere pedantry. Cf. the papyrus text of Antiphon the Sophist (Diels-Kranz, , ii. 346, frag. 44A, col. 2, II. 3 ff.):Google Scholar

^{page 268 note 2} Intrusion before o, ov, and w is particularly common in Private inscriptions in the fourth century before 323 B.C.

^{page 268 note 3} But this was the original form of the word, and so, although otherwise omitted in the period 323–146 B.C., this is a return of an original iota.

^{page 268 note 4} See pp. 259–60.

^{page 268 note 5} I am here concerned to modify certain statements made by Meisterhans's Gram-matik der attischen Inschriftenf³ and to present additional evidence which is not offered in that work.

^{page 269 note 1} IG is surely wrong here in reading

^{page 269 note 2} This calls to mind the famous poem of Anacreon, but the possibility that acquaintance with that poem influenced the drafter in his choice of form is very remote indeed.

^{page 269 note 3} For a discussion of the example quoted in § 24. 2, see CQ N.S. xiv (1964), 245–8.Google Scholar

^{page 270 note 1} But note that six of the seven State examples of elision occur in the same inscription, so that, if we except that inscription, this is a case of a preponderance of hiatus.

^{page 271 note 1} Note that shows more examples of hiatus than of elision except in State inscriptions. Note particularly that there are no examples of elision in Private inscriptions.

^{page 272 note 1} See n. i on p. 273.

^{page 273 note 1} In the period under discussion there are three examples of [1281. 10 (c. 866); 891. 6 (188/7); S. xvi. 94. 12 (173/2–168/7)], and only two of 945. 12 (168/7) and H. xi. 295. 58. 6 (152/1).

^{page 273 note 2} The nine examples quoted of this formula () are merely those counted down to 285/4. There are many more.

^{page 274 note 1} The stone actually reads and it is clear that what we have here is a confusion between and It would also be possible to divide the letters = ‘other things as many as he may wish in any respect’. This would save the stone-cutter's reputation, although the expression would be more natural without τι. Taken as IG takes it, there is an elision of either or (almost certainly the former). Reading there are two elisions, and

^{page 275 note 1} For ephelkustic nu, see below pp. 277–284.

^{page 275 note 2} In this section, and in section B which follows, I note all consonantal changes, whether I consider them to be genuine linguistic phenomena or exceptional forms to be explained in some other way.

^{page 275 note 3} See Grenfell, and Hunt, , The Hibeh Papyri, i. 49–55, no. 14, lines 28 and 41.Google Scholar

^{page 275 note 4} Cf. Buck, , op. cit., p. 133, § 166, 2: (formed after in Hesiod), instead of usual are frequent in Boeotian, and Thessalian also has and Google Scholar

^{page 276 note 1} Note that after 146 B.C. there are several examples of interchange between the spirants, e.g. 8439. P (s. i); 12098. P (s. ii/i); 7943. P (s. i).

^{page 276 note 2} Sturtevant, , The Pronunciation of Greek and Latin, p. 64, § 71, argues that the use of γ = [η] can be explained only if there was an actual change of γ = [g] to a nasal in some position, viz. before a nasal: so γv and γμ change from [gn, gm] to [ηn ηm]. He explains the spelling of on this basis.Google Scholar

^{page 277 note 1} There is a distinct word ‘grandson’. But as used in inscriptions in the formulae and there is no distinction of meaning between the two forms.

^{page 277 note 2} For before θ and φ see (c) (ii) above.

^{page 277 note 3} Cf. Meisterhans, § 44.

^{page 277 note 4} It is necessary to divide each subsection into (a) examples before vowels and those before consonants, and (b) those in pause and those in media oratione. Note that the trends before and after 323–146 B.C. are dealt with on pp. 283–4.

^{page 280 note 1} The only exception is k-3056.3 (320/19) where the strong break in punctuation is tantamount to making the last word of a line.

^{page 282 note 1} I exclude examples where there is no following word, i.e. the relevant parts of (d) and (e) and the whole of (f).

^{page 282 note 2} Note that in 487. 9–10 (c. 304/3) we find the third person singular of the present subjunctive of with a final nu: This may be a misplaced ephelkustic nu, but on the other hand it may have occurred because of the preceding infinitive (line 8).

^{page 283 note 1} The only exception is which always has the nu in all periods in the formula For this reason the frequency of has not been taken into account in the above examination of ephelkustic nu.

^{page 284 note 1} In order to give a more detailed analysis I have divided each section into 4 groups: (i) at the end of the oblique cases of the article; (ii) (iii) (iv) any other words ending in State: 6 instances, all assimilated, e.g. 1631. 254 (323/2), H. III. 68. 62. 2 (post 166). Sub-unit: 6, all assimilated, e.g. 1243. 21 (med. s. iii), 1224. a. 15 (c. 166). Private: one, unassimilated, 3779. 18 (med. s. iii).

^{page 286 note 1} Apart from these securely dated examples there are forty-five examples of assimilation in 1678, which is dated ‘ante 315’ and may therefore not belong to the period under discussion at all. The examples consist mainly of There is also one other example of doubtful date, 696. 5 (post 303/2).

^{page 287 note 1} Note that 1632 contains many examples of ‘general’ assimilation, as does 1631.

^{page 288 note 1} There are no instances of -v before gutturals in Private inscriptions.

^{page 288 note 2} There is also one ‘false’ assimilation, 521. 5 (fin. s. iv). This may well be a simple error, but it is just possible that the writer is doing his utmost to avoid assimilation which he considers ‘incorrect’. To achieve this he writes μ instead of v. State: 29 examples of only one of 467. 10 (306/5). Sub-unit: 2 of none of Private: 3 of none of

^{page 293 note 1} Cf. Meisterhans, , p. 119 (on First Declension Proper Nouns in ηs): ‘Doch lassen sich seit dem IV. Jahrhundert bei diesen Wörtern vereinzelte Übergänge in die III. Deklination nachweisen.’ P. 133 (on Proper Nouns in ): ‘Der Genetiv endigt im VI. und V. Jh. auf Seit 350 v. Chr. begegnen vereinzelte Formen auf die aber nicht durchdringen und später wieder verschwinden.’ Pp. 134–5: ‘Der Genetiv hat bis 350 v. Chr. die Endung -o(v)s. Von da an begegnen sehr häufig Formen auf -ov. Doch überwiegt bis zum Jahre 300 im allgemeinen noch die ältere Bildungsweise auf- ous. Von 300 an aber gelangen die Formen auf -ov gänzlich zur Herrschaft. Erst die Kaiserzeit, die Periode der allgemeinen Reaktion, wendet die alte Endung wieder in weiterem Umfang an.’Google Scholar

^{page 294 note 1} It is interesting to observe that the three examples of in the period 274–250 all belong to the same inscription, 678 (256/5); also that in the period 249–225 three of the six examples of belong to 681 (249/8) and the other three to 1534. B (247/6). This confines nine of the eleven securely dated examples to the period 256–246 B.C.

^{page 294 note 2} Note that thirteen of the twenty-one examples of -ovs in Category 1 of State inscriptions in the period 323–300 come from only two sources, 1631 (323/2) and 1632(323/2).

^{page 295 note 1} It is also the case that, since nouns in (from stem ) have epsilon twice in the stem and do not contract the first epsilon of the stem in the Accusative and Genitive, the resultant presence of epsilon before the inflexion in these cases is a visible marker which does not exist in nouns in in which epsilon of the stem has already contracted with -a and -os to produce (v) and -ous.

^{page 295 note 2} Cf. Meisterhans, , p. 136: ‘Im Akkusativ erscheint die alte Endung -η noch in einigen Beispielen. Sonst erscheint überall namendich seit dem IV. Jahrhundert (über das V. Jahrhundert sind wir leider so gut wie gar nicht unterrichtet. Erst die Kaiserzeit weist vereinzelt wieder auf).’Google Scholar

^{page 295 note 3} The iota here is the intrusive intervocaliciota: cf. p. 268 above.