II. ON THE RECESSION OF THE LATIN ACCENT IN CONNECTION WITH MONOSYLLABIC WORDS AND THE TRADITIONAL WORD-ORDER.'

PART I.

INTRODUCTION AND PROBLEM.

The present study was originally undertaken with a view to determining the probable cause of the frequent syllable-shortening which occurs in early Latin verse in connection with short monosyllables, i. e. sed illum3, sed aŭtem; in its present form, however, it will be found devoted chiefly to the preliminary task of determining the place of the grammatical accent in the wordgroups, ; —, ~ ~; and ~,. For it is evident that, after determining the place of the grammatical accent in the flexible tribrach groups sed ea, sed eni(m), etc., we shall be in a much better position to determine how far the accents séd illum, séd aŭtem and the like are due to their analogy.

1 This paper is an extension of a preliminary study on the same subject, an abstract of which appears in the Johns Hopkins University Circulars, Vol. XXIII. A study of the accent of the trisyllabic word-groups occurring in Terence has been published by the writer in Transactions Am. Philolog. Assoc. XXXIV (1903), pp. 60-103.

The scansion of ille, immo, quidquid, nēquis, hocin, etc., in some other cases as words of two morae, constitutes a wholly distinct problem. Upon this question the general principles laid down by Skutsch must be accepted as final; at the same time Skutsch's conclusions appear to me to call for certain modifications. Thus metrical theory does not require us to assume that a weak final short syllable which is neglected in iambic verse, i. e. illic, or a weak medial short syllable which is neglected in anapaestic verse, i. e. perdidi, suffers absolute 'apocope' or 'syncope'; it is sufficient to assume that such a syllable was greatly weakened in pronunciation. In general the colloquial iambic poets avail themselves of this license with a definite purpose, viz., in order to preserve unbroken the traditional word-orders, which exist in connection with the sentence-introducing pronouns and conjunctions, i. e. ille mé, unde tú, etc. A study of the question from this point of view will be published elsewhere, and it will be sufficient to point out here that, just as the existence of němpe is disproved by the non-occurrence of němp(e) with elision, so the existence of ecquis, etc., is disproved by the nonoccurrence of ecqu(a), siqu(em), etc.

The question of the recession' of the accent upon monosyllabic words which are closely connected both in sense and in pronunciation with some following word, is not a new one. Thus Ritschl in the sixteenth chapter of the Prolegomena, the chapter entitled 'de Accentu Logico', gives the rule (p. CCLVIII f.) that, so far as regards the accent, an iambic or pyrrhic dissyllable may coalesce with a preceding short monosyllable to form a trisyllabic word, i. e. pró equo, ét erus, quid agam. While Ritschl speaks prevailingly in this discussion of metrical coalescence, due to the fact that two short syllables are closely connected in forming a resolved arsis, there can be no doubt, I think, that he means to imply also a real coalescence in ordinary pronunciation. Valuable, however, as is Ritschl's study of this whole question, his treatment must appear to a critical student of the Latin accent at the present day as wanting at times in definiteness and precision. For the critical student must not only consider the question primarily from the view-point of actual coalescence, but must apply the necessary tests to determine the law of the accent in whole series of word-complexes and groups. Ritschl does not attempt to apply such tests, and it is doubtful whether he recognized the operation of a definite law in these processes. True, he not only holds, as has been already noted, that pró equo, ét erus, quid agam, etc., are accented as trisyllabic groups, but he correctly declares (Proleg., p. CCLXI) that the accent of de illo, ét iste, etc., is determined by the same principle, yet upon turning to p. CCLIII of the same chapter of the Proleg., we are perplexed to find a supposed example set éa (Trin. prol. 10: set éǎ quid húc) quoted in illustration of the thesis that monosyllabic particles of trite use and little weight are rightly placed extra arsim. If this view were correct, it would appear that the dramatists had known two forms of accentuation in the trisyllabic word-groups in question, viz. sét eă and set éă, which is far from being the case. For, with a single exception, which is only apparent (Cap. 329: ut éa-quae), the 27 cases occurring in the drama

2

1 For the sake of convenience the term 'recession' is here employed in general of the initial accentuation, ú, ~ ~; -, ~ ~, although it is not in all cases strictly applicable, see below p. 161.

Viz., Ba. 203; 472; Cap. 970; 942; Ci. 742; Ep. 265; 532; Men. 186; Mi. 346; 686; Mo. 160; Poe. 1015; 1265; Ps. 277; 1087; Ru. 1081; Tri. 330; 1168; An. 337; 837; He. 191; 334; Eu. 926; Ph. 480; 1015; cf. 1046; Titin. com. fr. 98; cf. Att. tr. fr. 432. Cf. also Seyffert's observation, Stud. Plaut., p. 27 n.,

tists in which eă, id ět, eam in etc., are preceded by a short monosyllable, all show the recessive accentuation sét ea, etc., and the reading set éǎ quid húc quoted by Ritschl from Trin. prol. 10 is only an unfortunate conjecture of Bothe's for the MS reading set ea húc quid introíerit, just as Fleckeisen makes a similar inadmissible conjecture sed éa servíbat, Phorm. 83, for the MS reading ea sérviébat.' The remarkable uniformity which appears to exist in the accentuation of sét ea and similar groups suggests an inquiry into the general tendencies of the republican accent.

THE REPUBLICAN ACCENT.

There are two periods of the Latin language, the accent-laws of which admit of being reconstructed even in minute detail. The first of these is the republican period, the accents of which are preserved in the dialogue verse of the dramatists; the second is the period of vulgar Latin which gave birth to the Romance languages and has left its accents embedded in the Romance forms. These two periods are separated at their furthest limits by an interval of nearly a thousand years, and the accentual changes which took place within this long period of time are numerous and in some cases far-reaching, yet so slowly is each single change of accentuation effected in the speech of a people 2 that it seems possible to trace with some precision the history of almost all the important changes in the Latin accent which occurred subsequent to the time of Plautus. According to the views which are held by most accentual scholars the Latin accent rested upon the initial syllable of words and groups until a time shortly before the beginning of the literary period (see the references given by Stolz in Müller's Handbuch II 2, p. 101 ff., 3 Aufl.). Thus in the time of Plautus the initial accent law was already superseded, but its effects were still very distinctly felt;

that the common formula "quis-hic (haec)-est?" is never accented on the second syllable.

1 Not admissible then are the accents marked by Hauler in his edition of the Phormio: prol. 8 et éăm; v. 284 ita éŭm, 605 si ab éŏ.

Thus in our own language in the case of many words derived from the Latin like confiscate, contemplate, demonstrate, etc., the contest between the Latin accent and the English recessive tendency has been going on since Shakspere's time and is not yet fully at an end, although in other words of this class, i. e. obdúrate, oppórtune, contráry, sepúlchre, etc., (Abbott, Shakespearian Grammar, §490) the Latin accent has long disappeared.

=

we shall best describe the republican accent then as strongly recessive in its nature and as seeking every opportunity of recession within the limits of the three-syllable law. The extreme recession of the accent is shown in three classes of words: I. In the inseparable composita which were usually written by the Romans as single words, i. e., éxplicat, réněgat, cómpater, désuper, etc. II. In the separable composita. Under this head I include first the recession in those separable composita which were not infrequently written by the Romans as single words, or whose parts were sometimes joined together by the ipév, i. e., cale facit, inter esse (intér erit), circum věnit, proptereos, inforo, quamdiu, intereá loci, anté volans, anté tulit (the last two written with the up by Donatus, Keil, IV 372, 2 ff. Schöll, De acc., p. 92; Diomedes, I 434, 36 ff.=Schöll, 1. 1., p. 95; Max. Vict., VI 193, 28=Schöll, l. 1., p. 98); and secondly the recession in those combinations which were only occasionally written together by the Romans, and which can only be called composita in the broadest, that is, in the ancient sense of the term; these latter include all the common phrases of the spoken language and also, to a large extent, as we shall see later on, the traditional Latin word-orders, i. e. sédenim (Priscian, Keil, III 93, 11f.), cúrita, quidego, quidea, nétime, nón potest, haúscio, certó scio, etc. III. In the regular accentuation of quadrisyllabic words beginning with three shorts (, proceleusmatic and fourth paeon words) upon the initial syllable, i. e. fácilius (cf. Stolz in Müller's Handbuch II 2, p. 101).

1 For a definition of composita, see the locus classicus in Priscian, Keil, II 177, 15 ff., and for a discussion of the separable composita see Priscian, II 183, 12; III 113,6; ib. 413, 14. The ancients included all prepositional phrases among the composita, as is evident from Charisius I 17, 3; Diom. I 436, 15; Dositheus VII 389, 4; ib. 409, 27, etc.

This and subsequent references to the Roman system of word-division as extremely fluctuating and uncertain (see, e. g., CIL. I index, p. 609 f.) are intentionally made. The practical necessity of adopting-often arbitrarilya fairly uniform word-division in modern texts of Latin authors obscures for most readers the whole subject of the Latin word-division and of the Latin separable composita, and is often misleading even to the critical student. The subject calls for a fuller exposition, but I can only refer here to my brief discussion of the Latin word-division in Trans. Am. Phil. Assoc. xxxiv 97-100, and to Eyssenhardt's reproduction, to some extent, of the variable worddivision of the MSS in his edition of Martianus Capella (Leipzig, 1866), viz. et enim and etenim, praeter ea and praeterea, etc.

In the Romance languages we find the Latin composita, both separable and inseparable, preserved in great abundance, and we observe further that many of the composita, which were separable in old Latin appear only as inseparable composita in Romance, i. e. Span. tambien=Lat. tam bene (tambene), Span. tampoco= Lat. tam paucum (tampaucum,) Span. ninguno=Lat. nec unus (necunus); cf. Corssen, Ausspr. II, p. 890. The coalescence in pronunciation of the more common monosyllables with the following word is regularly indicated also by the writing in literary Italian and by the doubling of the initial consonant of the second word, i. e. ellui=etlui, ebbene etbene, checcosa=quidcausa, etc. (Meyer-Lübke, Gramm. d. roman. Sprach. I, p. 508). But at this point all similarity ceases; for, as regards the form and accent of the composita, the Romance languages, as is well known, proceed from a period of thorough-going 're-composition' ('de-composition'); cf. G. Paris, Rôle de l'accent Lat. dans la langue française, p. 83; Meyer-Lübke, 1. 1., I, p. 495; Seelmann, Aussprache des Latein, p. 58 ff.; Lindsay, Lat. Lang., p. 199 f.; Stolz, Hist. Gramm. d. lat. Sprache I, p. 188. It is of course not to be denied that single cases of recomposition occur in the oldest literature; thus Stolz, Hist. Gramm. I, p. 187, quotes expars Turpilius, cited by Nonius II 138, 29 Müll.; requaereres Plaut. Merc. 633; conquaesivei CIL. I 551; conquaesiverit CIL. I 198, 38 etc., as well as late Latin inscriptional forms like reddedi CIL. VI 3, 20029; condederunt ib. 18850; but it was only at a very late period and only after a long conflict that these processes of disintegration and recomposition finally prevailed over the earlier tendencies towards composition and recession. Since, then, the Romance forms belong to a period of thorough-going recomposition, they commonly show the fall of the accent in the three classes of words just enumerated: I. In all inseparable composita in which the original composition was still felt,1 i. e., Late Lat. explicat, Fr. esploie; renégat, Ital. riniega, O. Fr. renie; compáter, Fr. compère, Span. compadre; desúper, Fr. desure; *ad própe (in old Latin regularly *ád prope, like the compound adverbs and prepositions: dé super, in super, dé foris, á foris, ád foras, póst modo; compare, for the last, Servius ad Ecl. I 30), Ital. apruovo, O. Fr. a pruef; *in fóris, Ital. infuori;

1 Only in cases where the original composition was no longer felt, was the recessive accent upon the prefix retained, i. e. collocat, Fr. couche, etc.; cf. G. Paris, 1. 1., p. 83.