Hendrik Schulze, The Manuscript Score as a Document of Performance Practice in Cavalli's Operas 1654–1661 :: Philomusica on-line :: Rivista di musicologia dell'Università di Pavia


Contributo di Hendrik Schulze


The Manuscript Score as a Document of Performance Practice in Cavalli's Operas 1654–1661



Some years ago I presented two papers at conferences in Dublin in 2000 and Naples in 2002. Both contained my findings concerning the manuscripts of Francesco Cavalli’s operas Xerse (1655), Statira (1656), Artemisia (1657), and Ercole amante (1662) as performance material. Recently they have appeared in an article in the proceedings of the Naples conference, which are edited by Dinko Fabris.[1] Building on my findings, in this present paper I will attempt to evaluate what exactly the scores in question have been used for, trying to establish a chronology of events. I will also present some thoughts on the implications on our subject at hand, problems of editorial practice of Baroque opera in the light of its performance. For reasons of time, I will limit myself on talking about Xerse and Artemisia only.

I will therefore begin by giving you a brief summary of my findings, and illustrate some aspects with new examples, considering each opera separately. Finally, I shall present my thoughts on how to handle those aspects in a modern edition.

The main body of extant scores of Cavalli’s operas is preserved in the Contarini Collection, housed at the Biblioteca Nazionale Marciana in Venice. They form part of the composer’s legacy, having been destined by him to be preserved for posterity. As Jane Glover[2] and Peter Jeffery[3] have demonstrated, most of those scores are fair copies, made in the 1660s and 1670s, presumably for preservation purposes. However, the majority of scores of those operas that were first performed between around 1650 and 1660 are autographs, seemingly preserving various stages of composition and performance in irritating complexity. It is to this group that the scores under discussion belong.



Xerse, composed on a libretto by Nicolò Minato, was first performed at the Teatro SS. Giovanni e Paolo at Venice on January 12, 1655. As far as we know, this performance was directed by Cavalli himself. Subsequent productions took place at Genoa in 1656, at Bologna and Naples in 1657, at Palermo in 1658, and at Paris in 1660, the last-named one again under Cavalli’s direction. After that, there were productions at Milan and Verona in 1665, at Turin in 1667, and at Cortona in 1682.

There are three scores of Xerse still in existence, one each at Venice, Rome, and Paris. By far the most interesting as regards of performance practice is the one from the Contarini Collection in Venice, the one Cavalli actually owned and used.[4] Although its core is in large parts a copy, made by several copyists whom Peter Jeffery has labelled X1, X2 and X3,[5] its first gathering was directly taken out of Cavalli’s composing score. In addition, it has been heavily revised both by Cavalli himself and by two other hands, Jeffery’s X3 and H1. When, in 1970, Martha Clinkscale prepared her edition, she came to the conclusion that the score in its entirety represented the final draft for the Bolognese production of 1657. This finding seems to be corroborated by the fact that a second extant score, a fair copy, preserved today in the Chigi Collection in Rome, represents exactly the same version, thus supporting the idea that this was indeed a fixed version to be produced at a certain date, and that perhaps the Roman score was prepared for use in that performance.

However, after closer inspection of the scores it must be said that the situation is much more complicated. To be able to evaluate them, one has to take into consideration what purpose these scores served.

According to Clinkscale, the Venetian score of Xerse was some sort of model to the Bolognese production, presumably to be copied for the performers. But this assumption does not bear closer inspection, for the score shows clear signs of having been used for actual performance. Most of the revisions are indeed not changes to the opera but clarifications and corrections, improving legibility and eliminating copyist’s mistakes. Interestingly, these corrections mostly concern the continuo part, whereas the legibility of the upper parts is sometimes difficult. In some instances, mostly in instrumental pieces, but also in choruses, the upper parts are left out entirely.

So the score clearly served as a performing score for a continuo player. Since most of the revisions involving the continuo part are in Cavalli’s hand, it is safe to assume that it was Cavalli himself who used this score to direct rehearsals and performances of Xerse from the harpsichord, as he usually was bound to do by contract. In fact, it can be shown that the first series of revisions seems to have brought the score right up to the version represented by the Venetian libretto printed for the performance in 1655 – i.e., the version we may expect to have been performed when the opera first opened. Thus the initial purpose of this score was to be used by Cavalli already in the 1655 production.

Apart from this first layer of alterations, however, there are at least four other layers of revisions – not three, as I wrote earlier – all made from the perspective of a continuo performer and leader of the ensemble. The first one, instigated also by Cavalli, involves some cuts, as well as transpositions of certain parts. Its main purpose seems to have been to accommodate an entirely different cast; it was thus clearly designed for a different production, which nevertheless was led again by Cavalli. Judging from the fact that the third layer of revisions was the one that represents the Bolognese production, the second production almost certainly was the Genoese one – that is, unless there was another performance of Xerse unknown to us, Cavalli must have been in Genoa in 1656 to direct his opera there.

As I said, the third layer represents the changes made for the Bolognese production in 1657. Interestingly, it was not Cavalli but another continuo player that seems to have directed these performances. The revisions are not in Cavalli’s hand but in that of Jeffery’s X3 and H1 – the latter being a hand found also in the Venetian score of Hipermestra, which was composed in 1654 and performed in Florence in 1658. Apart from cuts and minor modifications, they involve the exchange of an intermedio after the first act. This was done by crossing out part of the original tournament scene and inserting in its place an aria from the opera Artemisia, which was first performed earlier in the same year (1657).[6] In addition, the "Coro di spiriti" in the first scene was reduced from four to two voices. This entailed a reworking of the bass line, which is given at the foot of pages 5 recto and 5 verso.

It is in the hand of Jeffery’s copyist X3, like most of the revisions of this layer. Who was the player for whom these revisions were made? It is worth noting that the continuo player for this production, whom I have labelled "Continuo Player X", had needs different from those of Cavalli when directing an ensemble. While Cavalli apparently did not have any problems turning back pages when there was a second strophe to be sung, Continuo Player X seems to have been uneasy with this. He needed to have at least the bass line copied out for a second time.

He also differed from Cavalli in that in recitatives he had to have the notes of the bass accurately aligned vertically with those of the vocal part against which they sounded, whereas Cavalli apparently did not care about this; on the other hand, Cavalli seems to have taken more care about the correctness of texts the singers sung. This suggests that Cavalli was concentrating more on the text and its rhythm when accompanying recitative, whereas Continuo Player X focused more on the musical rhythm.

Most of the cuts for the Bolognese revision were marked not only by crossing out the passages to be cut but also by pasting sheets of paper over them. The fact that some of the cuts in this layer dispense with a preliminary crossing out and have been marked only by pasting over suggests that the latter process occurred very late during this stage. It clearly represents an attempt to make the already much cluttered score more legible for the performer.

Insertions for this revision were not always added to the score: an annotation on page 84v. (85v. in the old pagination) reads: "Qui và la scena aggiunta ad Eumete". This refers to a solo scene for the named character that appears in the Bolognese libretto in that place, but of which there is no further trace in the scores – neither in the Venetian nor in the Roman version, interestingly.

A further layer of revisions has been marked in a similar way by pasting over. This layer is not apparent in the Bolognese libretto but stays very close to it in its content, merely eliminating a few more passages, second strophes of arias, and ritornellos. Most of those cuts were never crossed out. Although I have yet to check the Neapolitan libretto, I strongly suspect that these revisions were made for the 1657 production in Naples, which we know was quite similar to the Bolognese one. It took place a few weeks after the Bolognese production, probably once again under the direction of Continuo Player X and with the same cast.

For the final layer of the Venetian score of Xerse, someone took out again all the sheets pasted over the cuts of the Bolognese and Neapolitan productions. In addition, some of the crossed-out material seems to have been inserted again, and some of the transpositions reversed. An example of a revision of this kind appears on page 128r. (129r. in the old pagination), where the instruction "si suona" is written next to a section of an aria that had earlier been crossed out for the Bolognese production. It is in a hand with which I myself am not familiar. However, part of the same revision looks suspiciously like Cavalli’s handwriting again. It is an insertion of recitative on page 180v. (181v. in the old pagination), which is written partly over the glue from the Bolognese revision, thereby demonstrating that it must indeed have occurred after the Bolognese performance.

During the same process, some of the added material for Bologna and Naples has been removed again, this time by simple deletion.

I have yet to trace the production for which these revisions were made. However, I have to admit that the chances are only slight that we will identify that production, since it is clearly not the Paris production, which is very different from what this layer represents, and is unlikely to be any of the later productions known to us, since these seem not to have involved Cavalli at all. At this point, we have to assume that there were preparations for another production after the Naples one, in which Cavalli was involved. But this production may never actually have taken place, hence the lack of a corresponding printed libretto for it.

To sum up the Venetian score of Xerse, here are the main points again:

1. The score was used primarily by Cavalli to perform on the harpsichord, and to direct rehearsals and performances.

2. It was used for at least four different productions and to prepare a fifth, which may never have taken place.

3. All those productions entailed heavy revisions, some of which were clearly necessary because of the different needs of a new cast.

4. The third production was the one at Bologna in 1657, which was directed by someone other than Cavalli, the mysterious Continuo Player X.

5. None of the productions apparent in the layers of the Venetian score was the one at Paris. Since that production was directed by Cavalli, too, he must have had recourse to another score for it.

Now let us briefly turn to the other two extant scores of Xerse.

In the case of the score from Rome,[7] the identification of its purpose is simple. It is a very neat and ornate fair copy, virtually free of corrections and without the signs of heavy usage that abound in the Venetian score. Clinkscale’s assumption that it represents the Bolognese version is a misunderstanding. Rather, it accumulates all the material contained in the Venetian score up to and including the layer of the Bolognese production; for example, it contains both the intermedio from the Venetian production and the one from the Bolognese production in the same sequence as they appear in the Venetian score. Although the first intermedio was not given at Bologna, neither of the two is marked as having been omitted. On the other hand, the solo scene for Eumete that was performed at Bologna is missing from this score, just as it is missing from the Venetian score. In other words, the Roman score represents the surface of the Venetian one without distinguishing between the different layers. Hence it does not at all represent a definite version of the opera but is, rather, a mish-mash of different stages, corrections and revisions, all based on the individual, and therefore incomplete point of view of the conductor’s score. The process of rehearsing and performing, as documented in the Venetian score, is lost in the Roman one.

The hand of its scribe differs significantly from that of any scribe involved with the Cavalli scores of the Contarini Collection. In addition, the layout differs from that of any Venetian score but is similar to other Roman scores. It would thus be safe to say that this copyist was not closely involved with Cavalli. Most probably, he was Roman. Since it was certainly the copyist who had to make the decisions when his source was ambivalent, most of the Roman score’s inconsistencies vis-à-vis the Venetian one are misreadings by the copyist and have nothing to do with any performance matters, let alone with conscious revisions made by the composer or someone close to him.

As a dedicatory exemplar, the function of the Roman score seems to have been purely commemorative, which means that nobody was expected to perform from it, and hence it did not need to be fully accurate. That explains why the Rome score contains several obvious mistakes that have not been corrected.

The same applies to the Paris score,[8] which is clearly a French copy commemorating the Paris production. There are no traces of it having been used for performance. Unlike the extant score of Ercole amante, the Paris Xerse score is thus not the score used by Cavalli for performing in Paris. It does, however, contain not only Cavalli’s opera but also the complete music for Lully’s ballets, which were given between acts. Since it was intended to commemorate a definite production, this score may be expected to represent that production as a definite stage in the opera’s evolution, even if it gives no information about the process of production. But even in this case there is no way for us to know for certain.



The source situation of Artemisia is slightly less complicated. The opera was first performed on January 10, 1657, at the Teatro SS. Giovanni e Paolo at Venice. Subsequent productions took place at Palermo and perhaps at Naples in 1659, at Milan in 1663, and at Genoa in 1665. None of these subsequent productions seems to have involved Cavalli.

There is one extant score, belonging once again to the Contarini Collection at Venice.[9] It is composite, the first two acts being basically Cavalli’s composing score, whereas the third act is a fair copy.

Again, the first two acts can be shown to have been used as the performing score from which Cavalli directed all rehearsals and performances of the 1657 Venetian production. Within this process, we may distinguish four phases.

1. Composition: in the order of libretto; few corrections.

2. Pre-rehearsal revision: strengthening dramatic structure by adding arias, rearranging scenes and adding funnier lines. Altering pitch for two singers. Copying of the parts.

3. Rehearsal: some cuts, some alterations to singer’s pitch, two arias newly set with instrumental accompaniment; elimination of character Cleante; score copied for Continuo Player A.

4. Performance: minor cuts.

The first one is the actual composition, which seems to have been a straightforward affair. By then, there must have been a finished version of Minato’s libretto available, and Cavalli simply composed what was there in the order of the drama.[10] From the beginning, he obviously had clear ideas about what he wanted to do. Only on rare occasions did he have to cross out a system or two that he just composed, mostly in order to give a different harmonic direction to a section of recitative.

The second phase, which was still pre-rehearsal, involved a reworking of the opera. One of the goals of this reworking was to improve the dramatic structure by rearranging some scenes, adding funnier lines, and supporting a structure of arias that was designed to facilitate the audience’s understanding of the complicated plot.[11] The other goal was more practical; it involved the changing of pitch for two characters. After that, the score was copied for the performers; there were individual parts for every singer, and perhaps also for the instrumentalists.

The third phase was that of the rehearsals. During this phase, some cuts were made, as well as a few transpositions for the singers (especially for one of them, a soprano castrato who sang the role of "Ramiro" – apparently, he could sing higher than expected, and Cavalli took advantage of that). In addition, at least two ordinary continuo arias were reset with a thorough-going violin accompaniment, and one was reworked to contain echoes. Finally, a character was cut altogether; he was to all intents and purposes redundant anyway, but several passages of the opera had to be reworked in consequence. Then the libretto was finalized and printed. At the end of this phase, the entire opera seems to have been copied out for the second harpsichord player, whom I shall name Continuo Player A.

The reason why I feel confident to make this last statement is that the third act, which is a fair copy, was obviously used as a performing score for a harpsichord player other than Cavalli. Since it represents perfectly the version given at Venice in 1657, as documented by the printed libretto, while matching none of the other productions known to us, I feel safe in claiming that it was indeed intended for use at this performance. It displays none of the characteristic modifications made during the first three phases I mentioned and therefore must have been copied later. It does, however, show the same characteristics as the revisions of the fourth phase, which is the period of performance. So it also appears that Continuo Player A did not take part in most of the rehearsal period, joining the ensemble for the performance period only.

During this period there have been some minor cuts, mostly marked in red crayon. These cuts do not appear in the printed libretto. The red crayon itself makes it difficult to decipher who wrote what, being too thick to write anything but strokes and stars, which are not very distinctive. They appear all over the score and look quite similar both in Cavalli’s part and in that of Continuo Player A.

But the scores of the first and second act differ from the third act in another fundamental respect – that of the characteristics of the continuo players. In fact, here we encounter the same differences that distinguished Cavalli from Continuo Player X in the Venetian score of Xerse. Like him, Continuo Player A needed to have exact vertical alignment of the bass line in recitatives. In arias, he also needed to write out the bass line of the second strophes in those cases that otherwise would have necessitated turning back the pages.

It is tempting to say that on account of these similarities Continuo Player A is indeed identical with our mysterious Continuo Player X, the more so as Continuo Player X was to introduce material from Artemisia into the Bolognese production of Xerse only a few months after the 1657 Artemisia premiere. Unfortunately, there is no trace of his handwriting in Artemisia, so we cannot be certain of this. However, if they were indeed the same person, I think there is a fair chance that we could identify Continuo Player X and Continuo Player A as Giovanni Battista Volpe, who is known to have regularly played second harpsichord for Cavalli throughout the 1650s, and to have stood in for him if for any reason Cavalli was not able to direct a performance himself.[12] But at this point, this is mere speculation.



What may we learn from these discoveries in Cavalli’s scores? Concerning our subject at hand, problems of editorial practice of Baroque opera in the light of its performance, I would like to make three points.

The first point is an obvious one. However, since some recent scholarly editions of 17th-century Italian operas seem not to be aware of it, I thought it worthwhile to mention it here. When editing a seventeenth-century opera, one should always, and at an early stage, determine what purpose one’s sources served. As we have seen in the case of the Roman score for Xerse, a dedicatory exemplar may preserve a mish-mash of different versions without differentiating between them. The result is the image of an opera that on the page looks perfect, but was never intended to be performed the way it appears in the score – for instance, in the Clinkscale edition of Xerse there appear two intermedi as scene 20 of the first act that have nothing to do with each other, and were originally intended for different productions. In René Jacobs’s recording of the opera, which is based on Clinkscale’s edition, we therefore encounter the claim that he is in some sense recreating the Bolognese 1657 production, thus claiming authenticity for his version, even though that is not the case. By taking the Roman score as a sort of "Fassung letzter Hand", the Clinkscale edition has lost all of the Venetian score’s numerous documentations of the original performance practice, the knowledge of which would be so beneficial to a modern musician.

At least in the case of Xerse and Artemisia, we still possess these messy autograph scores that yield so much information. In most other cases, however, we have only fair copies, sometimes of unclear origin – and this would be my second point. A good example is Cavalli’s Doriclea of 1645, which recently appeared in a modern edition prepared by Christopher Mossey. The sole extant source for this opera is a copy originating from the workshop of Cavalli in the late 1660s or early 1670s, at a time when nobody would expect this opera ever to be performed again. By taking the extant score at face value as that of the original version, Mossey ignores the purpose that his source served initially. As a result, he presents as a definite version a score which has a lot of characteristics similar to the Roman score of Xerse, which is at best misleading. Just to take one example, the five-part instrumental pieces that appear throughout the score are certainly not from the original 1645 version; the two middle parts appear to have been added at the time when the opera was copied, perhaps in order to make the composition look more opulent and more modern than it was in reality. The added parts look suspiciously like someone’s – perhaps the copyist’s – counterpoint exercise and do not at all enhance the musical value of the piece. After all, they were never intended to be played. They were there just to fill the space. Despite numerous attempts on part of modern performers to validate them, they do not justify the claim of authenticity made by editors and performers when adding instrumental parts for the sake of greater sonority or loudness to Cavalli’s originally three-part sinfonias and ritornellos, not to mention arias with instrumental accompaniment.

My third and final point concerns the incompleteness of the autograph scores. Despite having been used as performance material, they do not include all that was performed in the way it was performed. For instance, as can be shown in the case of Artemisia, transpositions made during rehearsal appear in the score only when they involve the bass part as well. Simple alterations of upper parts that do not involve an overall transposition do not appear in the score, since they were of no consequence to the continuo player who performed from it. Thus this kind of score is most reliable on the continuo part and less so on the upper parts. It is also bound to respect the order of the piece as it was performed, even if it only contains such tantalizing remarks as "Qui và la scena aggiunta ad Eumete".

To conclude, I would therefore argue that modern editions that would be most useful to both scholars and performers should concentrate on those autograph scores that can be shown to have been used for performance. While none of the types of extant scores discussed above can be said to be in any way complete, the autograph performing score at least provides information about the process of producing an opera and about its various stages and the normal types of revision it regularly had to undergo during this process. It will, to be sure, need a great deal of explanation and a large critical apparatus, but its value to both performers and scholars will be infinitely higher than a simple diplomatic reproduction of a fair copy, for which a facsimile may perhaps serve just as well.



[Bio] Ha studiato musicologia a Berlino, Princeton, Ferrara e Heidelberg; si è addottorato presso l’Università di Heidelberg con una tesi sulla figura di Odisseo nell’opera veneziana del Seicento. Dal 1999 al 2005 è stato assistente presso l’Università di Salisburgo. Attualmente è ricercatore presso l’Università di Heidelberg, dove collabora a un progetto sulla danza e la musica per danza nell’Ancien Régime. È il corrispondente per l’Europa della Society for Seventeenth-Century Music.

[1] Hendrik Schulze, "Cavalli Manuscript Scores and Performance Practice," in Francesco Cavalli. La circolazione dell’opera veneziana nel Seicento, ed. Dinko Fabris (Naples: Turchini, 2006), 39-58.

[2] Jane A. Glover, Cavalli (London: Batsford, 1978).

[3] Peter Jeffery, "The Autograph Manuscripts of Francesco Cavalli" (diss., Princeton, N.J., 1980) .

[4] I-Vnm, Cl. It. IV, 374 (=9898)

[5] Jeffery, "The Autograph Manuscripts of Francesco Cavalli."

[6] Martha N. Clinkscale, "Pier Francesco Cavalli’s Xerse" (diss., Minneapolis, Minn., 1970).

[7] I-Rvat, Chigi Q.V. 64.

[8] F-Pn, Vm 4/2.

[9] I-Vnm, Cl. It. IV, 352 (=9876).

[10] There seem to have been instances where Cavalli had difficulties in reading Minato’s longhand – sometimes he had to go back on the text to alter some words which in context did not make sense.

[11] Cf. Hendrik Schulze, "Plot Structure, Characters, and Aria Position in Francesco Cavalli/Nicolò Minato’s Artemisia (1657)," Musica & Storia 12 (2004): 91–102.

[12] Which he did for instance for the entire 1654/1655 production of Erismena; see Beth Glixon and Jonathan Glixon, Inventing the Business of Opera. The Impresario and His World in Seventeenth-Century Venice (Oxford - New York: Oxford University Press, 2006), 351.

Torna all'inizio della pagina

Firenze University Press

 Copyright 2005-06 © Università degli Studi di Pavia
 Dipartimento di Scienze musicologiche e paleografico-filologicheFacoltà di Musicologia

Registrazione presso la Cancelleria del Tribunale di Pavia n. 552 del 14 luglio 2000 – ISSN elettronico 1826-9001 | Università degli Studi di Pavia Dipartimento di Musicologia | Pavia University Press

Privacy e cookies