most important contributions which has been made to the subject, and it becomes necessary to understand their full significance. We may summarize Prof. Humphreys' conclusions as follows: The earliest Roman poets made no effort whatever to produce agreement in the close of the hexameter, but, as a consequence of the marked uniformity of Latin accent, the structure of the verse caused agreement to occur in a great majority of the cases (e. g. of the cases in Ennius). "Consequently, in the course of time, when the ear became accustomed to this agreement, it appeared to be a property of the verse" (TAPA. IX 40), and later poets required that agreement should always occur in this place. Quite similarly, the predominant use of masculine caesurae generally produced disagreement in the first part of the hexameter; hence disagreement became the invariable rule in this part of the verse, and the masculine caesura alone was admitted as the principal caesura of the verse. Therefore the relation of ictus to accent in dactylic hexameters is to be regarded as an artificial relation, and as the result of usage or convention. In addition to these conclusions on the dactylic hexameter, Prof. Humphreys declares elsewhere (TAPA. VII 112) that the invariable or almost invariable agreement' which is found in the third foot

'One of the questions which one might have expected W. Meyer to discuss in his voluminous but incomplete work, Beobachtung des Wortaccentes, is whether this agreement really exists. Since, however, he has overlooked this problem entirely, I wish briefly to discuss the question in the form, Did Phaedrus seek consciously to place a tonic syllable in the third arsis? Fortunately this question admits of a definite answer; for Phaedrus most freely allows a spondee or anapaest in the 4th ft., subject to the conditions of the dipodic law, and hence his verse-structure admits térrōr[em] héredís tuí just as well as gaúdi[um] héredís tuí (4, 20, 18), and exiti[um] úxorís peténs just as well as cúbicül[um] úxorís peténs (3, 10, 21). According to L. Müller's ed. min. elision occurs in Phaedrus between the arses of the 3rd and 4th ft. (as in quóni[am] indignos, córpor[e] et, vér[e] ad, hómin[i] ut) 118 times in all, and in particular cases of the elision of trisyllabic words of the value (pectore, péctori) occur 28 times. On the other hand, cases of the elision of trisyllabic words of the value do not occur at all in Phaedrus; for in App. 2, 4, where the ed. min. gives quaecúmque indúlgens Fortun[a], the MSS have quaecumque (or, q. .. que) Fortuna, and both Müller in ed. mai. and Havet read quae cu Fortúna indúlgens; also in App. 21, 7, where the ed. min. gives cum círcumspéctans érror[e], the MSS have orrore or errore, and Müller in ed. mai. reads óre ita, Havet reads aequor, and Hartman ómni[a]. We have a right to assume that words like audirem are about as frequent in Latin as words like audiam and no more in demand in other parts of the verse than the latter; hence I conclude that Phaedrus

of the trimeter has arisen in the same manner.

These clear and

definite statements (especially De acc. momento, p. 2; TAPA. IX 40; VII 112) evidently afford a general law for the accentual development of all forms of Latin verse', and, to make the significance of this law perfectly plain, I venture to restate it in the following terms:- The Latin metrical cadences or verse-forms were originally constituted entirely without reference to accent, and solely in accordance with metrical laws; no sooner, however, had they assumed a definitive form in this way than they began to respond to the influence of the accent in accordance with a simple psychological law. In view of the uniformity of the Latin accentual system, the result of observing the metrical rules was to produce at certain points of almost every cadence, (1) agreement of accent and ictus in the great majority of cases, (2) disagreement of accent and ictus in the great majority of cases. Wherever this result is brought about, the Roman ear is quick to note the relation which usually exists and to require in the end, i. e. in the course of the historical development, that it shall always exist, that is, to require that the agreement or disagreement shall be made invariable; in other words, the Roman ear remembers the hexameter, Sapphic or trimeter cadences (rÓVOL 'tunes', 'musical modes') at certain points by the relation which

consciously sought to place a tonic syllable in the 3rd arsis and for this reason rejected the 28 examples of Fortun[a] which the verse-structure would naturally produce. We find also no cases of elision like éxiti [um], but 2 cases like cúbicŭl [um] (3,10, 21; 4, 7, 13). [Observe that fourth paeon words like cubiculum were wholly withdrawn from the 3rd arsis after 150 A. D., when the accent had fully changed to cùbícůlum, cf. A. J. P. XXV 152, n. 1.] may add that, in order to give the words in question their proper accent in the 3rd arsis, Phaedrus employs the rare latent caesura in proláps[am] 3, 15, 6; secrét[um] ib. 10, 11; lanífic[am] 4, 5, 5. A long monosyll., however, is freely treated by Phaedrus as independently accented in the 3rd arsis, as 3, 14, 4 quám deridendús, where composition would give quam-déridéndus; so 4, II, 14, et al.; cf. also 4, 12, 3 própter vírtut[em].

1 An altogether similar view of the influence of the accent has been put forward by the well-known English scholar, H. A. J. Munro, in an extremely able article in the Transactions of the Cambridge Philosophical Society, Vol. X (1864), pp. 374-402. Cf. especially p. 377: “I wish... to show that before the third century Latin verses of every kind, popular as well as learned, were written by quantity alone; that on the different kinds of metre accent had no direct influence at all; that however sometimes consciously, sometimes unconsciously, certain poets sought sometimes a coincidence, sometimes on the other hand a contradiction between the ictus metricus of the verse and the

they bear to the accentual cadence or melody, which is its simplest standard of measurement, its most familiar register. In those more numerous parts of the verse, however, where no usual relation was established, but the effect of observing the metrical rules was to produce sometimes agreement and sometimes disagreement, the metrical cadence remained wholly unchanged, wholly unaffected by the accent, so long as the Roman quantitative system endured.

I trust that I have now made sufficiently clear the real character of Prof. Humphreys' theory, which my own studies have led me to adopt with confidence, and, I may add, with a genuine sense of relief. For this theory appears to afford a genuine via media, and to neglect neither the historical development, as does the theory of Ritschl and Langen, nor the historical facts, as does the theory of Corssen and Meyer. It assigns some value and some potential influence to the Latin accent without, at the same time, destroying the genuine character of Roman poetry; it moves also among the legitimate ideas of ancient classical metric', instead of transporting us somewhat violently into the domain of modern German and English accentual verse. Above all, it does not involve us in the difficult and obscure psychology of 'harmonische Disharmonie', that is, it does not compel us to believe with accent". Cf. also p. 393: "Rhythm we have now seen was in Latin as in Greek quite independent of accent which had no direct influence on it whatsoever. But as quantity on which it rested was divided into various portions by caesura, pause and due arrangement of words, it well might be that in consequence of the limited range of the Latin accent it might gradually obtain a certain indirect influence over some parts of the hexameter, as of the iambic or trochaic: habit being all-powerful in this as in more important matters." It seems a curious accident in this branch of philological study that, while the scattered notes of Bentley upon Latin accent and quantity are widely known and often quoted, this study of Munro's, which is no less deserving of notice, should be generally neglected and apparently be unknown to the present generation of English scholars! I may add that the articles of Humphreys and Munro, which are wholly independent of each other and yet reach identical conclusions, serve admirably to supplement each other and consequently seem almost to exhaust all the important phases of the subject, leaving only minor questions of detail for future investigators. Humphreys offers a much more rigorous and strictly scientific proof, while Munro contributes to the question the wide range of exact knowledge and the great critical acumen which one naturally associates with the English editor of Lucretius.

1 For examples of the 'conventions' of Sanskrit metric, cf. Bergaigne et Henry, Manuel Védique, p. 38 f.

Ritschl (Opusc. II, Leipzig, 1868, p. XII) that the Romans sought agreement in one part of the verse for the pleasure which it gave them, while in another part of the verse, and for the sake of an abstract balance or contrast, they sought with pleasure ('suchten mit Wohlgefallen')—disagreement. In view of these merits, it seems not improbable that the theory of Humphreys and Munro needs only to be better known to meet with wide acceptance among metrical scholars; it is even probable that precisely this theory is already held by many Latin scholars in a somewhat indefinite form. In any case we may note the probable consequences that would be involved in its acceptance. We should be compelled to modify a part, at least, of those views upon Plautine verse, which have tended to prevail during the last twenty-five years, and we should not only have a solution of the accentual problem of classical verse, but should be measurably advanced towards a final solution of the closely allied problems, which relate to the real character of the earliest Latin verse and to the origin of the later rhythmical poetry.'

ELMIRA COLLEGE, ELMIRA, N. Y.

R. S. RADFord.

1 See already the careful dissertation of Dr. J. J. Schlicher, Origin of Rhythmical Verse in Late Latin, Chicago, 1900, which, if it does not give a final solution of this problem, at least goes very far towards making such a solution possible.

IV. THE INFLUENCE OF THE INFINITIVE UPON VERBS SUBORDINATED TO IT.

The standard Latin grammars usually recognize the fact that an infinitive may have the power of shifting the mood of a verb that depends upon it. Now, the lists of examples given to illustrate the usage, as one finds them in Holtze,' Kühner,' Roby, etc., contain many citations that are not to the point, as, for instance, examples of indirect discourse, attraction by the subjunctive, descriptive relative clauses, and other irrelevant matter. The habits and limitations of the construction have not been clearly understood, or, to say the least, have not been explained. The reasons that have been offered as explanation of the possession of this somewhat arbitrary power on the part of the infinitive are hardly adequate. Evidently a more careful study of the question is called for. I propose in this paper, in fulfilment of a promise made when discussing the construction of attraction by the subjunctive, to explain the connection between this and the allied constructions of indirect discourse and of attraction proper, both in origin and in usage, and to define the scope of its influence. For this purpose I have listed all the verbs, whether indicative or subjunctive, that depend upon infinitives in early Latin to the time of Lucretius, not inclusive, and also in representative portions of classical and later Latin.

3

Let us first see what verbs exert this influence upon a dependent verb. The historical or descriptive infinitive seems to lack

1 Syntaxis, II, p. 191 ff. His list, which, by the way, does not distinguish this construction from that of indirect discourse and of attraction proper, includes, for example, instances of the second person singular indefinite subjunctive: Cist. I, 1, 25; Merc. III, 2, 7 (552), and of the anticipatory subjunctive: Aul. 12; Epid. II, 2, 94 (277). The list is wholly untrustworthy. 'Grammatik, II, p. 789. His rule, which may be taken as a representative one, reads: Daher steht der Konjunktiv in allen Nebensätzen, welche in einem genauen und inneren Zusammenhange mit einem durch den Konjunktiv oder durch den Accusativus cum Infinitivo oder durch den blossen Infinitiv ausgedrückten Gedanken stehen. Roby's lists consist mainly of examples of oratio obliqua (1772-6).

3 Attraction of Mood in Early Latin, Chicago, 1904.