labic form like ab-is was not permitted; similarly all the group-forms of idem were tetrasyllabic like sed-eodem, and a trisyllabic form was not allowed. In the same way we must explain the striking fact pointed out by Engelbrecht (Wien. Stud. VI, 236 ff.), that, while the simple verb eo and the compounds of eo with a long preposition show only the contract forms before s, c.g. īsti (Tri. 939), īsse (ib. 944), exissem (Ru. 534),1 yet, in the case of the compounds with a short preposition, the verse of the dramatists often requires and always admits the full forms with double i, e.g. ăbiisti, abiisse, abiissem, obiisti, subiisti, adiisti, rediisti, periisti, intěriisti; cf. adiese and adiesent (CIL. I, 196, 7 f.) in the S. C. de Bacchanalibus of 186 B.C. This is the special peculiarity in the early treatment of these forms which Brock (.., 78) seems to regard as inexplicable, and it is evident that in the second century B.C. the short i in hiatus of these forms was preserved from contraction by a psychological cause, that is, by the sense of logical analogy. The perfect of the simple verb is inflected ii, īsti, iit, and thus a single short is lost by contraction before s; a pair of shorts, however, has become closely associated in the Roman mind with the perfect forms ăbii, ăbiisti, ăbiit, etc., and, in view of this strong association, the loss of one of the pair, as in the contract form abisti, would make the form seem incongruous, and would seriously obscure its connection with the remaining forms of the perfect. Therefore logical analogy forbids the usual contraction to take place, or, to speak more accurately and to adopt the language which is suggested by Exon's admirable discussion of similar linguistic processes (Hermathena, XIII, 145 ff.), the contraction actually takes place, and the incongruous form interīsti is produced, but the sense of analogy immediately awakes, and a 'special sound-law' arises, in virtue of which i is retained before is, after a short prefix, as in interiīsti (interieisti, CIL. I, 1202).2 Special sound-laws,' however, often find special difficulty in maintaining themselves 1 The only exception is vēniisse (St. 232; Ps. 1090), which should perhaps be written venivisse. 2 Cf. also, for the quantity of the second i, Marx, Hülfsbüchlein f. Ausspr.3 10. fully, and forms like abiīsti, abiissem, which were doubtless in exclusive use in the second century B.C., were compelled, in the following century, to yield a portion of their territory.1 It should be added that this rhythmical epenthesis, this tendency to retain i, exists in a less degree in all verbs beginning with a short syllable, notably in petiisti (cf. Verg. Aen. ii, 25 nos abiisse rati et vento petiisse Mycenas), but the great number of the compounds of ire with short prefix caused analogy to be most strongly felt, and uniformity to be longest maintained, within this single group. (6) Question of Synizesis in Verse-Closes. As we have already seen (pp. 165,195, n. 1), the principle of metrical regularity excludes a contract form like m(e)ó from the close of all verses which conclude with an iambus.2 Similarly the contract dative mi, which, unlike m(e)o; is a perfect monosyllable, and occurs as such in all periods of the language (L. Müller, R.M.2 296), is allowed, so far as I am aware, only in verses which close with a trochee, e.g. Mo. 175 (troch. oct.); ib. 871 (bacch. tetr.). The exclusion of synizesis forms also from the close of verses which end with a trochee () is easily understood (p. 179). Synizesis is not, however, necessarily excluded from verse-closes of the form, and some certain examples are cited by Audouin (1.7., 69, 121, 228) from the diaereses of anapaestic verse, in which a resolved thesis such as túom would be extremely unusual, e.g. Ba. 1153 (p. 179, n. 2); ib. 1157 nilí sum. istúc iam prídem sc(i)ó (cf. Klotz, Grundz. 210, n.); ib. 1086; etc. Several possible cases (uncritically arranged) are also cited by Audouin from full anap. verse-closes, the best supported of which is Ci. 700 híc concílium f(u)ít (marked corrupt in ed. min.). We 7 8 1 Cf. Neue, III3, 435 ff., 465 ff., 472. 2 I reach at this point the same conclusions as Ahlberg, Procel. I, 92, but on wholly different grounds. It should be noted also that Pl. does not allow the shortened forms potin(e), viden (e) at the end of verses which close with an iambus (Seyffert, Berl. phil. Wochenschr. XVIII, 1577), and Dz. in his critical note suspects the one case of this kind in Terence (And. 476); we find the shortened form, however, in verses which conclude with a trochee, e.g. And. 299 atque aúdin? 7 are warranted in concluding that the synizesis forms are as well attested in such hemistich-closes and even in such verse-closes as we have a right to expect in view of the difficulties attending the identification of anap. verse. Whether the gen. sing. rei in Men. 764 (siét r(e)ì, bacch. tetr.) should be considered a case of synizesis in the close is very doubtful; the gen. sing. form is elsewhere always dissyllabic in Pl. (Maurenbrecher, Hiat. 156, n. 2; Leo, Forsch. 323 f.), but, in view of its monosyllabic use by Terence, it is by no means certain that it is an absolute dissyllable." (7) Specimen Verses. I may quote finally several verses which will serve to illustrate the various usages discussed in this paper. It will be observed that in general only metrical necessity or convenience leads to the employment of měí, měó within the verse: Cap. 740: Peric[u]lum vitae m(e) aé t(u)o stát perículó. St. 540: Dú(ae) ĕrant, quási nunc m(e)aé sunt. é(ae) ĕrant d(u)óbus núptae frátribús. Poe. 366: Méus ocellus, m(e)úm labéllum, méa salús, m(e)um sáviúm (i.e. s(u)áviúm). Tri. 329: Dé-měó: nam quód-tŭómst m(e)umst, ómne m(e)úmst autém tuóm. Cas. 614 M(e)am istúc transire uxórem ad úxorém tŭám. Summary. The results of the present study may be summed up as follows: Precisely that sequence of syllables and that position of the accent which produces iambic shortening in the case of vowels separated by a consonant has 1 Skutsch (Tépas, 131) needlessly rejects sc(i)o in the hemistich-close quoted above. On the other hand, it does not seem quite certain that O. Lat. synizesis can occur in a full anap. verse-close, such as that of the anap. oct., where it would be due entirely to the metrical accent. In the close of a full sentence we do not expect mos f(u) it, but rather f(u)it mås. 2 Acc. to Seyffert, Stud. Pl., 25 f., only twice does gen. rei fill any other foot than the last, viz. Ru. 487; Ad. 644. It is therefore somewhat similar to nihil, which never fills a whole foot in Pl., and never fills any foot except the last in Ter. and the metrical inscrr. Both these examples are instructive in their bearing upon the free admission of měó in the verse-close, and its rare use elsewhere. given rise to O. Lat. synizesis in the case of vowels which stand in hiatus. This synizesis does not occur in versecloses, since it is excluded from some closes by the conventions of the verse, and from others by the accentual conditions. Definite metrical proof of the extent of synizesis is afforded by word-groups like t(u)ám-rem, which show almost invariably a species of pretonic syncope. Finally, synizesis occurs most frequently in proclitic and enclitic words like the possessive pronouns or the substantive verb, which usually have little appreciable accent of their own, but it is also freely admitted in the case of those words which possess the ordinary intensity of tone, because these latter are themselves often subordinated in the sentence and placed beside words of still greater force and weight.1 From such beginnings as these, synizesis is free to develop even in the case of strongly accented words.2 1 For a fuller discussion of this point, see the supplement to the present article in Classical Philology, II, No. 5. 2 Addenda: P. 168. While synizesis is distinct from the hardening of i and u into full consonants, yet it is often the preliminary stage to such hardening and to the consequent loss of these sounds, cf. Corssen, II2, 754; Stolz, Müller's Handb. II3, 2, 32. P. 194, n. 2. See also especially Skutsch, Forsch. 136, n. I. P. 204, n. 2 (end). — Compare also O. Lat. hibus for his, ibus for is. 1 IX. The Title of Caesar's Work on the Gallic and Civil Wars. BY PROF. FRANCIS W. KELSEY, UNIVERSITY OF MICHIGAN. IT is noteworthy that the two editors, Nipperdey and Meusel, who in the last century contributed most to the criticism of the text of Caesar, adopted forms of the title of the Gallic War which are not only unlike but inconsistent with each other; and a third form is presented by Du Pontet, the editor of Caesar's text in the Oxford Bibliotheca, who combines elements that appear in the title as printed by the other two. Nipperdey considers that the correct designation of Caesar's Civil and Gallic Wars taken together is commentarii, each separate book being a commentarius, with the characterizing part of the title in the ablative; the title-page of his large edition has C. Iulii Caesaris commentarii cum supplementis A. Hirtii et aliorum, while at the beginning of the Gallic War we find C. Iulii Caesaris de bello Gallico commentarius primus. Meusel discards the word commentarius, plural as well as singular, adopting as the general title of the Gallic War C. Iulii Caesaris belli Gallici libri VII, and as the title of the first book C. Iulii Caesaris belli Gallici liber primus; he substitutes liber for commentarius, and has the genitive of bellum Gallicum in place of the ablative with de. Du Pontet uses commentarii as a common designation of the Gallic and Civil Wars, but liber of the individual books; at the beginning of the first book he has C. Iuli Caesaris commentariorum de bello Gallico liber primus. To trace the variant forms of the title down from the editio princeps to the present time might be of interest in another connection, but no light would thus be shed on the question what title Caesar himself gave to these works. The cause of |