With these greater poems of the period of rhythmical transition may be classed the lais of Marie de France, which, however, do not come perhaps within our time limits. Taking the lais as a whole, we reach an average of about 10 per cent. for broken couplets and 34 per cent. for three-line sentences which follow the break. But, analyzing the lais separately, we find that the proportion of broken couplets ranges from 1.6 per cent. in Chaitivel to 16.5 in Deux amants, while of three-line sentences after the break Chaitivel offers none and Laustic 75 per cent. Among the more important lais the one of Guigemar shows 10 per cent. of broken couplets and 25 per cent of three-line sentences following the break; Frêne, 13 and 30 per cent. respectively; Lanval, 9 and 35; Yonec, 8 and 63; Milun, 13 and 50, and Eliduc, 10 and 38. The variation among the different lais in sentences of three lines is much greater than the variation in broken couplets.

The next stage in the evolution of the narrative couplet seems to be reached toward the year 1165 in the works of Gautier d'Arras and Benoît de Sainte-More, both of them court poets of the second generation. The first composition of Gautier's pen which has come down to us is the romance of Éracle. Notwithstanding a division of the poem into three parts by the author himself, each part controlled by a different purpose and apparently dedicated to a different patron, and notwithstanding the fact that the last part is supposed to have been written a number of years after the first two, there is no appreciable variation throughout the whole romance in the treatment of the couplet. The proportion of broken couplets constantly remains at 18.5 per cent., and of three-line sentences at slightly over 31. If we compare these figures with those derived from a reading of Gautier's other

continue an old one. With so much latitude allowed, subjective interpretation of an author's meaning is bound to play a prominent part and influence the resultant reckoning. -After the statement made above in regard to the relatively wide difference between my count of broken couplets in Thomas' Tristan and the one made by its distinguished editor, I think it due to us both to print the lines where I still find breaks in the couplet, over and above those enumerated by M. Bédier on pages 33 and 34 of the second volume of his Tristan. These are: 2, 4, 16, 24, 75, 143, 343, 373, 425, 555, 597, 665, 667, 713, 911, 955, 957, 1098, 1112, 1120, 1148, 1203, 1229, 1245, 1249, 1253, 1255, 1313, 1331, 1359, 1417, 1461, 1551, 1607, 1623, 1715, 1735, 1737, 1763, 1773, 1811, 1829, 1841, 1859, 1869, 1891, 1941, 1945, 1947, 1995, 2003, 2005, 2027, 2035, 2051, 2077, 2191, 2223, 2281, 2283, 2349, 2375, 2393, 2473, 2529, 2553, 2563, 2611, 2635, 2659, 2663, 2754, 2971, 2997, 3045, 3113, 3135. I do not find breaks at 1741, 1777, or 2063. M. Bédier's 289 and 561 are evidently typographical errors for 287 and 551. I fail to identify his 1451 and 1453, which seem out of place.

romance, Ille et Galeron-assumedly later than the first two parts of Éracle, and differing from that poem in certain features of style, such as transposed parallelism-we find the percentage of broken couplets is identical with the percentage of Éracle, while the ratio of three-line phrases falls to 20 per cent. It should also be stated that, while Éracle offers as many as forty instances of overflow verse, Ille et Galeron presents barely ten. Consequently, a considerable difference may be said to exist between the versification of the two poems. Whether this difference was merely accidental, or was due to a change in the author's environment, cannot, of course, be determined.'

Gautier d'Arras, therefore, is fairly consistent in his treatment of the broken couplet, though not consistent in the matter of threeline sentences and overflow verse. But Benoît de Sainte-More, his contemporary, may rightly be accused of great vacillation, particularly in his Roman de Troie. So far as may be determined by the text of Joly's edition, that celebrated romance presents three phases of treatment in regard to the broken couplet. The first 8,000 lines show a ratio of 9.5 per cent.-or the proportion observed by the fragments of Tristan-with 34 per cent. of sentences in three lines. With the ninth thousand this percentage is raised to 15 for the one and falls to 20 for the other-not far from the rhythm of Ille et Galeron. Some 4,000 lines retain this average. The main body of the poem-lines 12,000 to 28,000increases these figures to 21.5 per cent. for the broken couplets, and lowers them to 15 per cent. for the three-line sentences. The remaining 2,000 lines return to a percentage of 15 in broken couplets, and give the very low proportion of 11 per cent. for the sentences of three lines. The reasons for such exceptional variations are not apparent. It is possible that Benoît, of the generation next to Wace, properly belonged to the transition period of this feature of style, and should therefore be classed with Thomas, the author of Enéas, and Marie de France. But his avoidance of the three-line phrase, after the first fourth of Troie, is much more pronounced than any other poet's. It seems to be peculiar to him,

1 The figures given for Eracle were those obtained by a second reading. The first reading showed 0.5 per cent. less in broken couplets and 4 per cent. less in sentences of three lines

individual, and not induced by outside influence. Unfortunately, the Chronique des Ducs de Normandie adds but a fraction to our knowledge of Benoît's manner. In that long narrative he quite steadily maintains a slightly lower percentage of broken couplets 13.7—than he had adopted for the closing lines of Troie, and one slightly higher-13-for the sentences of three lines.'

We now come to Chrétien de Troyes, to whom M. Meyer assigns the leadership in the new school of versification. And it is only necessary to glance at his Erec in order to assure ourselves that this primacy is well bestowed. For the first 2,000 lines of that poem we note a percentage of 15.5 of broken couplets and 28 of sentences of three lines. In these particulars, and in overflow also, we might find a likeness to Gautier's Eracle. But with the third thousand of Érec the ratio of broken couplets rises to 25.6 per cent., and the proportion of three-line phrases falls to 23 per cent. With the fourth thousand, and for the remainder of the poem-almost 4,000 lines-broken couplets form 37 per cent. of the whole, while the ratio of three-line sentences following the break is 20 per cent."

The other poems of Chrétien which belong to the same period of his career as the Érec do not vary in their use of this feature of style from the manner observed in the last 4,000 lines of that The average of broken couplets in Cligès is 33 per cent.

romance.

1A re-reading of the first half of Troie, in the two volumes of M. Constans' edition, now at hand, changes these computations, particularly in regard to the three-line sentence which follows the break in the couplet. The first 7,000 lines of this critical text (instead of the first 8,000 in Joly) would show a percentage of 10.5 in broken couplets and 39 in the sentences of three lines which follow. In the next 3,000 lines the percentage of broken couplets rises to 16.8, and the percentage of three-line sentences falls to 27.6. In the next 5,000 lines (10,000-14,882) the percentage of broken couplets reaches an average of 21.2, while three-line sentences form 32 per cent. of all those which follow the break. The general conclusion as to Benoît's attitude toward the broken couplet, which is given above, would seem to be confirmed, but the inference that he avoided three-line sentences is quite surely a mistaken one so far as Troie is concerned. This poem would belong to the class represented by Gautier's Eracle.

2 These figures were furnished by a second reading of Erec. They differ from the results of the first reading as follows: 0.5 per cent. less in broken couplets for the first 2,000 lines, 4 per cent. more for the third thousand, and 0.5 more for the remainder of the poem. The difference in the proportion of three-line sentences reached by the two readings is much greater, and illustrates again the uncertainty of such computations. For the first 2,000 lines it was 0.7 per cent. less at the second reading, for the third thousand 7 per cent. more, and for the rest of the poem 12 per cent. more. That the second reading may come nearer to the real manner of Chrétien would seem to be indicated by his subsequent handling of the couplet in his later poems.

(37 in the first 1,000 lines); in la Charrette, 34; in Guillaume d'Angleterre, 34, and in Iwain, 32. The percentage of threeline sentences in Cligès is 19; in la Charrette, 16 (but in the part ascribed to Godefroi de Laigni, 20); in Guillaume d'Angleterre, 20 (with a steady fall from 23 per cent. in the first 1,000 lines to 16 in the final 300), and in Iwain, 18. The ratio of overflow verses in these later poems is about the proportion observed by Érec. So that, if we make due allowance for errors and uncertainties, it is plain that, after a slight hesitation in the first 3,000 lines of Erec, Chrétien settled down to a definite scheme of varying the rhythm of the narrative couplet and deliberately broke every third one throughout his compositions. As to the length of sentence which should follow this break, he seems to have allotted to the three-line phrase a proportion of about one in five. His innovation becomes all the more striking when we compare his high percentage of broken couplets to the moderate increase made by Gautier and Benoît over the proportion of Énéas.

It would seem also that the other court poets of the time considered Chrétien too daring, and sided with temperate Gautier and Benoît. The author of the Comte de Poitiers, assigned to the years around 1170, observes an average of 17 per cent. for broken couplets and 27 per cent. for sentences of three lines.' A slightly higher percentage, but still one which remains within the limits set by Benoît, is found in Amadas et Ydoine, of uncertain date, but probably not many years younger than Cligès. Its proportion of broken couplets is 20 per cent.; of three-line phrases, 25. Overflow verse runs as high as twelve in a thousand, thus exceeding even the ratio of Iwain.

From the testimony of these few witnesses we may infer that Chrétien's pre-eminence as a versifier was not admitted by his contemporaries. Indeed, certain poems of the seventies or early eighties react quite decidedly against the liberty taken even by Gautier d'Arras and Benoît de Sainte-More. The Roman du

The Comte de Poitiers is thought to have been written by two different poets. The so-called second part of the romance presents 3 per cent. less of broken couplets than the first part, and 2.5 per cent. less of three-line sentences. It does not furnish any examples of overflow verse, while the first half rivals the ratio in Erec.

Mont-Saint-Michel (about 1170) reverts to the older manner entirely, with its 5.3 per cent. of broken couplets and 49 of threeline sentences. De David la prophecie1 (1180) is still more traditional, with 2.5 and 52 per cent., respectively. Partonopeus de Blois allows but 4.5 per cent. to broken couplets, and 45 per cent. to sentences of three lines. Ipomedon is only less reactionary with its 10 per cent. of broken couplets and 32 per cent. of three-line phrases. Neither of these romantic poems indulges to any extent in overflow verse.

Finally, the Breton lais not ascribed to Marie de France belong in great measure to the older school, whether because of Marie's example in this respect, or because of the assumed reaction against too free a handling. Melion offers 3 per cent. in broken couplets and 87 per cent. in three-line sentences; le Cor, 5.7 and 17.5 per cent., respectively (a quite exceptional ratio between the broken couplet and the three-line phrase); Tydorel and Graelent, 7.5 and 55 per cent.; Tyolet, 9 and 48; Désiré, 10 and 42; Épine, 12 and 50; Guigemar, 15 and 33, and Doon, 17 and 32-variations in broken couplets which are all included between the minimum of Rou and the maximum of Troie. Ignaure and the Mantel, obviously much later, come under Chrétien's influence with 33 and 27 per cent., and 42 and 18 per cent., respectively.

From the facts thus obtained we would suppose the history of the narrative couplet in the twelfth century to read somewhat as follows: At first a strict adherence to the primary conception of one couplet for one thought; then, toward the fifth decade, a slight deviation from this conception in practice, but not in principle. Representatives of this state would be found in the Ste. Julienne, St. George, and Wace's St. Nicolas and Conception. With Geoffrey Gaimar we enter on a new era, which is marked by a desire to vary the monotony of the rhythm without destroying the mold in which it was cast. The proportion of sentences of three lines remains high, and would characterize the spirit which governed this innovation.

With Wace, on the other hand, there is no progress.

1 Published in Zeitschrift für romanische Philologie, Vol. XIX, pp. 189-234.

After a