a. Copying from Exemplars: Textual Criticism and Notationless Exemplars

Perhaps the most obvious (and certainly the oldest) hypothesis for where the makers of chansonniers found melodies and texts is one that posits early trouvère sources already in circulation by the time the first surviving chansonniers were copied. This basic premise describes a variety of situations. The most widely circulated idea, originating from Gustav Gröber, is that individual leaves or groups of a few leaves of music and text must have been in circulation during the lifetimes of the first troubadours and trouvères.Footnote ³³ These song leaves (‘Liederblätter’) were copied out and circulated by the troubadours, trouvères and performers contemporary with them. Eventually, these exemplars were collected into increasingly large and increasingly well-organised fascicles or books (‘Gelegenheitssammlungen’), culminating in large, ornate chansonniers. In a handful of cases, Gröber believed songbooks were collected or directly influenced by the troubadours who had composed them, resulting in single-author volumes he dubbed ‘Liederbücher’.Footnote ³⁴ Hand in hand with Gröber’s model went the work of Eduard Schwan on the trouvère chansonniers.Footnote ³⁵ In Gröber’s view, the argument for transmission through ‘Liederblätter’, ‘Gelegenheitssammlungen’ and ‘Liederbücher’ would need to be tested by comparison among multiple surviving sources.Footnote ³⁶ Schwan undertook this type of work for the French trouvère sources, constructing a stemma codicum of manuscript families based on the comparison of errors, variants, author attributions and contents. The publications of Gröber and Schwan together constitute a corpus of evidence that has remained relevant for editors and has also provided the inherited narrative against which subsequent scholars could set up their own opposing views. Together they offer a model for the entire process of chansonnier transmission, from the moment of creation, to first transcription, to collection in surviving chansonniers. Gröber’s focus on the troubadours, rather than the trouvères, expands his time range backwards, so that he posits continuous written transmission starting at least as early as Arnaut Daniel (second half the of the twelfth century).Footnote ³⁷ The last part of that narrative, the arrival of songs in the chansonniers that survive, is more tractable, given the evidence, and that is what this article examines.

Leach, Lug and Haines’s work pushes beyond asking whether lost sources existed (they clearly did) and towards the work of describing their size, material (wax or parchment?) and function (transmission or planning?). Were written copies of songs collected in order to make multiple chansonniers as part of the same project?Footnote ³⁸ Might there have been a number of uniform stock exemplars kept by professional book-makers or puy members for the express purpose of copying, analogous to the pecia system in use for university books?Footnote ³⁹ Or were exemplars scarce and collected together with difficulty?Footnote ⁴⁰ And did exemplars contain both text and music as most trouvère chansonniers now do, or did these predecessors resemble the situation for troubadour song, with music copied infrequently?

The stakes of this last question are high, both for what we are to make of the surviving trouvère melodies and for how we imagine the circulation of written song sources in the thirteenth century. Leach has argued that the ‘very small ephemeral materials’ used by the copyists of manuscripts I and C must have contained some melodies, despite the lack of notation in the two sources copied from them.Footnote ⁴¹ In her view, ‘I chose to omit the notation’, whereas C’s makers were simply unable (owing, one assumes, to time or financial constraints that prevented the hiring of a skilled notator) to finish the project of filling its empty staves.Footnote ⁴² However, Leach makes it clear that not all of the ephemeral materials had notation for every song, particularly in one example in the jeux-partis section of I.Footnote ⁴³ Being able to show that a particular song did or did not have notation in an exemplar is a remarkable result, particularly when that lack or presence of notation does not correspond to the situation in the manuscript copied from it. Using the same exemplars for both text and music, as seems to have been the intention behind C, surely would have been more convenient. However it is clear that was not always possible. Elizabeth Aubrey has suggested, in her consideration of troubadour chansonnier R, that some text scribes were overly optimistic in leaving space for staves because they had no responsibility for finding or copying from notated manuscripts.Footnote ⁴⁴ In section IIb , I consider an instance where the makers of a surviving chansonnier that does contain notation had recourse to an exemplar that lacked it. I conclude that the music notator relied on a notated source that the text scribe lacked, or instead relied only on musical memory to notate the blank song. In light of this example, the rest of the article operates under the assumption that the existence of lost text exemplars cannot be taken as evidence of lost notated exemplars.

Let us now turn to a song that may well have circulated in a text-only exemplar, perhaps even in a single-author collection. Tout autresi con descent la rousee, RS 554 / L 199-10, is one unicum among many in manuscript V. Yet we must be careful not to imagine it as marginal on that account. It follows sections of songs by the major trouvères Thibaut de Champagne and Gace Brulé and of other composers whose pieces appear in a similar sequence in other sources: Gillebert de Berneville, Richard de Semilli and the Vidame de Chartres. Many of the unica in this particular section have been attributed to Philippe de Remi, author of various narrative poems, fatrasies, lais, and saluts d’amour and father of the jurist of the same name. This song explicitly portrays itself as being from the mouth of Philippe in lines 5–8 of the envoi:

In the absence of concordances in other manuscripts, or a marginal attribution in V, this is the most solid evidence we have of the song’s author.Footnote ⁴⁶ There is thus reason to associate the song with non-lyric traditions of poetry that circulated exclusively without notation.

To understand the process of copying a trouvère song, we must consider the shape of that song. The verse scheme of Tout autresi is standard. Two feminine rhyme sounds govern the entire poem, the a-ending (-ee) and the b-ending (-age), as demonstrated by the first stanza, reproduced below. Apart from the first stanza, where poetic line 5 is deficient by four syllables, each poetic line is decasyllabic.Footnote ⁴⁷ The eight-line rhyme scheme may be represented as follows: ababbaab (the text of the first stanza, below, is labelled accordingly). In this repertoire, the typical alternation of a and b rhymes in the first four poetic lines usually aligns with musical repetition to create parallel pedes, whereas the rest of the melody continues without significant repetition to the end. With Tout autresi, our knowledge of the musical form is hampered by the omission of an entire poetic line’s worth of music due to a scribal error.

The song begins at the end of the right-hand column of fol. 52^v (Figure 1). As is usual in all trouvère chansonniers, only the music for the first stanza is notated, with the text of subsequent stanzas (the residuum) following it in prose format. For this same reason, the music of Tout autresi follows directly on the text for the final stanza of the previous song. The scribe reached the end of the page on the first word of poetic line 3 (given in bold in the transcription above) and continued copying across the opening, on fol. 53^r (Figure 2). But here, in the middle of the poetic line, the standard chansonnier format of text and music together gives way to chaos. The text scribe neglected to leave space for the music at the top of the new folio and, for three lines, reverted to prose format. There are still six poetic lines left in the stanza and all of them need staves and music above them. When the scribe realised the problem, there was already text where those staves should have been.

Figure 1 Trouvère V, fol. 52^v: Beginning of [Philippe de Remi], Tout autresi

Figure 2 Trouvère V, fol. 53^r: Continuation of Tout autresi

Undoing the damage thus posed a challenge. Rather than immediately erasing the extra text, the scribe first repeated the text already copied, this time in a more appropriate position, after a space allocated for staves and music. It is no coincidence that the three lines of staffless text occupy exactly the space one staff would have taken up; it is exactly one staff of music that is still missing. The scribe intended to bring the first line of text, ‘bonne amour par une estroi-’ down two rulings and place a staff above it by erasing the intervening redundant text. However, as the attempted erasure of ‘et le courage’ shows, the correction was more easily planned than done. Erasures are always dangerous on parchment, and this folio is more than usually delicate. The scribe probably showed foresight in deciding against an erasure. Even without scraping, a hole opened in the third staff, over ‘est’. This must already have been present when the staves were drawn, as the staff terminates a little before the hole and starts again with a safe margin of distance (see Figure 2). The text on the verso side (Figure 3) displays similar adjustments, such as the use of abbreviation, in order to leave space around the hole. The decision not to erase may well have been based on the existence of physical damage on the same folio.

Figure 3 Trouvère V, fol. 53^v: Reverse of the hole in the music of Tout autresi

This is a scribe familiar with planning music notation, and their attempts to remedy a layout catastrophe are interesting in themselves. Still more interesting is the question of how a scribe who was expecting music notation could make such a mistake in the first place. There are so many ways that we know that music was always meant to continue (from the structure of the stanza, the layout of other sources, even the shape of the melody) that it is difficult to imagine how the scribe could not have known. We can only assume they lacked access to all those signals. The scribe must not have been thinking very hard about the form and order of the stanzas while copying them (which probably rules out transcription from memory). But it is also hard to imagine the scribe forgetting to leave room for staves while looking at an exemplar that contained notation.

It might be that the scribe began writing on fol. 53^r without looking back at the exemplar. But this would imply that the scribe had only just seen the staff spacing on fol. 52^v when they made their mistake. This would mean that something (such as dipping the pen in ink) distracted the scribe enough to make them forget staves were needed, but not enough to dislodge around two lines of text (a sensible transfer unit) from their memory. This turn of events is plausible, but the simpler possibility is that the scribe was copying from a source containing only text. Imagine that the scribe paused in the act of writing upon reaching the midpoint of the gathering, the end of fol. 52^v. On returning to the task, they found the leaving-off place marked precisely in the text-only exemplar, and blank ruled lines on the column onto which it was to be copied. The result is what we see.

The example suggests that music and text derived from different sources for at least this section of manuscript V. Those sources might conceivably have been leaves, rolls, or collections devoted to Philippe de Remi, perhaps even dating to the author’s own lifetime, though the attribution of the entire group of songs to Philippe remains speculative.Footnote ⁴⁸ The example serves to demonstrate what scholars have long known: trouvère chansonniers, even when organised into purportedly single-author sections, are messy affairs, with music and text arriving in them by separate routes. If the hypothesis of a text-only exemplar is correct, text and music followed different paths of transmission up until the moment of copying. The synthesis of text and music in V represents one possible combination of variants, perhaps one that was realised numerous times in parallel lost manuscripts and performances. What is sure is that we can identify one such moment of synthesis in V.

Further examples of absent music in other chansonniers, catalogued in Tables 1 and 2, suggest something similar, as the text scribes who left space for missing melodies evidently failed to predict whether it would be possible to add music later.Footnote ⁴⁹ Some of these empty staves may reflect situations similar to what we have seen in V. The evidence there rules out the use of a single notated collection as an exemplar for the Philippe de Remi section, calling into question the role of ‘Liederbücher’ in transmission and collection of music, and demanding that we use caution when assuming that evidence of lost written exemplars implies the existence of lost notated exemplars.

b. A New Leaf: Early and Late Notated Sources

What does the circulation of text-only exemplars mean for the lost sources behind other trouvère chansonniers? The exceptionally early dating of manuscript U and its high number of empty staves make it particularly worth revisiting. The earliest date, 1231, has been proposed by Lug, another neo-Gröberian and the reigning expert on U.Footnote ⁵⁰ For Lug, quirks of manuscript organisation prove the existence of sheets containing pairs of songs, used as exemplars for the earliest layer of U and for chansonnier C.Footnote ⁵¹ Taken together, Lug’s arguments imply the existence of written trouvère sources by the year 1231. If these small exemplars also contained notation, we can then take this as the start date for copying trouvère melodies between manuscripts, the terminus ante quem for a history of vernacular song’s transmission via notation.

Such a history would be fraught with uncertainty. The dating of most trouvère chansonniers is insecure. Lug’s dating of U has been contested.Footnote ⁵² Haines’s dating to the 1250s for M relies on the likely supposition that it was commissioned by Charles of Anjou for William of Villehardouin, not universally accepted.Footnote ⁵³ Yet these are two of the most securely datable trouvère manuscripts. Stones’s dating of manuscript a relies exclusively on art-historical grounds, placing its illuminations earlier than A’s and both manuscripts probably before 1297.Footnote ⁵⁴ The date of manuscript P has often been listed as around the 1270s, along with K, N and X.Footnote ⁵⁵ Its contents include songs by relatively late trouvères such as Colart le Boutellier, Gillebert de Berneville, Perrin d’Angecourt, yet no song can be definitively dated past the 1250s and most are also present in manuscript M. The inclusion of Adam de la Halle’s songs in manuscripts A, R, V and a tends to push them later, since his activity as an author is known to extend from the 1260s into the 1280s. Still, none of the individual songs in these manuscripts offers an indisputable terminus post quem in this range.Footnote ⁵⁶ Mary O’Neill has attempted to divide manuscripts into phases on the basis of musical palaeography, and according to her categorisation M, P and U fall into an earlier group than A, R, V and a.Footnote ⁵⁷ For V, further evidence exists in the form of coats of arms on the opening folio of the manuscript. These have been tentatively identified as belonging to the Artesian families of Bernastre and Varrennes as they would have appeared near the end of the thirteenth century or early fourteenth century, with the caveat that such armorial evidence remains equivocal.Footnote ⁵⁸

A definitive chronology of trouvère sources thus remains out of reach. What can be asserted with some confidence is that manuscript V is at least a few decades more recent than the earliest layer of U and that A, M, T and a likely fall somewhere in between. Any differences between U and V in terms of the information they furnish about lost sources therefore become interesting from a historical perspective. If we could find consistencies between the two, we could conclude that lost notated sources were a common feature in the production of trouvère manuscripts from the very beginning, and the discussion would be closed. Instead, we find differences that provide clues to a developing written culture in which manuscript makers adapted their habits to circumstances, changing them in the middle of the preparation of a single manuscript.

Empty staves in U have been variously catalogued and explained, usually as reflecting the limited knowledge, time and resources of the scribe.Footnote ⁵⁹ The later units of U are particularly limited in what notation they offer. The one song melody they contain is likely to have been entered late in the manuscript’s history.Footnote ⁶⁰ Table 3 offers a breakdown of the different codicological sections of U, informed by the work of Madeleine Tyssens and Christopher Callahan, along with the empty or missing staves in each section. Much of the information is duplicated from Tables 1 and 2, but Table 3 offers greater codicological detail for this one manuscript.Footnote ⁶¹ The compilers of Unit 2 (Gatherings XIV–XV) left no space for staves, and while the compilers of Unit 3 (Gatherings XVI–XXIV) left space for melodies in nearly every song, very few staves were actually drawn. Within Unit I, notation and empty staves alternate. Gatherings V and XIII contain empty staves throughout, while Gatherings II–III, VIII–IX, and XII are notated throughout. Still more confusingly, Gathering IV has a mere three notated songs, and Gathering X begins with the end of a fully notated song and then continues with nothing but empty staves until fol. 81^v, midway through Gathering XI. These three gatherings have given some scholars the impression of a notator with limited access to the repertoire, flipping through the text scribes’ work and notating if and when the music was available.Footnote ⁶²

Table 3. Fascicles of Trouvère U / Troubadour X (F-Pnm fr. 20050)

^a According to Tyssens, ‘Introduction’, pp. viii–xviii.

Gatherings VI (fols. 36–43) and XII (fols. 84–91) contain by far the most interesting instances of empty staves in the source and include clear new evidence for the U notator’s reliance on lost sources. The reason for their existence has not yet been adequately accounted for.Footnote ⁶³ As Tyssens has noted, this section contains numerous songs with only empty staves, and some partially notated songs: ‘at the top of fols. 37^v, 38^v, 89^v, the last two staves of a song remained blank’.Footnote ⁶⁴ To this short list may be added the final staff of the song Fine amors et bone esperance, RS 221 / L 65-35, which is fully notated on fol. 42^v (Figure 4) but whose final staff runs onto fol. 43^r where it remains blank (Figure 5). Thus, the word ‘pesance’, which straddles the opening, is notated only for its first syllable. Tyssens refrains from offering an explanation for these omissions, yet their oddity is precisely what yields evidence. It is hard indeed to explain this musical lacuna if the manuscript were bound at the time of copying, so much so that it serves as compelling evidence that the music scribe copied music into an unbound gathering. When the gathering is imagined as a stack of unbound bifolia (Figure 6), it becomes clear that the music copyist proceeded not by song but by bifolio side. Let us consider that the second bifolio of the stack comprises fols. 37 and 42. The flesh side includes fols. 37^r and 42^v and the hair side fols. 37^v and 42^r. The music copyist appears to have copied all the music appearing on the flesh side and left it to dry, perhaps intending to finish the hair side later. The copyist then turned his or her attention to the next bifolio in the stack: fols. 38/41, beginning with the hair side, fols. 38^r and 41^v. Both of these pages are notated, however the flesh side (fols. 38^v and 41^r) contains only blank staves. Furthermore, we find that the entire outer bifolio (fols. 36/43) has staves drawn but no music notated. We find too that while the innermost opening of the gathering, fols. 39^v–40^r, does contain notation, the outer side of that bifolio (fols. 39^r and 40^v) does not.

Figure 4 Trouvère U, fol. 42^v: Beginning of Fine amors et bone esperance

Figure 5 Trouvère U, fol. 43^r: Unfinished notation of Fine amors et bone esperance

Figure 6 Gathering structure of trouvère U, Gathering VI (fols. 36–43)

From the material point of view of a copyist, this makes perfect sense: why smudge wet ink when it would be perfectly simple to work on a new bifolio while the previous one dries? The scribe could work through a stack of bifolia in one direction before then turning each sheet upside down to work through again. This hypothesis, if true, holds implications for the transmission behind the source. The image of a copyist working through a stack of pages, regardless of whether the songs continue onto the next page, contradicts the idea that a jongleur took the unbound manuscript and ‘solicited contributions of various scribes as he carried the loose gatherings with him through the Lorraine’.Footnote ⁶⁵ It is also hard to reconcile with the idea that the notator copied what they could only when an exemplar came to hand or only when they knew the melody. It shows instead that the notator was copying onto a few unbound leaves at a time. We might further speculate that, if they were notating from memory, the copyist had enough confidence in their knowledge of the music to be able to finish copying a given song after interrupting work on it. Alternatively, that confidence might have been born out of working with an easily navigable notated exemplar, which should have made it possible quickly to relocate a song after setting it aside. Unlike the single-author collections often imagined as the earliest repositories of multiple trouvère songs, Gathering VI includes anonymous songs as well as songs attributed to several different trouvères.Footnote ⁶⁶ We would have to hypothesise a sizable collection of notated material for the bifolio-by-bifolio copying method to make sense. If the notator was working with single-sheet ‘first transcriptions from oral performance’, they must have assembled and organised a large number of them before copying.Footnote ⁶⁷

There are then two reasonable explanations for why the notator failed to complete their work in manuscript U: one is loss of musical knowledge and the other is time. Lug has proposed that the deadline for the completion of the manuscript became too tight for the ambition of the project, forcing it to a hasty conclusion.Footnote ⁶⁸ That would explain the situation whether the notator knew the songs or copied them from an exemplar. Either view is completely consistent with the evidence of Gathering VI and with what we know about the small written sources: the notator still had access to the melodies they had begun to copy, but never had the time to finish the other side of the folios where their texts continued. The idea that the notator knew the melodies but simply ran out of time is harder to reconcile with the Type 3 empty staff on fol. 89. As before, the beginning of the song on the recto side, Molt ai estat q’en bon esper non vi (PC 370.14), is notated. It is only in the portion that continues onto the incomplete verso that the notation fails. Yet the second song on that verso, S’om poguez partir son voler (PC 167.56), is fully notated, meaning that the notator must have had time to turn the page and carry on their task. Why would a notator who knew the song Molt ai estat have neglected to finish it when they clearly had time to work on that folio? The clear explanation is the loss of access to an exemplar. The notator chose, or more likely had, to prioritise S’om poguez partir by the time they finally began work on the verso because that was the song available to them at that moment. In the time between finishing 89^r and beginning 89^v, the exemplar was misplaced, had to be returned or, if it was a wax tablet, was erased.

Trouvère chansonnier U does not give us any knowledge of what sources, if any, existed during the lifetimes of the early trouvères. Studying its exemplars probably does offer the best picture we have of what song manuscripts looked like at the start of the period of collecting songs into large chansonniers. The case studies of empty staves supplement Lug’s analysis based on song organisation, by treating issues of notated transmission separately from the broader question of written transmission. The evidence leaves doubt about the existence and number of notated exemplars, suggesting more about scribal overconfidence than about any definite set of exemplars. The process of copying this source was clearly drastically different from what went on behind the projects of later manuscripts. That might be due to the quirk of the notator, differences in place or the fact that this source was compiled at a time when collecting notated melodies together was uncharted territory.

a. The Turn to Orality and the Defense of Marginal Melodies

The turn toward oral transmission of melodies in the work of Van der Werf in the 1960s and ’70s had precedents. Gröber himself acknowledged that it was possible ‘Liederblätter’ were copied by performers who had learned the songs of a troubadour, not by troubadours themselves; he also considered oral transmission of melodies more than likely.Footnote ⁷¹ Theodore Karp, though committed to a ‘trouvère manuscript tradition’, still acknowledged that stemmata constructed on the basis of textual comparisons failed to account for relationships between manuscript versions of melodies.Footnote ⁷² The perceived role of orality in the process of song transmission expanded in the middle of the twentieth century as a means of explaining variance. The initial Gröberian admission reappeared as an independent theory that replaced ‘Liederblätter’ with performers entirely. The new model, the ‘Repertoire-Theorie’ proposed by Friedrich Gennrich in the title of his seminal article, supposed that transmission from performer to performer nearly always served as an intermediary stage between the invention of song and its inscription in parchment.Footnote ⁷³ ‘Repertoire-Theorie’ does not preclude written transmission but merely insists that the first stage of that transmission occurred when performers transcribed every song they knew from memory.

Gennrich could not believe the degree of musical variance found in the chansonniers could arise from written copying.Footnote ⁷⁴ The process of collecting notated melodies, he argued, could not be as neat as the progression from song leaves to author collections to song books. He acknowledged that written collections, organised into author sections, might have appeared at an early stage and influenced the way song texts were arranged in chansonniers. But that did not mean the melodies were copied from them in the same sequence. Rather, for Gennrich and his followers, the operative unit of melodic transmission was the repertoire of an individual performer, hence the name ‘Repertoire-Theorie’. That performer would then either write or dictate their melodies, perhaps separately from the inscription of the texts. The repertoire of a single performer might not perfectly match the planned contents of a manuscript, thus explaining many blank staves, drastic changes in melodic family within a single source and in many manuscripts the violation of author groupings. Intentional variation was introduced by these performers and then either reproduced by transcriptions of their performances or copied by performers themselves as a study aid. Different performers would have different versions of a melody in their respective repertoires, and those repertoires and the melodies would change over the life of a performer.Footnote ⁷⁵

For Gennrich, ‘Repertoire-Theorie’ separated the melodies surviving in the chansonniers from their authors and thus called into question the task of critical editing and the search for authorial originals.Footnote ⁷⁶ Yet the subsequent turn toward orality in the work of Van der Werf served to reclaim melodic variance as evidence of a rich performance culture. In his extensive writings on both trouvère and troubadour traditions, Van der Werf argued that melodic change in an oral culture is likely to stem from both scribes and performers treating songs as variable objects.Footnote ⁷⁷ In Van der Werf’s works, both editions and studies, scribal alterations are often taken to be indistinguishable from changes made by performers.Footnote ⁷⁸ This realisation has been the single most enduring challenge to the idea of a written record that could enable the philological study of original trouvère melodic versions.Footnote ⁷⁹ A heavy suspicion of traditional textual criticism when applied to trouvère melody thus pervades his work and that of most trouvère scholars after him.Footnote ⁸⁰ However, for Van der Werf as for many others, melodic correction was still possible. If anything, melodic variance produced in performance could be relied on to reflect medieval sound, whereas changes made by scribes might reflect haste, musical incompetence and mechanical error. Manuscripts R and V in particular have borne the brunt of this new suspicion of scribes, until more recent work has defended their melodies. While Van der Werf’s assessment of these manuscripts can be determined only from interventions in his melodic editions, there have been numerous explicitly negative analyses of these chansonniers ever since Gennrich, as the following paragraphs describe.

The sloppy appearance of trouvère R prompted early speculation that it was copied swiftly by non-professionals connected to one of the puys, organisations whose membership included performers and whose events included song competitions.Footnote ⁸¹ The image is that of musicians with minimal scribal training hastily jotting down melodies as they were sung, perhaps even as they were invented. R was thus an obvious candidate to apply ‘Repertoire-Theorie’ to, a task which Johannes Schubert undertook as his doctoral dissertation.Footnote ⁸² For Schubert, the chansonnier (at least where music was concerned) was a collection of attempted holograph copies of first-hand transcriptions, thus not so different from Lug’s exemplars of U.Footnote ⁸³ R is unusual in that melodies in different sections of the manuscript resemble the notated songs from totally different manuscript families, at times the KNPX group, at others sources like A, M, T and a.Footnote ⁸⁴ In the ‘Repertoire-Theorie’, these different melodic families reflect the repertoires of the different performers who participated in the creation of the manuscript.

Schubert also offered another hypothesis, that some songs in R were ‘forgeries’ (Fälschungen), ‘faked’ (fingierte) by the notator.Footnote ⁸⁵ These songs were invented by the scribe when no exemplar was available, ‘to remove the appearance of incompleteness from the manuscript’ (‘der Handschrift den Anschein des Unfertigen zu nehmen’) by filling blank staves.Footnote ⁸⁶ Hans Spanke, too, had identified melodies that he considered to be the result of scribes with too much time and creative energy on their hands.Footnote ⁸⁷ In other words, these melodies had no existence prior to their notation. For the melodies shared in common with trouvère chansonnier V, Schubert deemed that these sources must have been deplorably copied, suggesting that the interventions and inventions of these notators were musical failures.Footnote ⁸⁸ The fact of their composition in writing delegitimated them in Schubert’s eyes, as much as written composition by a named trouvère would have sanctified them. A hierarchy of variance based on assumptions around performance grew up in editions of trouvère music. These supposed scribal inventions occupied the bottom of this hierarchy of variants.

Some twenty years after Schubert’s dissertation, Hans-Herbert Räkel took up the task of teasing apart variants produced by successive generations of performers from those due to scribal error.Footnote ⁸⁹ Manuscript R played a much smaller role in this analysis than in Schubert’s exercise in ‘Repertoire-Theorie’, but Räkel’s assessment reflected Gennrich’s and Schubert’s influence. Räkel dismisses R as riddled both with errors and with later reinventions by uneducated scribes.Footnote ⁹⁰ The result is, at best, a ‘representative workpiece by an admirer of trouvère lyric’ who took as their task not to capture melodic performances but to possess the texts and simulate the appearance of melody.Footnote ⁹¹ Rather than being copied for the sake of preservation, its melodies were fabricated by scribes to enhance the value of the texts.

While there are cases elsewhere of ‘music’ notation consisting of no more than square note-heads added to staves at random heights, the inauthenticity of R’s music is less obvious.Footnote ⁹² There are certainly instances where R’s notation looks much sloppier than that of most chansonniers. The alignment of music with text is frequently inscrutable (a charge that could equally be levelled at manuscript V).Footnote ⁹³ And R (like M and V) contains many melodies that are completely distinct from those that consistently accompany the concordant texts in other manuscripts. Yet there is nothing in any of R’s music (beyond some vagueness in text–music co-ordination and sloppy penmanship) to suggest that the notes are random or purely invented by a notator for appearance’ sake. There may be ‘flights of scribal fancy’Footnote ⁹⁴ in V as in R and M, in the sense that only one scribe ever presented a certain melody in a certain way. But this is no sure sign that the music as written had no connection to the music as sung.

The attractiveness of the idea that some melodies in these sources were scribal inventions lies in its power to explain why many melodies vary so dramatically from other sources. It is a peculiar feature, often remarked on, that the melodies of trouvère songs display difference beyond variance, with entirely different melodies in different manuscripts for the same song.Footnote ⁹⁵ These diverging melodies have been called ‘isolated’ or ‘marginal melodies’ or also the results of ‘Kontraposition’, a twist on contrafaction.Footnote ⁹⁶ The term ‘Kontraposition’ assumes that the same text was set to music multiple times independently. For Schubert and Räkel, this took place in writing. Music scribes, eager to complete their task despite a lack of exemplars for some melodies, might be expected to take matters into their own hands and adapt a stock tune to the text being copied. The quality of the resulting song would vary wildly along with the competence and conscientiousness of the notator. We do indeed see variance in style and coherence in the melodies of R and V, and there is evidence that one of V’s music scribes took considerable creative license to get out of trouble when aligning music above the text.Footnote ⁹⁷ It is thus worth taking seriously the possibility that some or all of these ‘Kontraposita’ were invented and asking what evidence to that effect can supplement melodic comparison.

It is in this context that Christopher Callahan has championed marginal or isolated mélodies as legitimate alternatives to their cousins.Footnote ⁹⁸ Callahan considers the three hypotheses in circulation to explain isolated melodies: invention by scribes, oral transmission of a ‘secondary lyric practice’ (in other words, invention by performers in the context of ‘Repertoire-Theorie’) or transcription from lost written sources.Footnote ⁹⁹ Callahan noted that isolated melodies were ubiquitous in R and V. In the latter source, he noticed that they appear clustered into groups. His focus was on the collection of songs by Thibaut de Champagne, and it is within the Thibaut section of V that a surprising break appears between concordant melodies and isolated melodies. It is a particularly interesting area of the manuscript in which to find extreme variance, since in the view of many scholars since Gröber single-author sections like that of Thibaut were most likely to be copied from extremely early written collections.Footnote ¹⁰⁰ Every melody in V before fol. 17 is very similar or even identical to the melodies for the same text in other sources, especially K, N and X. From fol. 17 to fol. 24, by contrast, there is a long string of melodies that are all ‘marginal’ or ‘isolated’, to translate Callahan’s terms (‘marginales’, ‘isolées’).Footnote ¹⁰¹ On the basis of the concentration of melodies in V, Callahan leans toward his third hypothesis: that the isolated Thibaut melodies in V were copied from lost notated sources. This seems to indicate the circulation of contradictory melodic versions of Thibaut’s song collections.

Callahan suspected that the beginning of the string of isolated melodies coincided with a new gathering but was unable to view the manuscript in person to verify the hypothesis. My own collation, based on examination of the manuscript, confirms this intuition. Fols. 1–16 make up the first two gatherings of manuscript V, and fols. 17–24 make up the third (see Table 4).Footnote ¹⁰² The music in the first two gatherings was copied entirely from a notated source with melodies in agreement with the other surviving sources. The music of the third gathering could reasonably have been copied from a lost notated source. Switches from concordant to isolated melodies stop lining up with gathering breaks at this point. The last two songs of Gathering III correspond closely to their concordant versions, and then the final song by Thibaut de Champagne, in the middle of Gathering IV, again has a unique melody. These alternations between concordances and isolation persist throughout the rest of the manuscript. For these later gatherings we might now ask whether the music scribe copied from a notated source with now-isolated melodies, if they relied strictly on their own musical knowledge, or if they resorted to invention. To answer that question, I turn once again to empty staves, this time found within a section of isolated melodies.

Table 4. Gatherings, Concordances and Missing Music in Trouvère V (F-Pnm fr. 24406)

b. Empty Staves and Lost Notated Sources

Leaving aside the appended fols. 119–155,Footnote ¹⁰³ V contains only one song fully lacking in notation, the fittingly texted Ne m’i sont pas achoison de chanter (‘I have no reason to sing’), RS 787 / L 65-52 (Figure 7). The song begins a new folio (fol. 33), indeed, a new gathering (Gathering V), and the staves are blank from the beginning of the song. These blank staves do not continue for long. The next song, Ne puis faillir a bonne chanson fere, RS 160 / L 65-51, has notation provided for most, but not all, of the first stanza (Figure 8). Notation begins not at the start of the song, but in the middle of the staff and even in the middle of the sentence, on the second word of poetic line 5, ‘meillor’. The rest of the song is fully notated. To demonstrate the abruptness of the beginning, the text of the first stanza is reproduced with translation below, with the start of the notation placed in bold type.

Figure 7 Trouvère V, fol. 33^r: Empty staves for Ne m’i sont pas achoison de chanter

Figure 8 Trouvère V, fol. 33^v: Incomplete notation of Ne puis faillir a bonne chanson fere

There is much to puzzle over here. It is easy to see why notation might be absent from Ne m’i sont pas: generally, staves were left empty due to lack of knowledge, lack of an exemplar or lack of time. It is less easy to imagine why the notator notated only the second half of Ne puis faillir. Had the notation begun again with a new folio, as we saw in the case study from chansonnier U, we might have suspected the notational lacuna came as a side effect of copying folio by folio. If it began with a new song, we might have attributed the lacuna to a lapse of memory. But only the second part of Ne puis faillir is notated; this is the cauda, which, unlike the opening pedes of most songs, is unrepeated and thus likely more difficult to learn and remember. Had the notation begun with a new poetic line, we might have argued instead that the easily-remembered opening was not worth notating but the ending was.Footnote ¹⁰⁴ If the song had come directly from a scribe’s knowledge as a performer, or were their own creation, we would have had notation from the beginning of the song or the beginning of a line. There is a chance, however slim, that the music was left out for dramatic effect to match the texts of Ne m’i sont pas and Ne puis faillir; but in that case, why is the eye-catching stunt reserved only for these two songs, when the themes of trouvères failing to sing and improving their songs are so ubiquitous throughout the repertoire, including elsewhere in this manuscript? The puzzle only resolves if we postulate a lost notated exemplar for V. While the abrupt start of notation does not correspond with a page break or a new gathering in V, it might have corresponded to such a break in the exemplar. If that exemplar were damaged, or if the notator had access only to part of it, they might well have seen only the second half of the song, starting from the second word of the fifth line of poetry.

Ne puis faillir can be found undamaged and with complete notation in the closely interrelated sources K, N, P, X and L. But the music that does exist in V has very little discernable relationship to the surviving music in the other sources. The music transcribed in Example 1 facilitates the comparison. The melody from V (also seen in Figure 8) is compared to L and to a regularised version of KNPX, and I offer staves for variant readings when they occur among these four sources. In poetic line 5, V’s music for the first five notated syllables, ‘meilleur s’en seroit’, corresponds roughly to the pitches in the other sources and matches their contours. V’s first three notes at the beginning of poetic line 7 seem to mimic those in the other sources but in transposition by a fifth. Elsewhere in the song, V reverses directions, replaces single notes with ligatures and viceversa and generally reorganises the modal shape of the piece. The KNPX and L versions centre around g with a secondary emphasis on outlining the octave d–d′; what survives of V’s version is firmly in an F mode. None of the three cadence points is the same for V as for the concordant sources. While the opening of line 7 suggests some relationship between the versions, the rest of the piece resembles what Schubert calls a ‘Falschung’ and Räkel calls ‘Kontraposition’. Moreover, the piece appears in a section of the manuscript primarily populated with isolated melodies.Footnote ¹⁰⁵ This is the primary reason for focusing on this example, as it raises the question of how such melodies come to be in surviving manuscripts.

Example 1 Comparative transcription of Ne puis faillir a bonne chanson fere

Examples of empty staves beginning in the middle of a song and in the middle of a folio are rare among trouvère chansonniers. I find such instances only in later sources, specifically V, a and R, and Ne puis faillir is the sole example that corresponds to a unique melody. In trouvère R, we see a close approximation of the situation in V, with the song Onques ne fui sans amour, RS 1964 / L 192-16, on fol. 85^v (Figure 9), though in this case the melody agrees with other sources. As usual, the text follows an abab structure, though the cauda contains a c-rhyme and a refrain.

Figure 9 Trouvère R, fol. 85^v: Incomplete notation of Onques ne fui sans amour

In poetic lines 5 and 6, between the words ‘son’ and ‘Par’, the notator has skipped almost (but not quite) an entire staff of music. The notation trails off after the second syllable of the fourth line of the lyric (in bold in the text transcription above) and only picks up again with the refrain, at the very end of the staff (also in bold). As in the previous example, neither the mise en page of the song nor the organisation of its text or melody explain why this particular segment of the notation is missing and a lost exemplar seems the most obvious explanation. The missing phrase would be just about the right length to correspond to a single staff if the piece were laid out in bicolumnar format, as it appears in all sources but R and Z.

The work of Spanke, Schubert and Räkel is impressive in its embrace of scribal agency, even if Callahan’s defense of isolated melodies against them remains convincing. It is reasonable to imagine that at some point, the role of scribe and inventor overlapped, as did those of scribe and performer. Determining when precisely this took place was of paramount importance for philologists in the early and middle twentieth century, as it could aid in sorting melodies that came through a respectable oral tradition from simulations of trouvère melody. Yet the specific idea that empty staves were filled with newly invented melodies cannot be reconciled to the apparent attitudes of chansonnier music scribes. The existence of examples like those above sets limits on the musical knowledge and creative liberties taken by scribes, even in the very sources considered by Schubert, Räkel and Spanke. Examples like these narrow the range of possible modalities of copying, as they demonstrate a certain level of care for accuracy over comprehensiveness on the part of the V and R scribes. These notators could not have been guilty of inventing melodies wholesale, otherwise why not fill in the missing lines with another ‘personal fantasy’?Footnote ¹⁰⁶

More generally, it is hard to reconcile the attitude behind Ne puis faillir’s blank staves for these melodies with confident scribal knowledge, either from memory or from invention. How could a scribe know a song well enough to copy it but not remember the repeating opening? And how could a scribe have the gall to invent only the second half of a song? It is perhaps easier to imagine the gap in R resulting from a faulty memory than it is for the V and a examples: the first four poetic lines (with repeating musical pedes) and the refrain (with repeating text in each stanza) would probably have been the most memorable parts of the melody. Other examples of missing music in R and a do correspond to page breaks and argue against memory issues as a major contributor to absent music: problems of codicological organisation seem a more likely culprit.Footnote ¹⁰⁷ Even in the case of Onques ne fui sans amour in R, it seems odd that anyone could remember exactly the first two notes of the cauda and no further. A notator who relied on a fragmented notated exemplar for the first five poetic lines and also knew the refrain by memory might have produced exactly what we see on R’s page.

The previous examples push ever closer to the conclusion that notated exemplars played a significant, perhaps even dominant role in the compilation of chansonniers such as R, U and V. Text scribes would have had little use for such exemplars and probably relied instead on more compact and less skill-intensive, text-only sources, as we saw evidenced by the text scribe’s error in Tant autresi con descent la rousee (see Figure 2 above). Music scribes for some large chansonniers on the other hand must have made use of notated exemplars, as seen in the cases of Ne puis faillir and probably Onques ne fui sans amour. The rest of this article seeks to show how empty staves can reveal what these lost notated exemplars may have looked like.

c. Preparatory Materials or Notated Tradition? Another Dinosaur Philologist post-Extinction

If not the ‘Liederblätter’ of the late nineteenth century, what kind of notated sources should we imagine being used as exemplars for sources such as R, U and V? The idea of ‘very small ephemeral materials’ being used as exemplars for copying songs may bring to mind the wide-spread use of wax tablets during the Middle Ages.Footnote ¹⁰⁸ In fact, Haines has considered the possibility that such tablets were used both for trouvère texts and their melodies in his codicological work and study of medieval book culture.Footnote ¹⁰⁹ This final section attempts to integrate wax tablets into the growing narrative of lost written trouvère sources and asks how they relate to the case studies we saw above.

Haines’s chapter on the manuscript context of Adam de la Halle is a rehabilitation of Gröber of sorts, in that Haines is interested in the earliest written vernacular songs and the earliest notation of their melodies. His defence of the ‘dinosaur philologists’ brings into focus evidence from library inventories, erasures in manuscripts and the same accounts of the lives of trouvères and author organisation used by Gröber and Schwan.Footnote ¹¹⁰ Haines also emphasises two types of media known to have played a large role in medieval written culture in general: the parchment roll and the wax tablet. Haines infers that both forms of written communication must have played a role in the transmission of trouvère music, without negating the fact that the creation, performance and memorisation of this music took place in a primarily oral culture. Haines, like most scholars today, still accepts that oral transmission and written transmission occurred simultaneously over the whole course of the trouvère period.Footnote ¹¹¹ Within this background, the ‘immediate ancestors’ of the extant chansonniers can be studied with greater certainty than first performances or transcriptions.Footnote ¹¹²

Thus Haines does not offer anything like a chronological progression from authorial originals to ‘Liederblätter’ to ‘Liederbücher’ and ‘Gelegenheitssammlungen’.Footnote ¹¹³ Haines’s work, along with Leach’s and Lug’s, serves as a moment of reorientation toward manuscript studies. He differs from Leach and from the current article in that he does not suggest any way that evidence from surviving chansonniers could be used to establish which type of medium was used in their copying. His focus instead is showing that trouvère songs transcribed in single-author collections (resembling Gröber’s ‘Liederbücher’), parchment rolls (resembling ‘Liederblätter’) and wax tablets, all must have existed. Haines’s review of the evidence is refreshing and it raises important questions. How did these lost writing materials relate to surviving chansonniers? One part of this issue is the question most deeply tied up with debates over oral and written transmission, that of whether song leaves or performance collections were used for copying. This has been partially answered above. Another question, brought up by Haines’s demonstration of how trouvère songs would look inscribed in wax, is whether chansonnier scribes themselves ever wrote songs out on wax tablets or scraps of parchment prior to inscribing them in chansonniers. In other words, was their act of copying into the chansonniers one of sight-reading, or a polished performance? The examples of empty staves considered here do not challenge any particular theory in Haines’s work, but rather shift the focus towards how these media (and which of them) could plausibly have interacted with the surviving chansonniers.

A wax tablet, due to the affordances of the soft material, can be erased and rewritten. For this reason alone, mistakes made on wax were less costly. Wax was exceptionally suited for the preparatory function described by Gröber for parchment leaves and van der Werf for leaves and wax, as an aid to scribes and composers.Footnote ¹¹⁴ Parchment was less suited to this purpose, since the process of inscribing ink on it was all but permanent. Binding parchment into roll or codex format added another layer of determination to the source. Yet any written-upon surface has the potential to serve multiple functions. There is the preparatory function just mentioned, but also the repository function (best exemplified by the chansonniers), and the mnemonic function exemplified in the performance copies, touched on above. Most importantly for Gröber and for Haines is the function of transmission, written surfaces carrying their contents from one place, or more precisely one mind, to another. It should be clear that parchment could fulfil any of these roles. That a wax tablet could be extensive enough to qualify as a medium for a small song collection, and thus serve as a means of transmission, is clear from the examples provided by Marilynn Desmond in her consideration of the role of tablets in the composition of long poems.Footnote ¹¹⁵ Needless to say, a written source could be produced with one purpose in mind and then later serve another, incidental function. This is what Gröber imagines when he describes preparatory ‘Liederblätter’ being used to transmit songs to construct larger collections, or what Gennrich envisages when he hypothesises memorial performance copies later becoming exemplars.Footnote ¹¹⁶ This is also precisely what Lug and Leach have demonstrated in their studies of shared exemplars for I and C, which Lug argues started life as preparatory materials for U.Footnote ¹¹⁷ The current case study aims to differentiate between different functions, if not media, of lost written exemplars. While it is impossible to identify the imagined purpose of a source at the moment of its creation without any direct knowledge of that source, it is possible to know how it was used when it became an exemplar for sources that do survive.

Given the way songs are notated in the trouvère manuscripts R and V, it seems unlikely any preparatory efforts were made by their music scribes. Notators behaved as though they were seeing exemplars for the first time, indeed, as if they were not the ones to have prepared whatever source they were copying from.Footnote ¹¹⁸ Furthermore, if wax tablets were employed as a means of determining layout, surely both notation and text should have been taken into account. Rather, examples like Tout autresi con descent la rousee (section IIa above) demonstrate that the text hand did not always know what the music hand would be doing.Footnote ¹¹⁹ A few further examples argue specifically against effective co-ordination between the planning of layout and music notation, even in one of the most elaborately decorated chansonniers, trouvère T. Staves in this source are frequently drawn for refrains, but only some of these refrains were supplied with music. Some entire songs were never notated, despite being situated between other fully-notated songs, such as Dame des ciels, RS 1353 / L 102-5, on fol. 32^v (Figure 10). Some stanzas of songs were mistakenly given staves, such as the final stanza of Trop est costumiere amors, RS 2018 / L 1-7, on fol. 81^v (Figure 11). Similarly, in the case of non-strophic lais, the scribes determined that it was unnecessary to copy out all of the repetitive and formulaic sections in full. However, there was disagreement about how much or how little notation was necessary, as can be seen in the second half of the Lai d’Aeliz, En sospirant de trop parfont, RS 1921 / L 265-659, on fol. 68^v (Figure 12), listed under Type D of Table 1 and Type 4 of Table 2. While the text scribe was parsimonious in providing staves, the notator still deemed three of them (those at the top and bottom of the page) superfluous in the final process of notation.

Figure 10 Trouvère T, fol. 32^v: Empty staves for Dame des ciels

Figure 11 Trouvère T, fol. 81^v: Superfluous staves for the final stanza of Trop est costumiere amors

Figure 12 Trouvère T, fol. 68^v: Superfluous staves in the Lai d’Aeliz

Such examples also undermine the idea of any consistent preparatory efforts to work out the notation itself, independently of the text, in wax. In general, notators may well have felt the need to decide on notational questions such as which ligatures to employ and where to place the clef before risking marring a clean copy of the text with notational mistakes. Errors in text–music alignment and transpositions by a third could then have crept in as the notation from the tablets was transferred onto parchment.Footnote ¹²⁰ While this is a reasonable hypothesis, it is incapable of explaining the gap we saw in Figure 8, at the beginning of Ne puis faillir a bone chanson fere on fol. 33^v of chansonnier V. If the notators themselves were in the habit of notating entire melodies in wax before copying them to parchment, what happened to the beginning of the piece? Damage to the wax tablet should have been easily remedied by the notator relying on their own knowledge of the song. The lack of notation rules out that knowledge.

In sum, text–music alignment, layout and notational choices show a distinct lack of planning in the very sources for which lost exemplars are most likely. Wax tablets were almost certainly employed at some stage of the construction process of some chansonniers, but their explanatory power over the errors and imperfections we see in chansonnier V is limited. Wax tablets play into this story only insofar as they are separated from the preparatory function so often associated with them. Notators may have copied from another source onto the wax tablet and then from the wax tablet into the chansonnier, in which case we are dealing with the existence of two lost sources: the wax tablet, and a now lost parchment witness to trouvère song. The only way wax could explain what we see in chansonnier V is if it was a lost source in the same way we normally imagine parchment to have been: the wax itself would need to have changed hands, potentially even across generations, allowing mistakes to arise from errors of interpretation. In that case, the function of the wax changes. It ceases to be preparatory material (prepared by the scribe for the scribe) to a medium of transmission, a lost ephemeral source carrying melody across space and time via the medium of notation. The act of copying from wax tablets is not in itself enough to explain the errors and omissions we see. We must posit that a lost notated surface (parchment or wax) changed hands. A scribe with knowledge of the specific song copied onto it, and a scribe with less knowledge of that song copied from it.