The Performance of Dufay's Paraphrase Kyries BY EDWARD KOVARIK IN INCLUDED THE FOURTH VOLUME of the Opera omnia of Guillaume Dufay is a group of ten three-voice Kyries which are polyphonic settings of plainchant Kyrie melodies.' Some of the settings include plainchant sections which are to be performed in conjunction with the polyphony, but as they are printed in the Opera omnia, there seems to be no consistent policy with respect to how plainsong and polyphony interact. In two cases, chant sections alternate regularly with the polyphony, in a third the chant appears as the first of each group of three acclamations, and in one setting an isolated section of chant appears at the very beginning; the remaining settings-the majority-lack plainchant sections altogether. It will be my purpose in the following pages to show that this treatment of the plainchant is a misinterpretation of the sources as they have come down to us, and that Dufay's original intention was certainly simpler and more consistent than the Opera omnia would lead us to believe. The Kyries are scattered about in eight fifteenth-century sources.2 No more than six of the ten appear in any one manuscript, and only two or three appear in direct succession. Nevertheless, they are so closely inter- related in style and by their use of plainchant paraphrase that it would be willfully bullheaded to consider them as other than a unified set of works.3 Four paraphrase Glorias written in the same style may be included as companion pieces to the Kyries.4 Among the Kyries, two works stand apart from the others. These are duplicate settings of two of the cantus firmi, indicated in Table I by 1 Fragmenta Missarum, ed. Heinrich Besseler, Corpus mensurabilis musicae, I (Rome, 1962), pp. 61-71 (Nos. 9-18); the work following (No. 19) is a later, freelycomposed Kyrie. Diplomatic versions of the Kyries in the original notation (plus analyses) are available in Rudolf Bockholdt, Die friiben Messenkompositionen von Guillaume Dufay, Miinchener Ver6ffentlichungen zur Musikgeschichte, 5 (Tutzing, I96O). 2 BL, BU, Em, Ao, Tr87, 90, 92, and 93; the sigla are those of the Opera omnnia. 3 See Charles Hamm, A Chronology of the Works of Guillaume Dufay Based on a Study of Mensural Practice (Princeton, 1964), pp. 79 and 86, where the Kyries are dated 1433-1435. Hamm's "Kyrie XVIII" is actually Kyrie IIa; two Kyries that Hamm reserved for later discussion (pp. 117 and I2z) also belong in this group, as Hamm himself surmised; they are Kyrie VI and Kyrie XIa. 4Three of the Glorias are grouped with the Kyries by Hamm, op. cit., pp. 79 and 86; the fourth, Gloria XV, was arbitrarily separated from them because it begins with the signature 0 and does not call for alternation with plainchant (p. 114).
THE PERFORMANCE OF DUFAY'S PARAPHRASE KYRIES 231 Kyrie or Gloria TABLE I Dufay's Paraphrase Kyries and Glorias Page and No. in Opera omnia, Vol. IV Kyrie I 68-69 (No. 15) Kyrie II 69-70 (No. I6) Kyrie IIa 70-71 (No. 17) Kyrie IV 62 (No. io) Kyrie VI 71 (No. I8) Kyrie IX 67-68 (No. 14) Kyrie XI 63-64 (No. Ii) Kyrie XIa 64-65 (No. 12) Kyrie XII 61 (No. 9) Kyrie XIV 65-66 (No. 13) Gloria IX 83-84 (No. 24) Gloria XI 85-86 (No. 25) Gloria XIV 86-88 (No. 26) Gloria XV 88-89 (No. 27) the letter a: Kyrie IIa (No. 17) and Kyrie XIa (No. 12). They differ from their companions in that they are written for two high parts (of approximately the same range) plus tenor, with the cantus firmus in the second high part-that is, in the middle voice5-whereas the other Kyries are written for the standard combination of discantus, tenor, and contratenor, with the cantus firmus in the uppermost voice. The two middlevoice works, though distinguishable by their unusual format, are closely related to the others in style and also through specific melodic resemblances. A comparison of either the two settings of Kyrie II or the two settings of Kyrie XI shows that for the first few measures of each section the manner of paraphrase is the same; that is, the voice carrying the cantus firmus is identical in both settings, even though it is the top voice in one case and the middle voice in the other. This identity lasts for just a few measures (at the beginning of each section), but it is enough to show that one Kyrie of each pair is, in effect, a revision of the other. Now, which are the originals and which the revisions? In various ways the middle-voice settings are simpler in compositional technique than their discant counterparts. The middle-voice settings are shorter, section by section, and their texture is less complex. The three voices move together in a broader, more homogeneous ensemble rhythm, with relatively few short note values (minims and semiminims), and with little use of conflicting patterns such as semibreve-minim in one voice and minimsemibreve in another. The identical cantus-firmus passages mentioned earlier are harmonized in a straightforward manner in the middle-voice settings but sometimes more subtly in the discant settings-with progressions comparable to the modern V-VI (see Ex. i). From this evidence, we may arrive at two alternative conclusions: in revising his works, Dufay either expanded the originals and made them more subtle and complex 5Because the two high parts cross, the cantus firmus is sometimes on top, but it is always in the middle at the ends of sections.
232 JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY (if the middle-voice settings are to be considered the earlier ones), or else he shortened and simplified them and removed the subtleties (if the discant settings are taken to be earlier). The first of these alternatives seems the more likely one. Among Dufay's works, these two Kyries are the only examples of middle-voice cantus firmus, and both are unica (one in the Bologna MS [BL], the other in MS 87 at Trent). By way of contrast, Dufay's discant paraphrase settings include eight Kyries and four Glorias, most occurring in two or more sources, some additional music for the Ordinary of the Mass, and a substantial body of similar settings of other liturgical chants such as Hymns and Sequences. Presumably, the middle-voice settings were a false start-one the composer abandoned in favor of the much more successful discant format. Apart from the two middle-voice works, the remaining eight Kyries form a cycle evidently intended for various classes of feasts throughout the church year. The settings are all similar in style, but their sections vary markedly in length, and this feature has been used to order the works as they appear in Table 2.6 Most of the pieces fall into a middle group in which three sections of polyphony take up some thirty to forty measures in modern transcription. The first and last items, however, differ substantially from this middle group. In the first, three sections require seventyone measures; in the last, a mere nineteen measures. The difference between the extremes of the table may be expressed as a ratio of about 3 /2 to i. This amount of variation cannot be explained simply on the basis of the underlying cantus firmi, for the plainchants involved, Kyrie II and Kyrie XII (similarly the longest and shortest of the group), vary in a ratio of less than 21/2 to I.7 Had Dufay so wished, he could certainly have made some attemp to minimize the differing lengths of his plainchant originals; he chose instead to exaggerate the differences. Evidently, the differing lengths are of some significance; perhaps it is a liturgical significance. If so, the longest settings were probably intended for the most important liturgical occasions, the shorter settings for lesser occasions. This principle, after all, applies in general to the underlying plainchants: in fifteenth-century graduals as in the modern Vatican edition, the longest and most elaborate Kyrie chants are usually associated with the highest feasts, the shortest chants with simple feasts and ferias.8 6 Most of the settings consist of three sections: Kyrie, Christe, and Kyrie. There are additional sections in Kyrie XI and Kyrie XIV, neither of which has been taken into account in Table 2. 7These figures are based on a note count of the chants as they appear in the Liber usualis or Graduale Romanum. Dufay's cantus firmi may have differed by one or two notes, but not by enough to affect the outcome of the calculations. 8 A brief comparison of the Kyries of Masses II, IV, XII, XV, XVI, and XVIII in the Liber usualis or Graduale Romanum will suffice to illustrate this principle. All of these chants appear with similar rubrics in fifteenth-century sources (see fn. 15, below).
Example i Kyrie IIa, Second Section Kyrie THE PERFORMANCE OF DUFAY S PARAPHRASE KYRIES 233 II, Second Section C.f&% "...- Chri - ste Chri - ste 0- 'ft Chri - ste Chri - - ste 8 Chri - - ste Chri - ste Kyrie XIa, Last Section Kyrie XI, Last Section C.f. A Ky - ri - e Ky - ri - - e Cf. -% 24 b Ct. Ky - ri - e Ky- ri - e T. - S Ky - ri - - e Ky ri - - e This is not to say that a direct correlation between length and liturgical solemnity holds throughout every collection of Kyrie chants. There are many individual exceptions, particularly among the chants that are neither markedly long nor short. Among chants of extreme length or brevity within the kyriale, however, the principle seems quite valid, and in his polyphonic settings Dufay has evidently respected it.9 Included in Table 2 are the rubrics that occasionally accompany our Kyries in the sources. In support of what has just been said, it might be noted that the longest settings are marked "solemne" (compare the modern rubric, "In Festis Solemnibus"), and that shorter settings are assigned to lesser occasions: feasts of apostles and martyrs, Sundays, or major semidoubles.'0 One might expect the paschal Kyrie to appear among the longer chants for the most important occasions, but in fact this setting, Kyrie I, was probably intended for the vigil of Easter rather than for the 9 Another method of analysis leading to the same conclusion is to compare the number of cantus-firmus notes (as closely as this can be determined) with the number of added ornamental notes in the discantus line. Such an analysis shows that the proportion of added notes is highest in Kyrie II, lowest in Kyrie XII. 10 A Marian rubric has been supplied for Kyrie IX because of the virtually unanimous agreement of chant sources and other polyphonic settings on this assignment.
234 JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY TABLE 2 The Length and Liturgical Function of Eight Dufay Kyries Number of measures Kyrie Rubrics* in the Opera omnia Kyrie II Sollemne (Em) Kyrie fons bonitatis (BL) 71 Kyrie IV Solempne (Tr93) 52 Kyrie XIV De apostolis (Tr87) 38 Kyrie XI De martyribus (Em, Tr87, Tr9o, Tr93) In diebus dominicis (Tr92) 36 Kyrie VI In semiduplicibus maioribus (BL) 34 Kyrie IX [B.V.M.] 34 Kyrie I Pascale (Ao, Em, Tr9o) 30 Kyrie XII 19 * The sigla for manuscripts are those of the Opera omnia. feast day itself, since contemporary chant sources usually include it at this point in the Temporale." The only Kyrie prosa actually associated with Dufay's music is "fons bonitatis." These words appear as a rubric in the Bologna manuscript and as a textual incipit (sometimes extended) in the other three sources of Kyrie II: in Em three full lines of the prosa (verses I, 4, and 9) are written in under the discantus. The appearance of other prosa incipits in the Opera omnia (Kyrie Cum jubilo, Kyrie Orbis factor, etc.) can be traced to their use as identifying tags in the modern Vatican kyriale. These texts have nothing to do with Dufay's music, and indeed many of them had passed out of use (except in England and parts of Germany) by Dufay's time.12 In repertoire and arrangement, Dufay's set of Kyries (Table 2) is very similar to the "Kyriale for Roman use"13 found in Italian and Franciscan sources dating from the thirteenth to the seventeenth centuries (Table 3).14 So far as I have been able to determine, chant sources from areas out- 11 See Dom Joseph Pothier, "Kyrie pascal," Revue du chant gregorien, II (1893-94), 12 I23- Information on the chronology and geographical distribution of the Kyrie prosas is available in Margareta Melnicki, Das einstimmige Kyrie des lateinischen Mittelalters (Regensburg, 1955). For example, an analysis of the entries listed under Melodie Nr. 16 (pp. 88-89) shows that the prosa "Orbis factor" appears in no Italian sources later than the twelfth century and in no French sources later than the fourteenth. 13 So labelled in some of the chant sources, it evidently originated (ca. I250) as a Franciscan codification of the liturgy of the papal court; see S. J. P. van Dijk and J. Hazelden Walker, The Origins of the Modern Roman Liturgy (London and Westminster, Md., i96o), pp. 325-28. 14 The complete kyriale, consisting of four-movement cycles (Kyrie, Gloria, Sanctus, Agnus Dei) is printed in van Dijk and Walker, op. cit., p. 328; in Kurt von Fischer, "Neue Quellen zum einstimmigen Ordinariumszyklus des 14. und 15. Jahrhunderts aus Italien," Liber amicorum Charles van den Borren (Antwerp, 1964), pp. 6o-68; and (from different sources) in my study, "Mid Fifteenth-Century Polyphonic Elaborations of the Plainchant Ordinarium Missae" (Ph.D. diss., Harvard Univ., 1973), pp. 63-66.
THE PERFORMANCE OF DUFAY'S PARAPHRASE KYRIES 235 TABLE 3 The Kyriale according to Italian and Franciscan Sources* Rubric Kyrie Gloria In maioribus duplicibus II ad lib. I In minoribus duplicibus IV IV In maioribus semiduplicibus VI II In minoribus semiduplicibus XIV XI In dominicis diebus XI XI or XIV In maioribus simplicibus XII XIV In minoribus simplicibus X VIb** XV In ferialibus diebus ferial Kyriet In festis et commemorationibus IX IX Beate Marie Virgine [Vigil of Easter] I * Roman type indicates chants set by Dufay; italic indicates chants not set by Dufay. Sanctus and Agnus have been omitted. ** A variant form of Kyrie XVI; see my dissertation (cited in fn. 14), pp. 134-35. t Not Kyrie XVIII, as Fischer and Van Dijk have it, but a Kyrie not in the modern Vatican kyriale, the "ferial Kyrie" (Melnicki's catalog, No. 7); see my dissertation, pp. 137-38. side of Italy normally differ in various respects from Tables 2 and 3-.1 English (Sarum) sources place Kyrie XIV at the head of the kyriale and include a long list of prosa texts to be used on important occasions. French sources place Kyrie IV ahead of Kyrie II, feature Kyrie ad lib. I (Clemens rector) in a prominent place, and seldom include Kyrie I. In German sources, Kyrie V appears somewhere near the beginning of the kyriale and a substantialist of other Kyries-most found only in German sources -appears later. Of German origin are several of the rubrics accompanying Dufay's settings ("solemne,""de apostolis," "de martyribus"), but these should probably be attributed to the German scribes of Em, Tr9o, and Tr93 rather than to Dufay.1'6 Two of Dufay's rubrics do match those of the Roman kyriale ("dominicis diebus" and "semiduplicibus maioribus"), but here again there is no proof that they originated with the composer himself; they may have been added by Italian scribes. More significant than the rubrics, in light of what has been said above, is the appearance in Tables 2 and 3 of the same group of Kyries in very much the same order. Particularly noteworthy are the positions of Kyrie II and Kyrie IV at the beginning, and of Kyrie XII near the end; the 15 Though reinforced by data gathered from Melnicki and elsewhere, the following observations are based primarily on a personal study of twenty-seven chant sources (three English, seven French, five German, eight Italian, and four monastic), most dating from the fourteenth through the sixteenth centuries. The sources include printed facsimiles, microfilms, and manuscripts in American libraries; they are listed and described in my dissertation, pp. 672-77. 116 The rubrics "de apostolis" and "de martyribus" in Tr87 are an anomalous case; presumably they were added after the manuscript reached Trent. Compare the Germanic prosa incipit "magne deus" added incorrectly to Kyrie HIa in the same manuscript (Opera omnia, IV, xxviii).
236 JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY ordering of the other Kyries is less important than the fact that they appear somewhere in the middle.17 It is not hard to see why Dufay may have neglected to set the last two Kyries in Table 3, for they are clearly the ones intended for liturgical functions of lesser importance. His treatment of Kyrie XII already represents the absolute limit of simplicity possible within the discant paraphrase style. A further degree of simplification could only be achieved outside of this style; that is, by leaving the last two Kyries to be performed in plainchant. Given this explanation, the overall similarity of Tables 2 and 3 certainly lends support to the theory that Dufay's paraphrase chant settings originated in Italy.'s Turning to the Glorias of the Roman kyriale, also included in Table 3, we find among them the four chants set by Dufay. These do not correspond precisely to the Kyries, however. Dufay apparently did not set the first three Glorias, whereas among the Kyries it is the bottom of the kyriale that is lacking. Perhaps some of Dufay's settings are lost, or perhaps his repertoire of Glorias differed somewhat from that of the Roman kyriale. If so, the Italian origin of these works cannot be regarded as fully established. This is a topic worthy of further consideration, but first it is necessary to return to the main issue, the use of plainchant in conjunction with Dufay's polyphony. Of the eight Kyries listed in Table 2, only four appear in the Opera omnia in conjunction with plainchant, and in only two of these is the chant treated correctly. For two of the Kyries, sets of plainchant sections exist in the sources but have not been included in the Opera omnia. Chant sections for Kyrie I are to be found in Em, folios 34'V35 (No. 66). (See Ex. 2.) Besseler lists this source in his critical notes but does not include the chant sections, presumably because they are written in German neumes and contain German melodic forms not found in Dufay's cantus firmus.19 Nevertheless, the chant sections do furnish evidence on the manner of performance that is now under consideration. Moreover, this is not the only case in which the chant found in a manuscript does not match the polyphony it accompanies; the same thing happens in one of the sources of Kyrie XI (see Ex. 4, below). 17 Except for the Marian Kyrie (Kyrie IX), which usually appears at the very end of the kyriale, and the paschal Kyrie (Kyrie I), which, as noted earlier, usually appears as part of the liturgy for Holy Saturday rather than in the kyriale. 18 Besseler has suggested that Dufay's paraphrase hymns and sequences (similar in style to the Kyries) were written while the composer was a member of the papal chapel (1428-33); see "Von Dufay bis Josquin," Zeitschrift fiir Musikwissenschaft, XI (1928), 3-4; idem, "Dufay in Rom," Archiv fiir Musikwissenschaft, XV (1958), 5. Alternatively, Hamm, Chronology, pp. 8o-8i, suggests that they may have been written in Cambrai or Savoy (1433-35)--perhaps Cambrai is the more likely candidate, since Hamm considers the alternation of chant and polyphony to be a northern characteristic. 19 For example, the third note of the opening Kyrie is c in the Em plainchant (an example of "German chant dialect"), whereas Dufay's polyphonic setting, like non- German sources in general, has a b. There are other differences as well.
THE PERFORMANCE OF DUFAY'S PARAPHRASE KYRIES 237 Example 2 Em, fols. 34'-35 (2) POLYPHONY Ky - ri - e ley - son (Kyvrie) POLYPHONY Chri - ste ley - son (Christe) (5) (6) POLYPHONY Ky - ri - e ley - son (Kyrie) Ky - ri - e ley - son Another Kyrie that is furnished with chant sections is Kyrie XII (see Ex. 3). The chant sections occur in Trent, MS 92, folio 3o0 (No. 1488), a source that is not listed in Besseler's critical notes.20 Besseler missed this source because the published inventory of the Trent codices, Adler and Koller's pioneering work of 1900, is garbled at this point: it cites the opening section of plainchant instead of the discantus voice of Example 3 Tr 92, fol. 130o w S1 -... 1 POLYPHONY -. SKy - ri - e lev - son (Kyrie) (3) (4) (5) POLYPHONY -.. 8 (Christe) Chri-ste lei - son Ky-ri-e lei-son (6) (7) POLYPHONYte (Kyrie) Ky- ri - e lei - son N.B. The natural sign ( ) in (7) cancels a signature of one flat in the preceding section of polyphony. 20 Tr92 is recognized in Hamm, Chronology, p. 168.
238 JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY the polyphony.21 Had Besseler known of the Trent source, he would undoubtedly have used it instead of the Aosta manuscript as his main source for Kyrie XII, for he preferred Trent manuscripts to the Aosta source in editing both Kyrie No. i i and Kyrie No. 15. As it turns out, the differences in readings between the Aosta manuscript and Trent MS 92 are not very substantial;22 the chief importance of the Trent source is that it contains the chant sections. One may object that in the two cases cited above the chant sections do not appear consistently: they are found in just one of the five sources of Kyrie I (another source giving only the first section which is published in the Opera omnia) and in just one of the six sources of Kyrie XII. This is quite true, but the inconsistencies with which these two chants are preserved in the variousources are not limited to these two instances. On the contrary, similar incongruities mark the preservation, in diverse sources, of the other Kyries-those for which chant sections have been published by Besseler. Thus, for Kyrie XI only two of the six sources give the chant sections that are published in the Opera omnia. For Kyrie XIV, one of the three sources gives the complete chant; another gives only the incipit of the first section. For Kyrie IX, one of the five sources gives the complete chant,23 and two others give only the first section, complete or in part.24 This suggests an important generalization: evidently the presence in a source of just one section of chant-or of just an incipit-should be viewed as a cue for the use of plainchant throughout; since the whole of the chant was readily availablelsewhere, there was no need to copy it in toto along with the polyphony. It must be said that the use of chant is contradicted in some sources by repeat marks which call for complete polyphonic performance. In the readings of Kyrie I and Kyrie XII in the Aosta manuscript, the ter mark (: I :) appears at the end of each section in each voice, suggesting that according to this source the use of chant is out of the question.25 Other sources of the same two pieces, however, do include the chant sections, and so we can only conclude that both methods of performance were acceptable to fifteenth-century musicians. Unanswered is the basic question: which method represents Dufay's original intention? 21 Sechs Trienter Codices. Erste Auswahl, ed. Guido Adler and Oswald Koller, Denkmiler der Tonkunst in Osterreich, Jg. VII (Vienna, 1900), p. 78. The error is understandable because the chant is written in void notes which superficially resemble mensural notation (see the incipit reproduced in Ex. 3). It might be pointed out, however, that void notation is used for the chant sections of Dufay's Kyries and Glorias throughout Tr87 and Tr92. 22 Omitting differences in ligatures and texting, there are just two minor variants in the discantus and none in the lower voices. 23 Except for the last section, which has been supplied by Besseler. 24 The clef is wrong in Tr9o, No. 882 and perhaps also in Em (where the chant appears at the end), but the melodic outline is clearly that of the first Kyrie. 25 Partial or inconsistent repeat marks also occur in conjunction with some of the other pieces; most of the sources, however, have no repeat marks at all.
THE PERFORMANCE OF DUFAY'S PARAPHRASE KYRIES 239 One approach to this problem is through yet another question: when chant is present, precisely how is it to be coordinated with the polyphony? As we mentioned earlier, Besseler specifies no fewer than three different relationships in the Opera omnia: ( ) The regular alternation of chant and polyphony, beginning and ending with chant (in Kyrie XI and Kyrie IX). i 2 3 4 5 6 7 8 9 Kyrie Kyrie Kyrie Christe Christe Christe Kyrie Kyrie Kyrie chant Polyphony chant Polyphony chant Polyphony chant Polyphony chant (2) The first section in chant; the remainder in polyphony (in Kyrie I). i 2 3 4 5 6 7 Kyrie Kyrie Kyrie Christe Christe Christe Kyrie chant Polyphony Polyphony Polyphony Polyphony Polyphony Polyphony 8 9 Kyrie Kyrie Polyphony Polyphony (3) One section in chant, the next two in polyphony (in Kyrie XIV). I 2 3 4 5 6 7 Kyrie Kyrie Kyrie Christe Christe Christe Kyrie chant Polyphony Polyphony chant Polyphony Polyphony chant 8 9 Kyrie Kyrie Polyphony Polyphony The sections printed in italics above are usually not written out in the sources but must be supplied in performance by repeating previous sections; it is the variety of ways in which this may be done that is now at issue. To put the matter plainly: a dispassionate reading of the evidence shows that the second and third alternatives given above are invalid. Only the first manner of performance is correct, and it should be applied uniformly to all five of the Kyries with chant sections and to at least one of the others; indeed, perhaps to all of them. To see why this is so, it is only necessary to look carefully at the way in which sets of plainchant sections are written into the sources. The chant sections are not arranged haphazardly; rather, they alternate with the polyphony in the following distinctive manner, which might be called the "seven-section short form." 7 8 9 Kyrie Kyrie Christe Christe Kyrie Kyrie Kyrie chant Polyphony Polyphony chant chant Polyphony chant To fill out the complete nine sections, an additional Kyrie is needed near the beginning, and an additional Christe is needed in the middle. The form of most plainchant Kyries, Kyrie i-3 (aaa), Christe 4-6 (bbb), Kyrie 7-9 (ccc'), allows repetition of either preceding section, but obvi-
240 JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY ously the most logical scheme is to repeat the first Kyrie (chant) as Kyrie 3 and to repeat the first Christe (Polyphony) as Christe 6-this results in a regular alternation of chant and polyphony. There is more than logic to support this interpretation. Two of the sources of Kyrie XI include chant sections: one of them (Tr87, No. 68) is arranged in the seven-section short form; in the other (Ao, No. 29), all nine sections are written out in regular alternation-the only such case among the Dufay sources-and it is reasonable to assume that both sources call for the same manner of performance.26 A comparison of the two sets of plainchant sections, incidentally, shows one important difference: at the beginning of Kyrie 7, Tr87 resembles the modern version of the chant (see Ex. 4a), whereas the Aosta manuscript includes two extra notes (Ex. 4b).27 Dufay's polyphony-kyrie 8-corresponds to the Aosta version.28 Example 4 (a) Tr 87 (b) Ao (and Dufay) _A j70 (etc.)0 (etc.) The chant sections of Kyrie IX are also arranged in the seven-section form in Tr92, No. 1421 (except that the last section of chant, Kyrie 9, is lacking), and in this case the unusual form of the original plainchant makes it possible to determine exactly where Dufay's polyphony fits in. The form of Kyrie IX is as follows: Kyrie i-3 (aba), Christe 4-6 (cdc), Kyrie 7-9 (efe'). Dufay supplies polyphony for sections b, c, and f; presumably his setting of section c is to be repeated (though this is not clearly marked in the sources). The result is again a regular alternation of chant and polyphony. Of the five Kyries with chant sections, the two just mentioned, Kyrie XI and Kyrie IX, furnish the strongest evidence for the regular alternation of chant and polyphony; indeed, these are the two pieces for which the Opera omnia specifies alternate performance. The seven-section short form described above, however, also occurs in sources of Kyrie XII and 26 This piece (Kyrie XI) is unique among Dufay's Kyries in that the composer specifically calls for two performances of the Christe, the first in three written parts, the second in fauxbourdon (same discantus, different tenor). 27 The variant reading in the Aosta manuscript is a common one found in a number of French, German, and Italian chant sources; see my dissertation (cited in fn. 14), p. 697. 28 This would seem to show that for Kyrie XI Ao is a better source than Tr87. This statement does not hold for the two manuscripts in general, however. Of the five Dufay Kyries in Ao, Kyrie XI is the only one that includes chant sections, and it may be a later addition to the manuscript, since it has an opening to itself and is written in a scribal hand different from that of its companions.
THE PERFORMANCE OF DUFAY'S PARAPHRASE KYRIES 241 Kyrie XIV, indicating that these, too, should be performed in alternation. About Kyrie XII no more need be said: the arrangement of chant and polyphony is illustrated in Example 3, and, even without the chant sections, it is clear that the last section of Dufay's polyphony represents Kyrie 8 rather than the longer Kyrie 9. The arrangement of Kyrie XIV in Tr87, No. 67 is as follows: 7 8 9 Kyrie Kyrie Christe Christe Kyrie Kyrie Kyrie chant Polyphony Polyphony chant chant Polyphony Polyphony This differs from the usual seven-section arrangement only in that the last section is set in polyphony rather than being left in chant. Besseler takes the last three acclamations, Kyrie 7-9, as a model and calls for similar performance of the first six sections (the third of the three arrangements illustrated above), even though in so doing he is forced to dis- regard the order of the Christe sections in Tr87, No. 67. A better interpretation would be the usual alternation of chant and polyphony, but with two sections of polyphony at the end.29 The fifth Kyrie, Kyrie I, appears in the Opera omnia with only the first section in chant. This is surely incorrect. As indicated earlier, the first section should be looked upon as a signal for performance in conjunction with plainchant throughout, and as a matter of fact, chant sections occur in one source of Kyrie I (see Ex. 2, above). The order here differs from the usual seven-section form (I attribute this simply to scribal misadventure), but it is clear that the arrangement begins and ends with chant, and it certainly seems to imply an alternation of chant and polyphony. In addition to the five Kyries with chant, there is a sixth which must have been intended for the same kind of alternate performance, even though no chant survives in the unique source, the Bologna manuscript.30 This is the setting of Kyrie VI. Like Kyrie IX, this chant has the unusual form aba cdc efe', and so it is possible to tell exactly where the polyphony fits in. A comparison with the modern version of Kyrie VI, however, yields somewhat puzzling results. The second and third sections of Dufay's setting resemble Christe 4 and 6 (c) and Kyrie 8 (f), respectively -just as they should.31 The first section of polyphony, however, resembles neither Kyrie 2 (b) nor Kyrie I and 3 (a); instead it is a setting of the melody of the modern Christe 5 (d). What is this section doing at the beginning of the piece? The answer lies in the variant forms of this melody found in chant sources of Dufay's time. Only Sarum sources and 20 This is so despite the presence of some contradictory repeat signs in Tr87, No. 76 and in Em. 30 No chant occurs in connection with any of the Kyries preserved in BL. 31 Except that Dufay's cantus firmus varies slightly from the modern form of Kyrie VI: his second section, Christe 4 and 6, begins cac cb instead of babc b.
242 JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY some French ones anticipate the modern form of Kyrie VI. In many French sources the melody of Christe 5 does not appear at all-the three Christe acclamations are identical and similar to the modern Christe 4 and 6-whereas in Italian sources the melody for the modern Christe 5 is used not only for Christe 5 but also for Kyrie 2. Dufay's cantus firmus, in other words, follows the Italian form of the chant. Can we conclude from this circumstance, and as well from the similarity of Tables 2 and 3 noted earlier, that Dufay's Kyries and their companion Glorias were written while the composer was a member of the papal chapel? Probably not, for although some of the variant forms used in Dufay's music are Italian, some others are French.32 Moreover, Dufay makes use of some variant forms which cannot be found in any chant source,33 and it is hard to believe that this could be the case if the chant forms were those in use in the papal chapel. Elsewhere, I have proposed that the settings may have originated in Savoy, as intermediate between France and Italy; at present, however, this is merely conjecture.34 Six of the ten paraphrase Kyries have now been examined, and in all six cases the evidence for alternate performance seems quite persuasive. In the remaining four cases the evidence is less clear-cut, though it remains consistent with the assumptions already made. In the settings of Kyrie IV and Kyrie XIa the final section of polyphony represents Kyrie 7-8 rather than Kyrie 9, and so the implication is that Kyrie 9-and, by extension, the odd-numbered sections preceding-should be performed in chant.35 The two settings of Kyrie II seem to contradict the established pattern because the last section of polyphony resembles the modern Kyrie 9 and not the modern Kyrie 7-8. Here, as in the case of Kyrie VI, however, modern chant books do not tell the whole story. An examination of early chant sources has shown that the traditional form of Kyrie II-most probably the form set by Dufay-has three identical sections at the end; that is, Kyrie 7, 8, and 9 are all similar to the modern Kyrie 9.36 Thus, the final 32 These statements are based on a study of the chant sources mentioned in fn. 15. Three examples: the third section of Dufay's Kyrie IV (m. 45) includes an extra note, b (third section of chant, notes 12-13 -- c b a instead of c a), a reading found in French sources from Chartres and Amiens, but not elsewhere. The second section of Dufay's Kyrie IX (mm. 19-20) includes an extra note, a (fourth section of chant, notes 9-o-- = g a bb insead of g bb); similar readings occur in a French source from Limoges and in German sources from Trier and Zwickau. In the second section of Dufay's Kyrie XIV, several notes are omitted (second section of chant, notes 5-7 omitted), a reading I have found only in one French source from Compiegne. 33 Among those I have consulted, at any rate. Compare, in this regard, Besseler's report (Opera omnia, IV, v-vi) that Stiiblein was unable to identify precisely the melodic forms used by Dufay for the "Spiritus et alme" tropes to Gloria IX. 34 Kovarik, A chant "Polyphonic Elaborations," p. 300. 35 incipit for Kyrie IV appears in one source (Em, No. 56), but no source preserves a complete set of chant sections. 36See Gabriel M. Beyssac, "Notes sur le Kyrie 'fons bonitatis,'" Rassegna Gregoriana, III (1904), cols. 531-44. The modern form of Kyrie 7-8 appears no earlier than the sixteenth century and must be viewed as an editorial abridgment giving the
THE PERFORMANCE OF DUFAY'S PARAPHRASE KYRIES 243 section of Dufay's polyphony, though it resembles the modern Kyrie 9 and not Kyrie 8, may in fact represent Kyrie 8. These settings, too, are consistent with the use of alternate sections of plainchant. Dufay's Kyries and their companion Glorias comprise the earliest known collection of Mass music written specifically for alternatim performance. It is possible to assume that this is merely the first visible indication of a convention of long standing: to assume, in other words, that the Kyrie was traditionally associated with some such kind of alternate performance.37 There is no concrete evidence to support such an assumption, however, and it certainly fails in the case of the Glorias: there simply are no polyphonic Glorias suitable for alternate performance earlier than this set by Dufay. Moreover, it seems clear that the alternatim procedure, particularly as applied to the Kyries, was unfamiliar to and not readily ac- cepted by Dufay's contemporaries and immediate successors. In many sources the chant sections were suppressed, and in some cases repeat signs indicating continuous polyphonic performance were added. Dufay's most celebrated contemporary, Binchois, also wrote a set of paraphrase Kyries, but his works do not provide for the use of chant; instead, they include separate settings for each new section of the original chant, and rubrics in the sources ("Kyrie primum et tertium," "Kyrie secundum," etc.) specify exactly how the works are to be performed. Of the other paraphrase Mass settings of the period, only a few isolated movements imitate Dufay's use of chant and polyphony.38 Not until the latter part of the fifteenth century-in the Masses of Heinrich Isaac-is the alternatim style applied systematically to all the movements of the Mass Ordinary.39 The general principle of chant alternating with polyphony is, of course, much older than Dufay's time: it occurs in the Chartres Alleluias last section a sense of climax. Beyssac further points out that in the prosa normally associated with Kyrie II, "fons bonitatis," the last three verses are all the same length and all equivalent to the modern Kyrie 9. 37 This view is developed in Christhard Mahrenholz's article "Alternatim," MGG, Vol I, col. 393. 3s The only other contemporary alternatim Kyrie I know of is Brassart's setting of Kyrie XII; for a modern edition, see Johannis Brassart: Opera omnia, ed. Keith Mixter, Corpus mensurabilis musicae, 35, Vol. I (n.p., 1965), pp. 65-66. Alternatimstyle settings of other Mass sections include a Gloria IV by "Aedvardus" in Perugia, Bibl. Aug., MS 431, No. 29; an anonymous Credo IV in Tr87, No. 179 (mod. ed. in Sieben Trienter Codices. Fiinfte Auswahl, Denkmiler der Tonkunst in Osterreich, Jg. XXXI [Vienna, '9331, No. 9); and an anonymous setting of an unidentified mensural Credo melody in Cambrai, Bibl. de la Ville, MS 6, No. 14 and MS ii, No. 22. In addition, an anonymous Gloria IX in Trgo, No. 924 and Tr93, No. 1734, consists of alternate sections only-the same plan as in Dufay's Gloria IX, which it followsand this may indicate alternate performance even though no chant is included in the sources. 39 See, for example, the five four-voice Masses from the Choralis Constantinus (modern edition in Heinrich Isaac: Five Polyphonic Masses, ed. Louise Cuyler [Ann Arbor, 19561). See, also, the eight five-voice Masses in Munich, Bayerische Staatsbibl., Mus. MS 3.
244 JOURNAL OF THE AMERICAN MUSICOLOGICAL SOCIETY (eleventh century) and in Leonin's Magnus Liber (twelfth century). Within the specific area of the Mass Ordinary, however, the only exam- ples to be found prior to the fifteenth century are polyphonic tropes, evidently intended for insertion in monophonic chants (e.g., the Sanctus and Agnus tropes in W, and elsewhere and the "Spiritus et alme" tropes preserved in numerous manuscripts).40 Polyphonic Kyries from the fourteenth and early fifteenth centuries often include prosas, with all nine sections of the text usually accounted for in the polyphony.41 Even Kyries without prosas frequently make provision for nine sections of polyphony.42 It might seem strange that the general principle of alternation could have been in use for many centuries without being applied to the Kyrie. The eleventh- and twelfth-century examples cited above, however, are solo polyphony used in conjunction with responsorial chants (solo- chorus); Dufay's works are presumably choral polyphony, and the structure of the Kyrie lends itself to the antiphonal style of performance that results (chorus-chorus). It might be noted that much of the other alternatim-style music that appears in the course of the fifteenth century is built around antiphonal chants: Introits, Hymns, Sequences, Magnificats. With these, Dufay's Kyries have many traits in common: they are relatively simple in style, occasionally making use of fauxbourdon; they are composed in cycles to accommodate the various feasts of the church year; and they make consistent use of the newly developed technique of discant paraphrase, an important stylistic effect being the appearance of parallel passages both in plainchant and in polyphonic elaboration. The Kyries are relatively little-known members of this body of liturgical music, but they deserve to be singled out as among the earliest examples both of a compositional type (chant elaboration) and of a specific per- formance procedure (alternatim style), which remain important in sacred music throughout the sixteenth century and beyond. University of Windsor 40On the latter, see Denis Stevens, "Polyphonic Tropers in i4th Century England," Aspects of Medieval and Renaissance music: A Birthday Offering to Gustave Reese, ed. Jan LaRue (New York, 1966), pp. 768-84. 41 See the examples in Fourteenth-Century Mass Music in France, ed. Hanna Stiiblein-Harder, Corpus mensurabilis musicae, 29 (n.p., 1962). Early fifteenth-century Masses by Dufay and Arnold de Lantins also fall into this category. 42 Dufay's Missa Sancti Jacobi, for example, as well as the Marian Masses of Lantins and Lymburgia in BL and Liebert's Marian Mass.