STOP Early Journal Content on JSTOR, Free to Anyone in the World This article is one of nearly 500,000 scholarly works digitized and made freely available to everyone in the world by JSTOR. Known as the Early Journal Content, this set of works include research articles, news, letters, and other writings published in more than 200 of the oldest leading academic journals. The works date from the mid-seventeenth to the early twentieth centuries. We encourage people to read and share the Early Journal Content openly and to tell others that this resource exists. People may post this content online or redistribute in any way for non-commercial purposes. Read more about Early Journal Content at http://about.jstor.org/participate-jstor/individuals/early- journal-content . JSTOR is a digital library of academic journals, books, and primary source objects. JSTOR helps people discover, use, and build upon a wide range of content through a powerful research and teaching platform, and preserves this content for future generations. JSTOR is part of ITHAKA, a not-for-profit organization that also includes Ithaka S+R and Portico. For more information about JSTOR, please contact support@jstor.org. THE RELATION OP THE ACCENT TO THE PYRKHIC IN LATIN VERSE By Albert Granger Harkness It is generally agreed that Latin accent is characterized by stress. Vendryes {Li 1 intensity initiale en Latin, Paris, 1902, p. 68) maintains that if this theory were true, and if Latin verse did not recognize this essential element of the language, the poetry of Rome must have been, absolutely artificial, and that it could not have been in any sense a national poetry. He allows that all testimony on this subject is in accord in affirming that the language of poetry and prose in Rome was one in character, that the verse of Plautus was composed for the people and in the language of the people, that the poetry of Virgil was in a true sense national and popular. He finds his solution of the difficulty in assuming that the accent of Latin was musical, like the Greek, and not one of stress. We must admit the general truth of his premises. No one can deny that the language would be spirit- less and colorless if it were deprived of its accent, and no one can deny that it would be artificial if it introduced a principle of accent entirely different from that of the spoken language. The solution of the difficulty is to be sought, not in abandoning the well-established theory of the stress accent, but rather in recog- nizing the part which accent plays in the structure of the verse. 2 I have sought to show (Trans. Am. Phil. Ass. XXXVI, pp. 82 ff. ) that when elision occurred between syllables separated by a sense-pause it conformed to definite laws of accent, and that, this type of elision, when the second element was short, was gov- erned by somewhat different rules from those which prevailed when the second element was long. This suggested that a 'The term "pyrrhic" is here used to denote two short syllables which form an integral part of a foot, as the resolved arsis or thesis in the drama or the thesis of the dactyl. 2 1 use the term accent to denote the recognized word-accent, often called the grammatical accent, and ictus to denote the verse-accent in contrast to the word- accent. Arsis is used to denote the strong or accented part of the foot. /■Classical Philology II, January, 1907J 51 52 Albert Granger Harkness special investigation of the relation of accent to the short syllables in Latin verse might yield results of some value. A special importance attaches to the relation of the accent to the short syllables in verse. One of the chief arguments of those who maintain that the word-accent is not an element in the verse- structure is derived from the law of Latin accentuation. The accent in the majority of cases falls on the long syllable, and thus in a large number of cases accent and ictus of necessity corre- spond. (Of. Oorssen Aussprache IT, pp. 972 ff. ; W. Meyer BedbacMung des Wortaccentes, Munch en, 1884, pp. 6 ff.). If it can be proved that there is a definite relation of the accent of short syllables to the verse, it is a specially strong argument that accent is an element in the verse-structure. We may observe in the Latin language a tendency to accent the first of two or more short syllables. After the penultima law of accent was established a word of the type ~ - ~ - was still in the time of Plautus accented on the first syllable. The accent on the penult of a dactylic word is not rare as compared with the accent on the penult of a tribrach. This irregular accent of the dactyl seems to reflect a tendency of the language and to be based on analogy. With this accent of the dactyl we may compare the secondary accent found in Plautus in the type _ z w a ^ .' The same tendency may be observed in dactylic word-groups which do not observe the rule of recessive accent so completely as do tribrach groups (Am. Jour. Phil. XXV, p. 406). 2 A word with two short syllables preceding the primary accent received a secondary accent on the first syllable. The existence of this accent may be traced throughout the history of the lan- II wish to guard against the idea that exdneratus, for example, represents the nor- mal pronunciation of the word. As there is a general tendency in Latin to accent the first of two short syllables, the verse-rhythm might take advantage of this tendency to modify the regular pronunciation. A tribrach word-beginning could not be accented on the second syllable. The verse-rhythm cannot go contrary to the laws of the lan- guage, but it may bring out its latent tendencies. In English compare doomed in prose and doomed in verse, or fdrefathers with the line : " The rude forefathers of the hamlet sleep." 2 A somewhat similar tendency may be noted in the Greek. The tribrach shows the tendency to accent the first syllable; the dactyl, according to Wheeler's law, shows the tendency to accent the penult (Brugmann Oriech. Gramm. s , p. 153). Eelation of Accent to Pykkhic in Latin Verse 53 guage from the period of the Saturnian verse to the time of the formation of the Romance languages (cf. Brugmann Vergl. Gramm. I 2 , p. 974; Lindsay Lai. Lang., p. 161). The impor- tance of this accent is minimized, or its existence utterly ignored, by those who do not admit that the word-accent is an element in the structure of Latin verse. With the recognition of this accent some of the main arguments used to disprove the harmony of accent and ictus tend to prove its existence. Havet 1 (M&irique grecque et latine*, Paris, 1896, p. 228) states as an argument against the intentional coincidence of accent and ictus at the end of the hexameter that, while the type mors sepelires has coinci- dence of accent and ictus, it is forbidden as the verse-close. It is generally admitted that monosyllables have a lighter accent than other words and, taking into account the secondary accent on the first syllable of sepelires, it seems reasonable to assume that this type of ending is avoided because the first syllable of the thesis would be almost as strongly accented as the arsis, and thus the usual end-rhythm would be interrupted. Havet states: "La coincidence existe dans consule inibit et dans tile animalis; or le premier type est permis, le second d6fendu." The first type is permitted because a short syllable immediately preceding a long syllable is without accent, as compared with the first of two short syllables. This type is not, however, a favorite one (cf. Rhein. Mus. XLV, pp. 402 ff.). It occurs .less than once in a hundred lines in Virgil and less than once in a thousand in Lucan. In the second type the first syllable of the thesis is accented and is also weighted by the elision; 2 in this case again the thesis would be made too prominent as compared with the arsis. 3 1 Havet's view is taken as representative because in his admirable treatise he pre- sents a very concise summary of the arguments of those who hold that in verse "l'accent n'est rien" (loc. (At., p. 227). 2 This theory assumes that the final vowel in elision was sounded and not wholly suppressed. For authorities for this view see Trans. XXXVI, p. 82, nn. 1 and 2. 3 This form of ending is not entirely avoided by Roman poets. We have thirty examples in Virgil (cf. Rhein. Mus. XLV, p. 242). It is chiefly confined to endings containing Greek words, or else it involves the elision of -que. This enclitic is espe- cially light in sound and it would add relatively little to the weight of the following syllable (cf. Trans. XXXVI, p. 93). This would seem to show clearly that the weight given to the thesis by the elision was one reason for the exclusion of this type. Kor- den (Aeneis VI, Leipzig, 1903, p. 427) says: "Noch nicht sicher erklart ist, aus 54 Albert Granger Harkness The only other argument used by Havet against intentional coincidence of accent and ictus in this part of the verse is the use of the type glaebaque versis. It is not agreed by all authorities that the enclitic changes the main accent of the word. Many grammarians maintain that the main accent of this type is on the antepenult. But granting that -que throws back its accent, we must still admit that glaeba retains its own accent. Verse appar- ently had the privilege of recognizing either of these accents as the main word-accent (cf. p. 71). This appears to have been one reason why this enclitic is so frequently used in poetry. Another theory of those who are opposed to the recognition of accent in Latin verse maintains that the special rules, which in reality produce coincidence of accent and ictus, are derived from the Greek, and accordingly have no reference to accent. Havet's brief summary of this theory begins as follows (loc. cit., p. 230) : " Si Virgile, qui a tant developpe la coupe hephtemimere, evite la coupe ennehemimere, ce n'est pas qu'il songe h 1' accent; c'est plutot par la meme raison qui fait que l'lliade et l'Odyssee, si riches en coupes au trochee troisieme, ont si rarement la coupe au trochee quatrieme." What ground can be found for compar- ing a caesura in the fifth foot of Latin verse with one in the fourth of Greek verse? There might be excuse for attempting to explain the disfavor shown in Latin for the hephthemimeral caesura as compared with the penthemimeral by reference to a corresponding avoidance in Greek verse of the fourth trochee as compared with the third trochee. In both Greek and Latin verse the fifth and sixth feet are regarded as forming a metrum which has a certain unity and which is governed by laws of its own. welchen Griinden nun auch schliessende Worte der Form ducunt argento, o bona mater, dextra rigat. amnem vermieden wurden." The recognition of the word-accent as an element in the structure of this part of the verse is a sufficient explanation for the avoidance of the specified types. This view, too, is in harmony with Norden's theory, as set forth on p. 391: "Das Bestreben, dem gegen Schluss des Verses fallenden Rhythmus durch prftgnante Worte ein Gegengewicht zu geben, ist so alt wie der lateinische Hexameter selbst." The falling rhythm would not be so marked in the types specified and thus would depart from the norm. If the poet recognized the fall- ing rhythm as characteristic of this part of the verse and adapted the thought to it, it would seem to be sufficiently clear proof that the accent was recognized as an element of the verse-rhythm. Relation of Accent to Pyrrhic in Latin Verse 55 The remaining part of this paragraph in Havet is as follows: "Si Plaute, dans certaines conditions, recherche les p^nultiemes breves (§273), c'est probablement pour des raisons de meme ordre que celles qui, dans d'autres conditions, font rechercher aux tragiques grecs des finales breves (§225)." Here again we have the rules applying to the final dipody compared with those apply- ing to other parts of the line. Furthermore, we are not told what reason there is for imagining that rules applying to the ultima of Greek words should apply to the penult of the Latin. 1 I would suggest that the special rules of Greek verse have to do largely with the unity of the verse and with the unity of its parts, even to that of the half-foot, while the Latin rules, relaxing something of the strictness of the Greek in this respect, are con- cerned especially with the relation of accent and ictus. The Greek hexameter wishes to give unity to the final metrum ; it accordingly avoids the fourth trochee. 2 For a like reason it avoids as a rule a sense-pause within the limits of the fifth and sixth feet. The Latin largely discards these rules, but gives unity to this part of the verse by means of coincidence of accent and ictus. 3 The Latin drama does not so definitely distinguish the last dipody from the preceding part of the verse by employing a dif- ferent relation of accent and ictus in the two parts of the verse. The dipody in comedy follows its Greek model more closely than does the final metrum in Latin hexameter. 4 Greek tragedy is 1 This, however, does not touch the essential question in regard to the coincidence of accent and ictus, i. e., why at the beginning as well as at the end of words we so rarely meet with the accent x -i x . Professor Radford (Am. Jour. Phil. XXV, p. 265, note) has well pointed out that " a supposed rule that the two closing shorts of a polysyllabic word cannot be used as a resolved arsis cannot explain the non-occurrence of ci'ilami|tatem." 2 This theory is not opposed to the rule that there shall not be a strong sense-pause after a fourth spondee unless elision occurs at the end of the foot. This latter rule merely implies that the last two feet shall not be too violently separated from the rest of the verse. 8 The difference between the Greek and the Roman method of handling this part of the verse may be seen by comparing the difference in feeling shown in regard to sense-pauses. Though all sense-pauses are rare in this part of the verse in Greek, yet the most common is that occurring after the fifth arsis. This sense-pause in the Latin is almost unknown. 4 That the Latin preserves the unity of the dipody in comedy in much the same way as the Greek, whereas the Latin hexameter does not observe the rules of the Greek in regard to the verse-close, favors Lindsay's view (Capt., pp. 362 ft.) that the accent 56 Albert Granger Harkness subject to more definite and exact laws than Greek comedy and hence affords a better opportunity to study the tendencies of this kind of verse. Here the dipody has a unity of its own and is subject to definite laws. Porson's law will illustrate this fact. 1 If the thesis of the next to the last foot were long and were fol- lowed by a pause, such as would necessarily exist before the pronunciation of a word or word-group of the minimum length of three syllables, 2 it would have greater prominence as compared with the arsis than was permitted in this part of the verse. Latin comedy also recognizes that the final dipody has a unity of its own and has its own special characteristics. 3 The rules regulating the use of tribrachs, anapaests, and dactyls evidently embody the desire to preserve the unity of the half -foot. For example the short syllable receiving the ictus is never the final of a word of more than one syllable (Havet loo. cit, p. 115).* The rules regu- lating the use of tribrachs, anapaests, and dactyls in Latin comedy concern the word-accent. 5 It is the purpose of this paper to consider the relation of accent to the two short syllables when they are the equivalent of a long syllable, as in the resolved syllable of Saturnian verse, or of comedy, or when they form the integral part of a foot, as in the of the last dipody is more in harmony with the prose accent than is generally allowed. The sentence-accent will be considered in connection with epic verse in my second paper. 'This law is well set forth by Havet (loe. cit., pp. 109 ff.). It may be summed up as follows : When a verse of iambic trimeter or catalectic trochaic tetrameter ends with a word of the length of three half-feet and the preceding syllable is a final, that syllable is short. 2 The view that there was a tendency to make a pause before a long word in Greek and Latin receives support from the syllable-lengthening in Virgil. When the length- ened syllable is not followed by a sense-pause, it is usually followed by a word contain- ing at least three syllables. This is true in all cases in which the succeeding word is a Greek word. 3 Jacobsohn (Quest. Plaut., Gottingen, 1904) treats of the characteristics of this part of the verse ; cf . Class. Rev. XX (1906), p. 35. 4 In trochaic verse the first syllable of the tribrach and anapaest has the ictus, and in iambic verse the first short of the tribrach and dactyl. In Greek comedy the first syllable of an anapaest is rarely the final of a word of more than one syllable (toe. cit., p. 125). For other similar limitations to the division between two words of two short syllables forming a half -foot, see Klotz Orundzuge, pp. 252 ff. 5 Of. Lindsay Oapt., pp. 70 ff . A summary of these rules is given in the Class. Rev. XX, pp. 31 ff. Relation of Accent to Pyebhic in Latin Vebse 57 lesser ionic. I trust that this will be useful in illustrating the unity of Latin verse, and that a general survey of this limited field may throw light on certain subjects which are not sufficiently clear when viewed by themselves. Thus it does not seem to me that we can decide in such a case as Most. 402, aedi\bus habi\tat, whether we are to regard the foot as a tribrach with the first two syllables forming the arsis, or as a dactyl with long arsis and resolved thesis, without reference, not merely to the older Satur- nian verse, but also to the forms of verse which followed the drama. 1 wish to establish the following rule for certain Latin metres: When two short syllables are the equivalent of a long syllable (not including the irrational long), the first of the two short syllables has an accent (primary or secondary), or if this is lacking, the second of the two short syllables has an accent as compensation. In certain kinds of verse the accent falls as a rule on the first of the two short syllables, as in the Saturnian verse and the iambic measures of Horace and Phaedrus. The drama is somewhat less strict in its adherence to this law. Epic verse comes under this rule only so far as concerns the second and third feet. Let us examine the Saturnian verse in its relation to this rule. Though there is no agreement among metricians with regard to the principle on which the verse is constructed, whether it is quantitative or accentual, yet there is substantial agreement in regard to certain characteristics of this verse, i. e., the division of the verse, hiatus at the division, and in regard to resolution. Havet (De Saturnio Latinorum versu) has collected all the extant material. Lindsay (Am. Jour. Phil. XIV, pp. 144 ff., 312 ff.) following Havet's order, gives a list of all the full lines. I shall observe the same order in referring to lines, and shall confine my references to the 123 lines which are regarded as most authentic (c£. loo. tit., pp. 321 ff.). Reserving the resolutions found in pyrrhic words to be con- sidered last, we find the following types of resolutions: 1. ~ ~ - 12, subigit omne Loucanam, opsidesque abdoucit; 17 examples. 2 w w w ^ 123, noctu Troiad exibant capitibus opertis ; 1 example 58 Albeet Granger Harkness 3, - w - 42, gondecorant saipissume comuiuia loidosque; 6 examples. 4, w ~ _ s 36^ /j, oc es £ factum monumentum Maarco Caicilio; 5 examples. 5_ w w „ - ^ 109, siciliensis paciscit obsides ut reddant; 1 example. 6, - ~ _ w - m j onerariae onustae stabant influstris; 1 example. 7. We have the following examples of resolutions in pyrrhic words: 16, quibus set in longa licuiset tibe utier uita ; 33, Romam redieit tri- umphans. ob hasce res bene gestas; 38, Bene rem geras et ualeas; dormias sine qura; 44, utei sesed lubentes bene iouent optantis; 52, neque tarn te oblitus sum, Laertie noster; 67, simul ac dacrumas de ore noegeo detersit. In 30, semol te orant se uoti crebro condemnes I prefer the scansion with elision rather than resolution. This is in accord- ance with the view of L. Mtiller [Der sat. Vers 1885, §§ 42, 43), and Leo {Der sat. Vers 1905, p. 70), who may be taken as representatives of the quantitative theory, and it is given as an alternate scansion by Lindsay (loc. cit., p. 315), a representative of the accentual theory. Mtiller regards the resolutions found in pyrrhic words as exceptional and, when possible, removes them from the text. In 38 he would read ben for bene and in 52 he prefers the reading neque enim to neque tarn. There does not seem to be sufficient reason for regarding this type of resolution with suspicion. It appears, however, to have been used with certain limitations, though our material is too limited to allow us to draw very definite conclusions in this respect. It will be seen from the lines cited above that all these resolutions, with two exceptions (16, 44), 1 the latter only an apparent exception, occur either at the beginning or end of the verse, and in these two positions exceptional usages are found throughout the history of Latin verse. In 44 beneiouent is written in the inscription as one word. It is to be regarded as one word, or as the phonetic equivalent of one word, i. e., a word-group accented as b6ne-iouent. Thus in this prosaic verse of the Faliscan cooks the ictus falls apparently 1 In 16, one is tempted to read licuiset to avoid the two resolutions in the same hemistich. Synizesis, however, does not occur in later Latin poetry in an anapaestic sequence (Trans. XXXVI, 168). It is possible to consider this case an example of the total suppression of u so far as the verse is concerned (Lindsay Lat. Lang., p. 268; Trans. XXXVI, p. 171). Relation of Accent to Pyrrhic in Latin Verse 59 on a syllable of secondary accent. quibus (16) is the most independent and disconnected of these pyrrhic words, but the relative is an especially weak word and is regarded as an enclitic by the Roman grammarians (cf. Lindsay Lat. Lang., p. 167). In the Greek and Latin classical drama the arsis is not often formed by a pyrrhic word (cf. Christ Metrik d. Griechen u. Homer 2 , p. 56). In Saturnian verse pyrrhic words do not seem to have received a marked sentence-accent and they apparently conform to the same general rules as apply to this class of words in the drama. These seven types include all the resolutions found in the 123 lines to which we have referred . We accordingly see that certain resolutions are strictly excluded from this verse. Thus they do not occur in the last two syllables of dactylic and tribrach words, nor does the final syllable of a word form the first half of a reso- lution. The rule in regard to secondary accent is as strictly observed as the rule in regard to primary accent. Thus the type - ^ w _ w is excluded. 1 The usage of Saturnian verse in respect to the resolved syllable is more closely related to the usage of the drama than to that of epic poetry. It is, however, more strict than the drama in observ- ing the relation of the resolved syllable to the accent. This is a phenomenon of the verse which should be taken into account by the advocates of the quantitative theory who do not admit that the word-accent is an element in this verse and who accordingly will not even recognize such a harmony between accent and ictus as is to be found in comedy. It seems difficult to imagine that this verse, so calm in tone and so simple in form, was in its method of delivery so far removed from the spoken language. When the Saturnian verse fell into disuse, it was replaced in epitaphs not by epic verse but by prose. 2 iCf. Miiller loc. cit., pp. 49 ff. ; F. Leo Der sat. Vers, Berlin, 1905, p. 47. Leo scans (p. 58) fulse vtrom, and classes with this hemistich two other cases which he regards as somewhat doubtful. Klotz (loc. cit., p. 256) regards the above citation as an example of resolved arsis. There does not seem to be good ground for regarding this as a case of resolution, as there is one syllable less than the normal number in the hemistich. 2 If we are to class as Saturnian verse such a variety of rhythms as are given by Leo in his learned monograph (pp. 69 ff.), it would seem necessary to revise our conception of Saturnian verse and to consider it as a generic term, denoting a variety of metres, 60 Albert Granger Harkness Again, this relation of the resolved syllable to the word-accent is an argument against the extreme accentual theory which does not make quantity one of the essential elements of this verse. There appears to be no wide separation between the principles on which Saturnian verse is constructed and the quantitative verse of a later period. It would seem clear that whatever be our theory with regard to this verse, we must recognize the accent of the resolved syllable as an element in the rhythmical effect. 1 Though we do not find in the case of the drama the same wide divergence of view in regard to the fundamental principle on which the verse is constructed, we find here also two hostile camps with hardly a prospect or hope of reconciliation. Those who, like L. Miiller, hold the view that there is no intention on the part of the poet to introduce harmony in any respect between the accent and ictus, maintain (loc. cit., p. iii) that W. Meyer has satisfactorily demonstrated this fact, while the advocates of the accentual theory feel as confident that Langen has completely refuted the arguments of Meyer (Lindsay Capt., p. 361, n. 2). Not to refer to what has been accomplished by such writers as Bentley, Ritschl, Christ, and Klotz, it seems safe to say that Lindsay and Radford, to take examples from writers in our own tongue, have demonstrated that there is in certain cases at least an intentional effort to bring about harmony of accent and ictus. For example, Lindsay has proved this for words of the form ~ ~ « s {Philologus LI, pp. 364 ff. ; Brugmann Vergl. Gramm. F, as does the term lyric. It is difficult to convince oneself that the two following schemes belong to the same measure (p. 70): He points out (p. 31) that, if the accent on the final syllable is not pleasing to our ear, yet the same rhythm occurs in the drama. It would have been helpful in this con- nection if he had illustrated the melody of such a hemistich as (p. 69) inserinuntur, jc. *, j. j. ^, by giving us parallels from later literature, for some readers will be forced to confess that such rhythms also "unser Ohr nicht leicht und erfreulich beruhren." iln comparing notes with Professor Radford I find that he has independently arrived at the same conclusion in regard to the relation of the accent to the resolved syllable in Saturnian verse as that here set forth. It is a pleasure to learn this fact, as it is an added argument for the correctness of the theory. Relation of Accent to Pyrehic in Latin Vekse 61 p. 973) and Radford in the case of recession of accent in certain word-groups (Am. Jour. Phil. XXV, pp. 147 ff.). It is not my purpose to review the arguments in favor of the theory that the accent is an element in the rhythm in comedy, or even to state what may be claimed as proved in regard to the harmony of accent and ictus, but to direct attention to the relation of the accent to the resolved arsis and thesis. An interesting paper on this subject has recently been published by Professor Exon (Class. Eev. XX, 1906, pp. 31 ff.). He attempts to estab- lish the following law: "Except in the first foot of a colon, a resolved arsis or thesis must, in dialogue metres, begin in a syl- lable which would bear in prose the primary or secondary word- accent, or a sentence accent." 1 That which the author of the paper recognizes as the crucial point is the "ease in which a resolved arsis seems to begin in a short final syllable" (p. 34) . These final syllables he regards as metrically lengthened. This 1 Having in mind the general views on Latin verse presented by the standard authorities, one is startled by the first sentence of the article : " No satisfactory attempt has ever been made to account for the existence of the strict rules observed by Plautus, and the early Roman dramatists generally, in the use of dactyls, anapaests, and proce- leusmatics in dialogue verse." Christ (loc. cit., p. 55) says: "dass die lateinischen Komiker, .... fast durchweg die Verse so bauten, dass die erste der stellvertretenden Kurzen den Accent hat," and more to the same effect ; cf . p. 59. I shall not attempt to criticize this paper in detail, though one is often tempted to ask just what is the meaning of some of his expressions. For example he says (p. 32) : " A long syllable accented in prose may not be shortened in dialogue under the law of breues breuiantes." In the next sentence in speaking of mdlestus he refers to the ictus as "canceling the word-accent." To the uninitiated he appears to profess one belief and to act on another. His first statement of the rule seems to be in harmony with the accent theory advocated by Lindsay, Skutsch, and Ahlberg, and his application of it seems to be in accord with the ictus theory of Klotz, Seyffert, and Leo. The chief defect of the paper seems to me to be that the author is not satisfied with maintaining the general truth of his proposition, but that he does not wish to admit any exceptions or to recog- nize any principle as capable of modifying the one under consideration. He is thus led to regard the secondary accent in such words and word-groups, as commoditatem and propter-amorem, as falling on the second syllable rather than on the first which is its recognized place. (Of. Lindsay hat. Lang., p. 161.) Professor Radford also communicates to me the following criticism of this accen- tuation: "It is both contrary to the old Latin principle of recession (not to mention the earlier principle of initial accentuation), and is disproved by the absence of pro- celeusmatics like sed exonerates, et obsiqueratur (cf. Trans. XXXVI, p. 195, note 1). Further, such old Latin forms as surrupuisse (Plautus), consoluerunt (S. C de Bacch.) with their weakened short vowel, clearly point to the original initial accentuations, surrupuisse, consoluerunt; the later accentuation that has resulted from such an original is clearly not consoluerunt, but consoluerunt. 62 Albert Granger Harkness theory is by no means a new one, though no reference is made to the arguments in its support by such scholars as Fleckeisen, Use- ner, Koch, Crain, and Buchholtz (cf. Christ loc. cit, p. 60). This theory was afterward modified by Bticheler (Grundriss d. lot. Decl.), but even in this modified form it has been discarded by practically all metricians. 1 These facts seem also to be overlooked by Professor Sonnenschein in his suggestive article on "Accent and Quantity in Plautine Verse" in Class. Rev. XX (1906), pp. 156 ff., when he refers to the changes which will be wrought in our theories of Plautine verse, "if Professor Exon succeeds in establishing his scansion of cases like aedibus habitat.'''' Professor Exon begins his argument in favor of this lengthen- ing with the statement that the succeeding syllables are always of such a kind that lengthening is at least possible. We, find how- ever, many such endings in iambic and trochaic metres as Merc. 540, puer est ille quidem, stulta, and Merc. 1008, filius apud nos tuos. If he excludes such cases from the "dialogue metres," it would seem impossible to maintain that this type, which often occurs in dialogue and which so closely resembles in character the cases under consideration, was treated by the poet on an entirely different principle; for here lengthening is impossible. Continuing our author says: "First, the next following syllable may be long, as in finger e fallaciam.'" He does not mention the fact that in all his illustrations (the last three should be omitted as they do not relate to the case in hand) the apparent short arsis occurs before the final dipody (cf. Lindsay Capt., p. 42; Jacob- 1 As representing the present view which approaches nearest to Exon's we may mention Havet (loc. cit., p. 140) who regards this short final syllable in the arsis, which was necessarily followed by a brief pause, as producing rather the effect of a long than the effect of a true short syllable ; or the view of Brix (Mil. Glorfi 1901, note on line 27) : "EineandereBeurteilung ist moglich resp. notwendig in den Fallen, wo Personen- wechsel eintritt, also etwa syllaba anceps angenommen werden kann." Again when Professor Exon adds (p. 36) that the acceptance of his theory of the lengthening of the final syllable in the arsis "would even enable us to introduce a little order into the scanning of Saturnians," he overlooks the fact that Reichardt ("Der saturnische Vers in d. torn. Kunstdichtung," N. Jahrb.f. Philol. XIX, Supplbd. 1893, pp. 207 ff.) maintained this theory as an important discovery of his own. His theory, however, was not hailed with great enthusiasm by the world of scholars (cf. Gleditsch's review in Bursian's Jahresber. CII, p. 44, and the reviews there cited). This theory of lengthening is accepted by Leo (loc. cit., pp. 20 f.) for certain positions in the verse. Relation of Accent to Pyrrhic in Latin Verse 63 sohn, pp. 5 ff.) . Even if we grant that the syllable in this posi- tion is lengthened, the author would- have gained but little for his main contention, for it has long been known, or, as he says: "It has long been suspected that the point in the line that immediately precedes the final iambic dipody had some special characteristics." He next adds that which to me is quite unintelligible: "But far more significant than such instances is the fact that, wherever a syllable can be said to remain short in arsis, it must be followed not merely by one, but by two short syllables, capable of forming a resolved thesis." This very fact is interpreted by Klotz (loc. cit., p. 265) — and it would seem justly — as evidence that the final syllable is not lengthened. If instead of the two short syllables which are usually found after the final, there was but one remain- ing to form the thesis, this would be good ground for considering the syllable in question as lengthened. This is followed by an interesting discussion of Luchs's law in relation to endings of the type ~ ~ | - - | - - . Though we grant his contention that this comparatively rare form of verse-ending is not allowed in comedy, even though it is usually recognized as legitimate, and that the final arsis before the dipody is length- ened, it would be but another proof of the exceptional character of this part of the verse. Such lengthening would not be any stronger evidence that the syllable in question was lengthened in the hundreds of cases which occur in other parts of the verse, than the exceptions, which the author admits occur in the first foot, would be proof that they existed in other parts of the verse. 1 1 While Sonnenschein (loc. cit., p. 159) does not regard Exon's law as proved, he points out how greatly our view of early versification must be modified, if it should be established. After granting that the quantity of the arsis may be an essential element in the verse, he still claims that Plautine verse is semi-quantitative and that the scheme for the iambic verse, for example, may be represented as follows : x-x-|x-x-| x - wx | . In this formula x stands for a syllable of indeterminate value. He adds further, apparently in support of Exon's theory : " It would even be possible to argue that in semi-quantitative verse there is no a priori objection to a rise (arsis) being occasionally formed by a short syllable, the quantitative defect being supplied in some cases by a slight pause after the syllable in question." It is to be noted that in this case he does not regard the defective quantity of the arsis as helped out by accent, but by the pause which follows. In other words, Sonnenschein seems unconsciously to admit that the quantity of the arsis is after all the essential element. In fact can we even admit that substitution of the long syllable for the short in the thesis implies that quantity is not an essential element of the verse ? On this principle the iambic 64 Albekt Geangee Haekness We may sum up the relation of the accent to the resolved arsis and thesis as follows: As a rule the resolved arsis and thesis receive an accent on the first of the two short syllables, except when the first short syllable of the arsis is a final, and in this case the second short syllable has an accent. Other exceptions to the general rule that the first short syllable receives an accent are more common in relation to the secondary accent of words of five syllables or more (e. g., exoneratus, commoditatus) than in rela- tion to the primary accents. For example, these exceptions are more common in the word-type - ~ - - a than ~ ~ ~ ( Am. Jour. Phil. XXV, p. 267 ) . Again, exceptions are more common in the lyric measures, especially the anapaestic, than in the dialogue measures, and more common in the first foot than in the other feet. I now turn to hexameter verse 1 to consider the relation of the accent to the two short syllables in the dactyl. It might seem most natural to begin with Ennius, but though he showed won- derful genius in adapting Greek verse to the conditions of a language widely different in character, yet the task of necessity remained incomplete. Lucretius carried forward the work so auspiciously begun. The tendency to give a distinctively Roman character to hexameter was interrupted by Catullus, who was under Alexandrine influence. Virgil openly broke- with the traditions trimeter of Horace would be only partially quantitative, for here every other thesis may be long. Other lyric measures also allow substitutions. In the first Asclepiadean we have as substitutes , — w w, — w, — . These substitutes differ from those of the iambic verse. These may be called fixed substitutes, while those in iambic verse may be called possible substitutes. The character of the verse in both cases depends on the substitutes used, and the quantity of the syllable remains an essential factor of the verse. The inference to be drawn from the fact that substitutes apparently varying in quantity are allowed both in early and later verse, is the well-established fact that syllables are not absolutely fixed in quantity but that their actual length is to a certain degree determined by the verse-scheme in which they are used. Thus a dactyl sub- stituted for a trochee is in reality shorter than a dactyl substituted for a spondee. 1 The caesura is regarded by some as the most essentia] feature of hexameter ; and its laws are thought to be sufficient to explain the general verse-structure and even to account for the differences between the Latin and the Greek hexameter. These advocates of an over-rigorous doctrine of the caesura are led into all manner of inconsistencies, and one is almost forced to the conclusion that they imagine that verse is constructed with reference to the caesura and not for the purpose of expressing thought. It is to be remembered that the caesura in hexameter is free and shifting, and that it does not determine the character of the verse to the same extent as does the fixed caesura of some forms of lyric verse. Relation of Accent to Pybkhic in Latin Veese 65 of his Roman predecessors. It is one of the main objects of his work to give a Greek background to his Roman scenes. Every- where there are reminiscences and echoes of the Greek past. He interrupts the Latin rhythm to introduce Greek words and Greek rhythm, as Dante breaks the flow of his verse by interweaving Latin lines, or as Milton does by quoting phrases from the Bible. 1 Ovid did not possess the genius to blend in so masterly a way elements so different in character. Though the influence of Greece may be traced in his verse, it is confined within comparatively narrow limits. It is not till we come to Lucan that we find the Romanizing of hexameter verse carried out to its logical conclu- sion. The Pharsalia affords us the best opportunity to study the hexameter in a strictly Roman garb and practically free from Greek influence. Though we may not possess the "Lucanian proclivities of those modest scholars who," as a writer in the Quarterly Review (CXXVII, p. 129) tells us, "committed their cherished Pharsalia to memory from end to end," yet we must admit that Lucan possessed real genius and that he has at times used the Latin language with masterly force. We may not agree with Voltaire that Lucan owes nothing to his predecessors, neither his beauties nor his defects, or with Merivale, who is "almost tempted to imagine that Lucan had never read Virgil," yet a study of the parallel passages which Heitland (Haskins Phar- salia, with Introduction by Heitland, London, 1887, pp. ex ff.) has so carefully collected to prove the indebtedness of our poet to Virgil, tends to confirm the view that Lucan, though an admirer of Virgil, was not even under his spell. Every other writer of epic verse during the Silver Age is a servile imitator of the author of the Aeneid. Lucan's avoidance of Greek rhythm and his strict adherence to law, though they may be defects in the poet, render him especially useful in a study of the verse- structure. Before turning our attention to the relation of the accent to the short syllables of the dactyl in the second and third feet, let 1 It is this distinct purpose of the poet which justifies the introduction of the Greek rhythm, especially in connection with Greek words. Judged from this point of view the usage of Virgil cannot be condemned as an error of judgment (cf. Havet, p. 55). 66 Albert Granger Harkness us note the general rules of the relation of the accent to the ictus in the different parts of the verse. The first foot in Greek and Latin metres usually allows greater freedom than the other feet. This applies especially to the accent of the hexameter. In the second and third feet we have conflict of accent and ictus. The fourth foot forms the connecting link and admits either harmony or conflict. The tendency to harmony in the fifth and sixth feet increases from the time of Ennius (cf. Trans. IX, pp. 39 ff.). In Lucan there is almost complete uniformity in this part of the line. To make clear the contrast in accent between the second and third feet on the one hand, and the first and fourth on the other, it is only necessary to call to mind that a dactylic or spon- daic word or ending cannot form the second or third foot, but it may form the first or fourth foot. In Virgil I have noted some ten lines in which a dactylic word or ending forms the second foot. The third foot is never thus formed in Virgil, for we are not to consider as such nescio qua, Georg. iv. 55 ; this is used as a word-group, united under one accent. 1 There are no exceptions in the case of spondaic words or endings. 2 In Lucan I have not noted any exception to the rule that the second and third feet shall not be formed by dactylic or spondaic words or endings. Omitting for the time the cases in which trochaic words or endings are used to form dactyls, as in verte manus, the rule of accent in the second and third feet, as exemplified by Lucan, is as follows: The accent falls on the thesis, except in a few cases of elision. Examining the rule first in reference to the long syllables, we find that the thesis always has an accent, primary or secondary. 3 1 With this we may compare the frequent shortening in comedy of -o in nescio, a shortening which is regular, or normal only in the word-group nescio quis (Klotz loc. dt., p. 96). 2 A foot formed by inter in a word-group like inter se, united under one sentence- accent, is not treated by the poet as a spondaic word (cf. L. Mailer R&m. Melr?, p. 466). 3 We may recognize a light secondary accent on the first syllable of the types _£ a si , j. 41 w j£ . Some metricians regard this secondary accent as an element in the verse of comedy (cf. Lindsay Capt. p. 358; Glass. Rev. XX, p. 157). For hexameter it is only necessary to assume that this initial syllable was more prominent than the final of a word of two or more syllables. This long syllable preceding the main accent was doubtless lighter in accent than the first syllables of the type J- — 1! •*., ± — C ~ s. In Lucan the thesis is never the second syllable of the type — > •" <" — , though this occurs in Virgil, as v. 591, and in the fourth foot in Lucan. Relation of Accent to Pybbhic in Latin Vebse 67 When the thesis has only a secondary accent, the arsis is usually the final of a word of two or more syllables, but if the arsis is a monosyllable, it is without a sentence-accent; and it is never a noun. The monosyllables most frequently used in the arsis, when the thesis has a secondary accent, are the sentence-enclitics, especially est, which rarely ranks as an independent word and resembles rather the formal enclitics, such as -que, -ve, and -ne. A noun may be used in the arsis, if followed by the primary accent of a dissyllable or trisyllable, as i. 45, res acta est. If two monosyllables form the foot, the one in the thesis has the more prominent sentence-accent, as Lucan x. 432, donata est nox; i. 360, | et ius \ est; v. 117; 664; vii. 374; x. 511. The following are somewhat exceptional: i. 349, dat, qui; 681, hie, o; x. 410, dant. Pro. In these cases the monosyllable in the arsis is closely associated in thought with the preceding word and forms practically the final of a word-group. At the end of the word-group the monosyllable would not he prominent. Accordingly we never find a monosyllable lengthened in arsis. 1 The monosyllable at the beginning of the next clause would have a certain prominence from its very position. This tendency is seen in the lengthening of syllables at the beginning of a line, as is frequently the case in Homer, and in the prominent accent at the beginning of the line in comedy, which results in iambic shortening. This same principle is illustrated by ix. 1032, hoc non, and 1062, non his. The arsis in both cases is closely united with the preceding word — scalus hoc, nee non — while the theses are more closely associated with the following words. The rule that the theses shall have the more prominent accent has an appre- ciable effect on the word order. Thus to observe this rule of the verse the natural word-order may be reversed, as in viii. 636, ius | hoc, and x. nox | haec. In case of elision the accent may fall on the arsis. Let us examine the rules which govern Lucan's verse in this respect. In the first foot there are frequent examples of elision in the second syllable of the thesis of the dactyl, as well as in the first i Mailer Rem. Metr?, p. 396. 68 Albert Granger Harkness syllable of the thesis, i. 169, vomere et. This elision also occurs in the fifth foot in the dactylic word-ending, x. 298, consurgere in. When the elision is in the second short syllable of the first foot the second element of the elision is not an accented word. There is one apparent exception in the use of ait, as in vi. 230, parcite ait. ait is a subordinate sentence-word and does not have a prominent accent. 1 Neque enim or nee enim occurs in the second, third, and fourth feet. That in this word-group neque and nee were hardly distin- guishable is supported by various evidence, but we need not go beyond the MSS authority of Lucan. Some of our best Lucan MSS give nee in many cases, and in some of these instances our standard editions accept the reading nee. See Francken and Hosius 2 , i. 632; iii. 464; ix. 42, 243, and the MSS authority quoted in connection with these lines. 2 In the second, third, and fourth feet elision never occurs in the second syllable of the thesis. Lucan extends the restrictions which apply to the second and third feet also to the fourth, to a greater extent than do the other Roman poets. The careful avoidance of this type of elision in this part of the verse is in harmony with our general theory of the accent, i. e., that the first of two short syllables shall be the more prominent. To weight the second syllable of the thesis with elision would tend to make it more prominent than the first. The elisions in the first four feet are distributed as follows: in the first, 156; in the second and third, exclusive of 15 examples of neque enim, 156; in the fourth, 418. In the first foot the main sentence-accent may fall on the arsis, as in i. 169, vomere et, or on the first syllable of the thesis, as in 173, hide irae. Either of the two words forming the elision 'The same exceptional use characterizes inquit (Trans. XXXVI, p. 86, n. 2). Trampe (De Jjucani arte metrica, 1884, p. 20) rightly regards ait as an atonic word, but when (p. 30) he wishes to show that writers other than Lucan have an accented syllable after the elided dactylic word, he classes erat as an accented word. 2 The fact that Lucan avoids such combinations as neque erat, neque ego, neque adhuc, which are used by Ovid, is most naturally explained on the assumption that in these cases the final vowel of neque was not so fully suppressed. The usage of Lucan in respect to neque seems to throw some light on the mooted question of the suppres- sion of the vowel in elided -que. Kelation of Accent to Pyrrhic in Latin Verse 69 may be a subordinate sentence- word, or an emphatic word; or both may be subordinate, or both prominent. The two words may be closely connected in thought, or be separated by a sense- pause. The only restriction in regard to the words forming the elision in the first foot seems to be that the second shall not be a long word. In comparatively few cases does it extend beyond the second arsis. In the latter part of his work Lucan does not observe this restriction so fully, and uses words containing as many as 8 morae (cf. ix. 357, 378). The 156 elisions in the second and third feet may be classified as follows : I. Those in which the first element of the elision is not an inseparable enclitic, -que, -ve, -ne. There are 34 elisions of this type. They may be arranged as follows: 1. The thesis is formed by the first two syllables of a tri- syllable. a) The thesis is formed by a noun: i. 244, ut notae fulsere aquilae; ii. 385, submovisse hiemem; iii. 427, nam vicina operi; 578, hi luctantem animam; 623, effugientem animam; 751, festinantem animam; v. 354, en, quantum Fortuna umeris; vi. 714, descendentem animam; 732, infelicem animam; vii. 769, inspirasse animas; 795, et lustrare oculis; viii. 157, quod, sub- missa animis; 266, ingentis praestate animos; ix. 46, ancipitis tenuere animos; 96, iuris habete animos; 407, incendit virtute animos; 949, exegere hiemem; x. 23, membra viri posuere adytis; 108, excepere epulae; 155, infundere epulas; 412, Thessaliae subducta acies; 422, et districta epulis; vii. 576, promovet ipse acies; ix. 763, mors erat ante oculos. b) The thesis is formed by an adverb: ii. 552, opposuit. Parthorum utinam; vi. 768, sit tanti, vixisse iterum. 2. The thesis is formed by the first syllable of a dissyllable. a) The thesis is formed by a noun; iv. 565, exegere enses; vi. 542, effodisse orbis; x. 145, inposuere orbes. b) The thesis is formed by an adjective, i. e., omnis: vii. 284, Romanos odere omnes; viii. 536, adsensere omnes; vii. 650, eminus, unde omnis. Ipse acies, ante oculos, and unde omnis are regular in regard 70 Albert Granges Haekness to the accent of the thesis but depart somewhat from the type of the others. 1 It will be noted in regard to the list quoted that the examples are similar in character. In each case the two words forming the foot are closely related in thought. This fact is of special signifi- cance, since the tendency in hexameter and in the more elevated styles of poetry in general is to separate words grammatically belonging together. The formation of word-groups, such as are in harmony with comedy, is as a rule avoided. The tendency in the case of the examples under discussion, in which the words are united owing both to the thought and the elision, would be to lighten the accent on the first of the two words, i. e., the arsis, and to have the main sentence-accent fall on the second word, which is usually a noun. This conclusion receives support from a similar tendency to be observed in a different type of elision (Trans. XXXVI, p. 93). In the majority of cases the second word is a noun. Nouns are found to predominate where a specially strong accent is desired, and avoided where a strong accent is a disturbing ele- ment, as, for example, in elision when the elided syllables are separated by a sense-pause (Trans., loc. cit., p. 98). Even when the word containing the thesis is not a noun, it will be seen that it has especial emphasis, as in the case of the adverbs quoted under 16). It may also be noted that the large majority of these examples occur in the last half of the Pharsalia. I think we may observe in the latter part of the poem a tendency on the part of the poet to emancipate himself somewhat from the thraldom of law to which he felt subject in the earlier books. Trampe (loc. cit., p. 54) 1 Book iv. 811-24, are pronounced spurious by Francken. In 811 we have a quibus omne aevi senium, which is somewhat exceptional in character, and in 824, concessa est f emere omnes. That this kind of elision which occurs so rarely in the first half of Lucan should be found twice in this suspected passage is an additional reason for rejecting it. In x. 136, Haskins and Francken read discvhuere illic reges. This case is the most exceptional to be found in the text of our author, as illic would not have a prominent sentence-accent (Lindsay loc. cit., p. 167. 5). I feel confident that toris is the correct reading. It has the high authority of the manuscript M, and it has also the approval of Hosius (JV. Jahrb. f. Phil. 1893, p. 351). It appears to be a reminiscence of Ovid, who twice begins a line with this hemistich, Metam. viii. 566; xii. 155. Relation op Accent to Pyrrhic in Latin Verse 71 states that the last two books are less perfect in form, but the evidence cited illustrates greater freedom of style rather than less regard for form. II. Excluding neque enim, there are 122 cases in which the first element of the elision is an inseparable enclitic. The enclitic is -que except in two cases and in these -ne occurs. In the majority of these elisions the primary word-accent falls on the second element of the elision, as in those belonging to Class I. While there are no exceptions to this rule when the first element is not an enclitic, there are 30 exceptions when -que is used. This is nearly one-fourth of the whole number. This difference in treatment of the two types clearly implies that the accent before -que was not the equivalent of the primary accent of other words. Lucan makes it very plain by the structure of the verse that the caesura does not occur in connection with the elision, except in two lines which contain proper names, vii. 541, 711. Twenty of these elisions are in the third foot. 1 The sense-pause in the second foot is marked by punctuation in Prancken's edition in 18 of these lines and the trithemimeral is clearly the main caesura in the two lines which have no mark of punctuation. In 8 lines the elision is in the second foot and the caesura is in the fourth. In 6 of these the sense-pause, which corresponds with the main caesura, is indicated in our edition by punctuation, and in the two remaining cases the same caesura is clearly indicated by the sense. 2 The elisions in the fourth foot may be divided into two classes: (1) those in which the first element of the elision is an insepar- able enclitic. The enclitic is -que, except in some half-dozen cases. The entire number is 220. In 165 of these the second element of the elision is a preposition, or a preposition in compo- sition. (2) those in which the inseparable enclitic does not occur. In 130 of these the second element of the elision is a preposition, or a preposition in composition. In 39 cases the second element H. 11,422; ii. 373, 591; Hi. 270, 387, 456, 476, 615; iv. 27, 261, 642; v. 622, 629; vi. 567, 653; vii. 357; viii. 746; ix. 742, 1039. 2i. 68; ii. 98, 410; iv. 620: v. 307, 675; viii. 5; ix, 794. 72 Albert Gbangeb Haekness is an unemphatic monosyllable. 1 The remaining cases number only 29. Though sense-pauses occur with unusual frequency in Lucan in this part of the verse, they do not occur in connection with this elision. Usually the words forming the elision are very closely connected in thought. The poet evidently felt that if this transitional foot were made prominent by a strong accent and weighted by elision it would detract from the impressiveness of the end-rhythm. For the sake of brevity I shall limit the consideration of Ovid to the first six books of the Metamorphoses. Ovid follows the same general principles which we have seen exemplified in Lucan, but in minor points he shows greater freedom. In the first foot his percentage of elisions is nearly twice as great as in Lucan. An elided dactylic word is sometimes fol- lowed by an accented syllable. This elision appears to be used to heighten the artistic effect by giving variety and emphasis. 2 Elisions in the second and third feet. — The percentage is about twice as high as in Lucan. Classifying them as before, we consider: I. Those in which the first element of the elision is not an inseparable enclitic. In Ovid we have two classes: 1. Those in which the elision is in the first syllable of the thesis of the dactyl, or in the spondee. 3 2. Those in which the elision is in the second syllable of the thesis. When the first element of the elision is the final of a dis- syllable, the accent is on the first short syllable of the thesis, as may be seen from the examples: i. 250, sibi enim, 503; ii. 596; iv. 17, 215, 426, 452; v. 26, 593. 4 lest occurs 21 times; et, 9; ac, 5; an, 2; haec, 1; hoc, 1. 2 Of. ii. 118; iii. 153; iv. 39; v. 647. 8 The following correspond to the norm of Lucan: i. 469, 714; ii. 87; iii. 39; iv. 765, 780 ; v. 94, 134. The following are slightly different in character : ii. 520, 803 ; iv. 447, 465; v. 259, 476, 477; vi. 154, 497. The following are exceptional: i. 478, 759; v. 225; vi. 269. This type is justified by the pause-elision. Similar exceptions are frequent in Virgil (see p. 76, n. 2). 4 ii. 305, qui dederat currus, nisi opem ferat, omnia fato, would appear to be an exception and to have the main sentence-accent on the second syllable of the thesis. If this clause is accented as a word-group, nisi opim ferat, an accent which is favored and helped by the verse-rhythm, it does not present ah exception to the rule. Relation of Accent to Pyrrhic in Latin Verse 73 When the first element of the elision is the final of a dactylic word, the following word is always an iambic word, though in the first foot a monosyllable often occurs as the second part of the elision. The reason for the difference in treatment is the fact that in the second and third feet the thesis must receive an accent. 1 II. Those in which the first element of the elision is an inseparable enclitic. 1. The elision is in the first syllable of the thesis of the dactyl, or in the spondee. a) There are 69 elisions which have a primary accent in the second element of the elision and thus conform to the norm. b) Those which do not have a primary accent on the second element of the elision. There are 8 elisions of -que followed by et. 2 et belongs more closely to the following word than to the preced- ing; it often introduces a clause so that there is a definite sense- pause before it. This type of elision is thus usually justified by the law which governs pause-elision. 3 There are only 15 remaining cases, a much smaller percentage than in Lucan.* In these the second element of the elision is formed by the secondary accent of a polly syllable. Ovid does not introduce this type of elision without a definite justification. In iv. 10, 15; v. 254, it is justified by the Greek character of the line. In the remaining cases 5 it is justified by the force and vigor of the thought which is strengthened by the 1 The only question is whether the iambic word retains its word-accent in the sen- tence. If it does, the second syllable of the thesis receives the accent, but this accent, as we have seen, is not the one favored by Latin poetry. If the iambic word forms a part of a word-group we have the regular accent on the first syllable of the thesis. I have noted the following examples of this type of elision : ii. 85 ; iii. 289, 417 ; iv. 215, 752 ; v. 197, 315, 588. The answer to this question must be deferred to my second paper. 2i. 117, 131, 193; ii. 795; iii. 134, 515; v. 459; vi. 514. 3 In Lucan there are but two such elisions ; iv. 27 ; vii. 711. 4 1 do not include in this number the type -que ddituque, iii. 226, 595 ; iv. 15, 744 ; vi. 597; cf. p. 71. I also omit vi. 607, -que in humum, since the preposition with its case forms a well-recognized word-group and in this type the primary accent is on the preposition, and vi. 185, quoque audete. quoque, as a sentence-enclitic, would not have a sentence-accent and would hardly be more prominent than the first syllable of audete. Thus the foot would conform to the general law of accent which prevails in this part of the verse. H. 231 ; ii. 232, 400; iii. 109, 530; iv. 408, 527, 689; v. 233, 360, 421 ; vi. 47. 74 Albert Gkangeb Habkness special weight of this rhythm. With one exception the elisions occur in the third foot, whereas by far the great majority of the regular elisions of -que (Class a) occur in the second foot. When this elision occurs in the third foot the remainder of the line has coincidence of accent and ictus.' Though Ovid is in so many respects less strict in his verse structure than Lucan, his use of this type of elision shows how superior he is in the artistic structure of the verse. 2. Elisions with -que in the second syllable of the thesis. In all of these neque forms the first part of the elision and the second part is formed by a subordinate sentence-word. 2 It is an extension of the type neque, or nee enim (cf. p. 68, n. 2), and the sentence-accent falls on the first syllable of the thesis. The treatment of the fourth foot does not differ materially from that in Lucan. The percentage of cases in which the elision is formed with -que, or is formed without either -que or a preposition or a preposition in composition, is higher than in Lucan. Ovid, in contrast to the usage of Lucan and Virgil, does not favor the use of the preposition or preposition in composition except when preceded by -que. 3 1 shall attempt to summarize briefly Virgil's treatment of elision in the thesis of the first four feet, omitting the last six books of the Aeneid. Virgil employs this elision in the first foot somewhat more frequently than Ovid, but compared with the number of elisions occurring in the verse as a whole the percentage in the first foot of Virgil is far smaller than in Ovid. In Virgil the second word forming the elision is frequently a long word, as Georg. ii. 80, plantae immittuntur, 277, indulge ordinibus. Virgil's treatment of the fourth foot differs in a marked degree from that of Lucan and Ovid. In Lucan there are three times as many elisions in the fourth foot as in the first; in Ovid twice as many. In Virgil there are less in the fourth than in the first foot. In Lucan, in a majority of the elisions, -que forms the first I iii. 109, mox umeri pectusque onerataque bracchia tellis, is somewhat exceptional. 2 enim occurs 17 times, adhuc i times, erat , erit, eum once each. 3 Elision occurs in the second syllable of the thesis in ii. 376. Eelation of Accent to Pyeehic in Latin Veese 75 element; in Ovid nearly 75 per cent, are so formed. In the Eclogues and Georgics only about 20 per cent, of the elisions have an inseparable enclitic as the first element. In the second and third feet the percentage of elisions in the Eclogues and Georgics is about twice as high as in Lucan. Con- sidering first those elisions which do not include the inseparable enclitic, we find that the Eclogues and Georgics contain about twice as many as all Lucan. While in Lucan there are no excep- tions to the rule that the second element of the elision should have a primary accent, and only three in Ovid, in Virgil we find that one-third of the whole number are exceptions. Even in those elisions in Virgil which are regular in regard to the accent, both words forming the elision may be unimportant sentence- words, or the first may be a subordinate word, such as a sentence-enclitic. The second of the two words forming the elision does not appear in any case to be the less prominent and thus have the lighter accent. In this respect they conform to the norm of Lucan. In Virgil about one-third of the elisions in which the first element is formed by the inseparable enclitic are irregular. The exceptional cases, including both those formed with the enclitic and those without, appear to be used with a definite purpose. These exceptions do not indicate that the poet devoted less attention to the outward form of the verse than did those whose verse is regarded as more "correct." It is not so easy to classify each line in Virgil and to determine the exact relation of the rhythm to the thought as in Ovid. We may, however, note the conditions under which these elisions occur, and in most cases we can feel the special relation of the exceptional rhythm to the thought. The Eclogues do not show in the use of this elision the boldness which appears in the Georgics and which attains to its full development in the Aeneid. In the Eclogues there are 24 of these elisions; 13 are entirely regular and some of the remain- ing number would hardly be felt to be exceptions, as atque adversos, x. 45. The exceptional form is used a few times to give a conversational tone to the verse, as in vii. 8; x. 21, and also when the elision is in the second syllable of the thesis, iv. 88, 102; vi. 6. 76 Albert Gbangeb Haekness We may classify the exceptional elisions in Virgil as follows: 1 1) Those which have a sense-pause between the vowels of the elision. When the second element of the elision is a syllable long in quantity, the rule here under consideration yields to the rule of pause-elision; when the second element of the elision is short the present rule sometimes prevails. 2 2) Those which are introduced to give a light, conversational tone and thus break the regularity of the stately hexameter. 3 3) Those which occur in lines containing proper names, especially Greek names. 4 4) Those which are used to reinforce the thought. 5 The effect of the rhythm produced by elision in the first syl- lable of the thesis somewhat resembles that of the feminine caesura, since the word-accent in both cases falls on the arsis. It is used in a similar way and is subject to similar rules (cf. Havet, pp. 100, 101; Norden Verg. VI, pp. 421 ff.). 6 1 1 omit those eases in which the arsis of the foot containing the elision is formed by a sentence-enclitic, such as atque or a dissyllabic preposition. It is to be noted that these classes are not absolutely distinct. Some of the excep- tions are of such a nature that they might be classed under either one of two or three headings. For example, some of those characterized by pause-elision also contain Greek names. 2 These elisions are of special interest because the subject of pause-elision is here approached from the opposite point of view from that adopted in Trans. XXXVI, pp. 82 ff., and the same conclusions are reached. The most marked exception to the accent of the thesis in Ovid comes under this head of pause-elision (see p. 72, n. 3), and in Virgil we have the following numerous cases in the first six books of the Aeneid: i. 101, 119, 295, 458, 566, 653, 660; ii. 7, 413, 465, 523, 705, 772; iii. 311, 622, 652, 657; iv. 87, 244, 274, 667 ; v. 20, 206, 630, 703 ; vi. 289, 447, 650. s The effect of this type of elision is clearly seen in Horace's Satires, but is most effectively illustrated by Sat. ii. 3. This kind of elision is a characteristic feature of his satire and differentiates it from his epistles, and still more from the epic verse. We have noted illustrations of this type in the Eclogues; they are comparatively rare in the Aeneid. We may class as such i. 326 ; iii. 382, v. 411. M. 416, 550, ii. 393, iii. 271, 351, 589; iv. 7, 75, 348, 367; v. 223, 407, 537, 576; vi. 289, 447, 650. 6 The conditions under which these are used are the same as we noted in Ovid. Comparatively few of these elisions seem to be employed to give variety of rhythm in contrast to emphasis: i. 359, 420, 465; ii. 16, 20, 497, 593, 609, 653, 722; iii. 38, 298, 315, 357, 382, 427, 549, 658, 674; iv. 168, 280, 297, 532, 564, 602; v. 70, 307, 318, 396, 403, 411, 447, 480, 504, 567, 570, 705, 780; vi. 150, 244, 279, 594, 597, 636. 6 The reason for the unvaried use of trisyllables in the dactylic foot and of the dissyllable in the spondaic foot, and the avoidance of the type amata erat, correspond- ing to ii. 56, vacet tibi, and of amata est, corresponding to i. 99, fuit non, involves the consideration of the sentence-accent, which will be taken up in the second Kelation of Accent to Pykkhic in Latin Vekse 77 Not including cases of elision already considered, the thesis of the dactyl in the second and third feet in Lucan is formed as follows : 1) By a pyrrhic word, by the first two syllables of a tribrach, or of an anapaest. The first short syllable of the dactyl receives the accent. 2) By the first two syllables of a word of four or more syl- lables. The first short syllable of the dactyl receives the secondary accent. There is but one example in our text of the type ----- . : 3) By a monosyllable closely connected in thought with the following word which consists of two or more syllables. These cases, though comparatively infrequent, resemble those under b. Compare p6r-hiantis with pirhibebant. 4) By the final syllable of a trochaic word or ending. This is without exception followed by an iambic word or its equivalent as regards accent. 2 It would be sufficient to note this fact in order to show that this part of Latin verse accords with my general theory, i. e., that, if the first of the two short syllables does not receive an accent, the second receives a compensatory accent. But this theory would not account for all the phenomena involved in the case. Two explanations have been advanced to account for the fact that the two words forming this type of dactyl are closely related in thought, and that the second of the two is an iambic word. paper. The relative infrequeney of elision in this part of the hexameter has often been noted. Trampe (loc, cit., p. 20) refers to the restricted use, in this part of the verse in Lucan, of words other than those ending in -e. In the case of the dactyl the propor- tion of those in -e forming the first element of the elision is not much larger in this class of elisions than it is in the first book of the Aeneid, as a whole. I have not found any explanation advanced for the rarity of this elision, or even mention of the fact that in Lucan, without exception, and in other poets, as a rule, the anapaestic or spondaic word forms the second element of the elision. l i. 488, sedibus exiliere patres. This may be an example of an exceptional usage which appears in comedy ( see p. 52) and is found in Virgil ( xi. 900) . It is possible that Lucan wrote sedibus ex saliere. In Lucretius (v. 949) the approved reading is quibus e scibant, though the MSS gives quibus escibant. Lucan MSS (ii. 273) vary between in casum, and incassum (Haskins loc. cit., note on passage), ex often stands after its case, as in Lucretius; in Lucan the preposition in is similarly placed (vii. 776). 2 The only eases in which the iambic word does not occur are in the two types : (1) i. 209, erexitque urbem et, a pyrrhic word followed by a sense-pause ; (2) iv. 283, nulla data est, a pyrrhic word followed by est , which is the only word thus used ; and in (3) two instances of sat est, v. 137 ; ix. 388. 78 Albert Granger Harkness The usual explanation is that by this method the feminine caesura is avoided and the masculine caesura is introduced into the following foot. This theory does not adequately account for the existing phe- nomena. If the view is accepted that the caesura is not deter- mined by the thought, the fact that the words forming the dactyl are as a rule so closely related loses its significance. If, on the other hand, the theory is accepted that the caesura accords with the sense-pauses the regular occurrence of the iambic word is without meaning in lines of the following type: i. 406, urguet rupe cava pelegis: non corns in ilium. Let us examine Trampe's explanation as given in his admirable monograph. He admits but two examples of the feminine caesura in all Lucan (loc. cit, p. 71). But if we apply his own theory, that the caesura may occur between words as closely related as the preposition and its case (p. 53) instead of two feminine caesurae, we might have nearer two thousand. Again, we cannot admit the distinction which he makes between the two cases cited and a score of others ; for example, between ix. 109, sic ubi fata, caput ferali obduxit amictu, and vi. 719, haec ubi fata caput spumantiaque ora levavit. These all belong to one general class and are treated in the same way by the poet. Accordingly, we do not find any case like the fol- lowing in Aen. i. 290: accipies secura; vocabitur hie quoque votis. Professor Heitland (Jour. Phil. 1898, p. 10) suggests an explanation which attempts to account for the regular occurrence of the iambic word. He says: "That a feminine division fol- lowed by an iambic word afforded the best possible means for making a strong pause is shown by the history of the ending of the Latin hexameter." This theory implies that the pause is the most essential element to be taken into account in the verse-close of the pentameter, and not the accent or ictus. This view does not find support either in the usage of other metres or of prose. In my second paper, which will appear in a later number of this volume, I shall consider the sentence-accent in relation to hexameter verse and shall seek to prove that the structure of the verse is largely affected by word-groups and their accent. Brown University