PART TWO—
THE VIEW FROM WITHOUT:
PERFORMANCE AND POETICS REFLECTED IN THE CHANSONNIERS

Four—
Mouvance in the Manuscripts

"High-fidelity reproduction" of troubadour lyric, then, may not necessarily have meant verbatim replay of an original. The customary fidelity, as both poets and transmitters inform us, relies most upon bona fe and drechura, "good faith" and "rightness": a singer or writer is true to the song when realizing it in its best form. The singer acts as the poet's representative in more than one way, assuming some of the duties of the maker along with the maker's persona. The variety of transmitters mentioned by the poets—not only trained performers hired to reproduce the master's voice, but also friends, patrons and their courts, village boys from the jongleurs' home towns^[1] —makes mouvance appear the natural result of many years' melhuramen (improvement), some for the better and some for the worse.

Under these circumstances, the greatest wonder is that many of the troubadours' songs remain essentially unchanged, often in more than one manuscript tradition. Mouvance operates selectively: it leaves certain poems almost untouched no matter how far they travel, while it alters others so drastically that even the most Lachmannian of editors must sometimes print more than one version. The purpose of Part Two is to uncover some of the principles that govern this selectivity.

Of particular interest in the search for those principles are the disturbances in stanzaic sequence that characterize the troubadour chansonniers . Here is mouvance on a large scale: a single poem comes down to us with its stanzas in nine different arrangements, another with nineteen; the transposition (or, to borrow a term from mathematics, "permutation")^[2] of stanzas offers large-scale variants of a type that has generally been ignored in stemmatic reconstructions of "original" texts, yet ob- 

viously affects the meaning and structure of each text in a profound way. Since most of the conscientious editions of troubadour poems indicate the various stanzaic sequences preserved in the manuscripts, the degree of stanzaic transposition undergone by poets' works can be quantified, studied, and compared more simply than can other kinds of variants. The study of this type of large-scale variant has the advantage of independence from the stemmatic system (which is based on comparison of word and phrase variants) of establishing "authentic texts": manuscripts with shared word and phrase variants do not necessarily share a stanzaic sequence."^[3]

That stanzaic transposition is distributed unevenly among the Occitan poets has been documented, but the reasons for this unevenness have not. William D. Paden delineates three areas of investigation that might explain the variance among poets in the stability of their texts; he suggests that "permutation" might correlate with the period of time in which a poet produced his work, with his degree of stylistic difficulty, or with his ability to write and thus to produce a written archetype for future manuscripts. Toward this last question Paden contributes much, including a substantial study of Guilhem IX's literacy. He concludes tentatively that the count could not write. These findings have meaning for the study of other troubadours in at least two ways: first, an illiterate Guilhem would not have established the author's manuscript as the first step in transmitting a kind of poetry for which he set so many other precedents; second, unless Guilhem wrote, the remarkably low rate of stanzaic transposition In his songs (22 percent; see Appendix A) cannot be explained as the effect of an autograph copy on the manuscript tradition.^[4]

Paden also argues that sung transmission grew less popular as time went on. His study of the occurrence of the names of jongleurs in poetry, as a function of the date, shows a clear historical trend in the poets' use of a device that includes the transmitter in the message itself: naming the jongleur within the poem (1979, 4–7; 1984). Mention of jongleurs declines gradually—not in the number of occurrences but in the percentage of known troubadours who name jongleurs—especially between A.D. 20 and 1340. Since many of the jongleurs named have been documented as living near the poets or in their households (e.g., Raimbaut d'Aurenga's Levet in 1173), it is safe to assume that the named jongleur is not merely a writer's affectation. The decline in the practice of naming the longleur within the poem proves first of all that mention of 

jongleurs had gone out of fashion, and only secondarily that writing had overtaken singing as the primary mode of transmission. But clearly the weight of evidence points to a rise in writing, and a decline in sung transmission, from about 1250 onward. If we conclude, with Paden, that the increasing scarcity of troubadours who name jongleurs in the thirteenth and fourteenth centuries^[5] reflects "a gradual shift from oral to written transmission" and, concurrently, "elevation of the vernacular from the status of low language into competition for that of high language in the culture of Occitania," then we also observe that, along with the decline of the joglar, "the lyric pulse weakened" (Paden 1984, 13–14). The present study concentrates on the troubadours of the twelfth and early thirteenth centuries, when the "lyric pulse" still beat unabated.

One caution to drawing conclusions from the incidence of jongleurs' names: there are many loopholes in the criteria for determining what is the name of a jongleur and what is not. I have shown in preceding chapters that the recipient of a poem is often asked to memorize it, so that he or she—without losing class status—partakes of the jongleur's role in repeating the song. Both types of names—the senhals and the jongleurs' professional names—include fanciful, descriptive nicknames as well as ordinary given names. A tornada may contain several names and senhals . Might the "Joglar" of Azalais de Portcairagues, or one of Guilhem de St-Didier's three "Bertrands," or Rigaut de Berbezilh's "Bels Bericles," ever have performed as a jongleur? These poets are merely the first three to come to mind; there may be many more such cases. None of these three poets, nor others addressing potential transmitters whose names have traditionally been taken as senhals for a nonperforming addressee, were counted as having "named joglars." Yet their inclusion might change the percentages—and thus the appearance of a trend—rather dramatically, since the percentages are based on numerically small samples.^[6]

The third issue, that of stylistic difficulty as a control on the mutability of lyric texts, will receive careful attention here. Friedrich Diez observed the function of elaborate rhyme schemes as a mnemonic aid not only in fixing the internal structure of each stanza but also, for some poems, in fixing the order of their stanzas.^[7] Martín de Riquer, in his introduction to Los trovadores, explains the development of many of the metrical forms that draw rhymes from one stanza into the scheme of the next—coblas doblas, alternadas, capcaudadas, capfinidas, and so on—as the result of a deliberate effort on the poets' part to bring transposition under 

control.^[8] The recent recognition that some poets both acknowledged and defied the tendency of lyrics to alter in circulation has led to the further claim that poets conceived complex versification as, among other things, a safeguard for textual integrity: "Some troubadours came up with the idea that a complicated metrical structure, a complex word arrangement might be appropriate in order to protect the composition of the poem from outside influence. Indeed, a public of literary experts could be expected to detect the mutilations of the avol chantador right away and denounce them as such" (Mölk 1979, 5).

That complex verse forms were developed expressly "to protect the composition of the poem from outside influences" is very plausible, but it requires verification.^[9] " Mölk's statement assumes the drive to textual integrity, the perception of rhythmic and rhyming structures as potential tools for this purpose; it also assumes that "outside influence" would be perceived as undesirable, as worsening the songs it changed. Finally, though time increased the number of complex metrical forms that had been invented, we cannot say that time brought an increase in their use . On the contrary, the use of certain kinds of complex forms actually declined between 1100 and 50 (see Appendix A, Fig. A-9).

Taken as a hypothesis, Mölk's notion is well worth testing: we cannot understate the importance of such a discovery, should it prove true. This would indeed be, as Mölk calls it, an "obviously significant moment in literary history": the moment when poets in a performing tradition, where outside influence could be expected, began to look on their works as artifacts; the moment when they rebelled against the instability of their medium, when they idealized a verbatim textual integrity beyond the capability of available means of publication. It would be an exceptional development, in the midst of medieval tradition favoring anonymity and adaptation, if the idea of authorship suddenly became applicable to contemporary lyric when authority had heretofore been reserved for the ancients.

Did the troubadours really view metrical complexity as a possible remedy for mouvance? The texts indicate that some did. Yet the story is not that simple. Complex verse forms do stabilize the order of stanzas; but the poets do not increasingly use them. Only 19 percent of the 552 poems in my survey avail themselves of one type of stabilizing device studied: stanzaic linkage, that is, connecting one stanza to the next by establishing a pattern in the rhyme scheme that requires their contiguity. 

What is more, the use of this type of device drops to an all-time low among the very poets who perfected it (see Fig. A-9, poets 13–16). Clearly, something is wrong with this theory that the poets, relating stylistic complexity to textual integrity, "tried" to use complex verse forms as a protective measure.

The "solution" of stylistic complexity proves to have been effective in practice, but its use responded to the "problem" only intermittently, whether deliberately or accidentally. The relation of "problem and solution" finally depends on evidence that elaborate metrics were perceived as a guard against outside influence, and that at least some troubadours both knew transposition to be a problem and believed stylistic complexity capable of solving it. Some poets show very little awareness of textual change in their works; some openly welcome "outside influence" while still using certain kinds of "stylistic complexity" that promote textual stability (for example, Jaufre Rudel uses stanzaic linkage). More disconcerting, in view of the historical movement around 1180–1195 to abandon stanzaic linkage, is the need to face the distinct possibility that certain poets chose not to "solve" the "problem."

The opinions of the poets on the subject of "textual integrity" vary widely. Some apparently believe, for instance, that only inept tampering with a poem qualified the singer as an avols chantaire, whereas skillful tampering that actually improved the song could earn the singer the praise of its original author as well as of the audience. Nearly all troubadours take a stand against bad renderings of their works, but few object openly to "outside influence" in itself. Part Three will probe more deeply the vocabulary, the imagery, and the cast of characters with which the poets illustrate "textual integrity" and its alternatives; for now, then, I will only sketch out some of the questions behind my statistical survey.

The external evidence presented here and in Appendix A offers a way of looking at mouvance from a point of view not available to the troubadours themselves. It can tell us what actually happened in transmission, regardless of whether the typical processes were known or viewed as problematic by the poets, and regardless of any intention they might have conceived, whether to subvert those processes or to encourage them. This evidence will tell us which poems—when, by what poet, using which techniques of rhyme and meter—were most likely to produce versions with variation in the sequence and array of stanzas. It will also tell us which variables made no difference to that likelihood. 

Instability in Transmission: Popularity and Repertories

Since some poets say that their songs will improve or appreciate with repetition, I decided to test the hypothesis that their poems might have fulfilled this forecast in the chansonniers —that is, that the more often a song was repeated, the greater its chances of being recast by transmitters. Because of the special character of this manuscript tradition and its apparent status as the last step in a much longer process of transmission—its status, perhaps, as the result of translation from a "fluid" medium to a "monumental" one—the rate of survival in written copies probably reflects not only popularity with copyists but also frequency of performance. Based on my survey, the "average" troubadour song (extant in more than one copy) survives in 8.8 manuscripts, with its stanzas in 2.1 sequences and 3.2 different stanzaic arrays (including abridgments and fracturings).^[10] Deviations from the norm in the rate of manuscript survival can serve as a rough index to popularity in the jongleurs' repertories. Paden believes that jongleurs had very small repertories (1984, 93). If this is so, one might expect jongleurs to be very selective in composing those repertories, favoring songs that not only suited their voices and stage personalities but also consistently pleased audiences (and brought coins to the cap).

I am assuming that essentially self-supporting jongleurs, both those mentioned by the poets and later arrivals, performed the songs. Even if it is true that certain jongleurs, those mentioned in the poetry, "worked with only one poet" (Paden 1984, 94), it does not necessarily follow that every jongleur had a partner-poet who consistently provided his songs. Mailolis, who approaches Bertran de Born requesting a song, is described by Bertran as a stranger to him, and an obnoxious one at that; he gives him a song ostensibly designed to reward its singer with a rain of overripe tomatoes rather than of kudos and coins (B Born 27 [P.-C. 80, 24]). Since the jongleur apparently received his payment not from the composing poet but from the addressee or audience, we may expect that jongleurs chose to perform certain "successful" songs more often than others. It would be naïve to suppose that a popular favorite song like "Can vei la lauzeta mover" ("When I see the lark move") was performed only by a single jongleur whom Bernart de Ventadorn designated as his partner and then consigned exclusively to parchment on the retirement or death of that lucky jongleur. Certain poems inevitably "pay" better than others—as the poets were well aware, given the tenor of their discussions of "clo- 

sure" and "integrity" (treated in Part Three). The kind of "exclusive franchise" involved in the supposed one-to-one relationship of poets to jongleurs would conform to a concept of literary property, exclusivity in performance, and authenticity that was shared by some poets in the school of Marcabru but denied by Jaufre Rudel, Bernart de Ventadorn, and Giraut de Bornelh.

My interpretation of the indices of manuscript survival (M and MS) therefore supposes that popularity with copyists can be traced to popularity with jongleurs and ultimately with audiences. We recall that Bernart Amoros compiled poems that he had "seen and heard " in their native lands. I arrived at an index of the rate of retransmission (manuscript survival, M) by very mundane means: by counting the total number of copies of a given poet's works. To get MS (manuscripts per poem), I divide M by the number of different poems copied. These popularity indices, rating each poet as to his work's survival in manuscript copies, were then measured against several different indices of susceptibility to stanzaic transposition.

Rate of transposition versus year, when plotted on a graph, shows "clouds" of thoroughly random "scatter" reflecting the wide variation in stability among poets composing at the same time; the correlation of date and percent-with-transposition is virtually nonexistent up through 1250 (Fig. A-6). Thus, chronology cannot explain why some poets keep a larger percentage of their poems intact—at least in regard to the arrangement of stanzas—than do others. We still need an explanation of these variations (index TR). Did some poets deliberately fix the ordering of stanzas in their works, while others—not from artlessness but by choice—give free rein to stanzaic transposition? Or do the unvarying poems reflect sheer luck with transmission? The proportions of unvarying songs were doubtless influenced by poets' expressed intentions, by the formal properties of individual poems, by trends in performance or in copying, and by transmitters' reverence for certain poets. Manuscript survival—a finger on the pulse of the transmitters rather than of the poets—proved by far the single most important factor in all measures of stability or instability.

Popularity Correlates with Rate of Transposition

The rate of transposition (TR) can be explained as a very close correlative of popularity (MS). When the question to be answered is, "Why do the 

poets vary in the proportion of their work that manifests stanzaic transposition?"^[11] we can say confidently that it covaries regularly with the number of manuscripts per poem. Nothing else correlates so strongly with rate of transposition. The number of manuscripts per poem is the key factor in predicting the degree to which a poet's work will manifest stanzaic transposition (Fig. A-5). With a correlation coefficient of .62 between TR and MS, even the conservative analyst can say that popularity "explains" about 40 percent (r² ) of the differences among poets in their rates of transposition. This is considered quite a strong correlation for statistics involving social phenomena, since the activities of human beings are always too complex to be accounted for by only two variables. The significance level for this correlation confirms its importance: the probability that these data could have come from a "no-slope population," that is, from a sample with no real trend, is .0018, or eighteen in ten thousand.

Stepping back to look at the relationships among primary data, however, we find that "percent transposed" (= T/P) versus "manuscripts per poem" (= M/P) is not the whole story. These two derived indices, both filtered through the number of poems (P), register as a kind of shadow of an impressively regular interrelationship between transpositions and manuscripts. The number of poems showing transposition (T) is directly proportional to the number of manuscripts surviving (M), as Figure A-3 in Appendix A illustrates. The odds of this correlation coming from a no-slope population are only one in ten thousand.

The close relationship between manuscript survival and transposition rate shows that, to a great extent, stanzaic transposition is a phenomenon of the transmitting culture. It also indicates that authorial interference (rewriting and republication of poems) was clearly not the only source of new versions. The alternative (that the number of poems showing transposition, T, "causes" M, the number of extant copies) is improbable: it would mean that the number of songs a poet revised and reissued after "first drafts" had been placed in circulation somehow tended to make him popular with his own and later generations or increased the tendency of his poems to survive in many copies.

Time—the period in which a poet composed his songs—is again somewhat discounted as a significant factor in determining the rate of transposition. Guilhem IX and Guilhem de Montagnagol, the earliest and latest poets in my survey, appear close together on the graph (Fig. A-6). The distribution of transpositions in proportion to manuscripts does not discriminate between earlier and later poets. 

If popularity were the only factor in determining rate of transposition, the graph mapping the relation of TR and MS (Fig. A-5) would show a fine line and not a wide band. The popularity distribution makes it possible to see the transposition rates in perspective and suggests further ways to explore the reasons for great differences in transmission among poets productive within the same span of time. It filters out, for example, the proliferation of manuscripts of Bernart de Ventadorn's poems when we compare him with his contemporary, Raimbaut d'Aurenga: even considering their respective popularity ratings, Raimbaut's lyrics are significantly more stable than Bernart's. Yet the spread between them is not as wide as the transposition figures alone might suggest.

In drawing the trend line for manuscript survival versus rate of transposition, one notices that poets on either side share certain stylistic tendencies: Bernart de Ventadorn, Peire Rogier, and Jaufre Rudel may not be equally popular or equally mutable, yet they all fall well above the trend line, requiring fewer manuscripts than the others to produce transpositions; none of them has been accused of practicing trobar clus . Peire Rogier is famous for warning Raimbaut d'Aurenga not to be so learned but to play the fool if he wished to please audiences. On the other side of the line, among poets whose songs change less than the number of copies warrants, we find Raimbaut d'Aurenga, Gmilhem de St-Didier, Giraut de Bornelh, and Arnaut Daniel—all of them innovators in versification, most of them identified as the heirs of Marcabru and Guilhem IX rather than of Jaufre Rudel. This pattern confirms Paden's idea, arrived at on the basis of transposition figures alone, that "stylistic difficulty correlates with low permutation in Marcabru, Raimbaut d'Aurenga, and Giraut de Bornelh, all poets associated with trobar clus —although the case of Arnaut Daniel, with higher permutation, tests the rule" (1979, 3).

When we take into account his popularity, Arnaut Daniel's above-average rate of transposition still places him well below the trend line, since a higher percentage of his poems remain stable than one would expect from the number of extant copies. Thus, the popularity/transposition distribution, as it divides the troubadours roughly into two fields, supports Paden's tentative rule about the effects of stylistic difficulty, especially since it makes it unnecessary to except Arnaut Daniel from the general pattern.

With regard to the number of poems with transposition versus the number of manuscript copies (Fig. A-3), only a few poets do not conform to a very regular statistical pattern. "Out of line" on the unstable side 

(with an excess of transpositions per manuscript) are Bertran de Born, Bernart de Ventadorn, Raimon de Miraval, and Peire Vidal. Who among us would be willing to declare Bertran and Peire incapable of "stylistic difficulty"? On the stable side (with an unusually high proportion of manuscripts to transposed poems) are Guilhem de St-Didier, Raimbaut d'Aurenga, Giraut de Bornelh, and Folquet de Marselha. Coincidentally, the first three of these were contemporaries in close literary contact; could they have shared some "secret weapon" against transposition? Raimbaut's poems tended to be abridged more than average but transposed less (Fig. A-4). Conversely, Bernart de Ventadorn's were transposed more than average but abridged less!

What constitutes stylistic difficulty has been partially explained, though not completely captured, by quantification of the two elements studied here: the length and "linkage" of stanzas. Songs with what I call "linked stanzas" have rhyme schemes that join together two or more stanzas in such a way that their sequence cannot be disrupted without a noticeable breach of pattern. These rhyme schemes and their stabilizing effects will be discussed in detail further on; at this point, it suffices to know that stanzaic linkage, by itself, stabilizes poems but not poets . For each poet,1 found the percentage of poems with linked stanzas and observed its influence on other elements of stability. This percentage of poems with linked stanzas (abbreviated LK+) can be taken into account when we try to explain variations of stability.

Stanza Length and Version Production

An unexpected result of my statistical survey is that the length of stanzas probably did affect the stability of troubadour songs. The longer the poet's average stanza length, the more manuscripts it took to produce a transposition (Fig. A-10). Stanza length also correlates strongly with version production, when we define "version" as any alternate array, be it an abridgment or a transposition.

The stabilizing power of longer stanzas may partly explain Giraut de Bornelh's success in "evading" mouvance —if one can say he "evades" it without implying that he tried to evade it. He rarely uses the fancy rhyme schemes that discourage transposition, yet by any index his work is extremely stable. Among the twenty-three poets I surveyed, he holds the record for the longest stanzas, averaging eleven to twelve lines per stanza (II.48). His rate of version production is among the lowest. Likewise, 

Arnaut Daniel uses very long stanzas (averaging 10.83 lines), rarely with stanzaic linkage. His version production is unusually low, and even his rate of transposition is low when we consider his popularity. "L'aur'amara," with its seventeen-lines-long stanzas, comes down in only two sequences in nine manuscripts (CDHNU and IKN² a); the frequency of its rhymes certainly qualifies as a kind of stylistic difficulty.

There was also, as scholars have long observed, a chronological trend to compose more and more lines per stanza (Fig. A-8). Although the trend to longer stanzas could be viewed as "progress" toward an identifiable "goal" of creating stable poetry, there are two major drawbacks to this view. First, on the average, stanza length increased by only two lines per hundred years. Such a slow trend cannot have been the reaction to a "problem" viewed as urgently important. Second, long stanzas did not imply the creation of fixed texts. They stabilized, they perhaps reduced the number of possible combinations and permutations, but they did not absolutely or effectively dictate the sequence of stanzas.

As a very slight trend among the poets surveyed, linked stanzas and long stanzas occur in inverse proportion: the longer a poet's stanzas tend to be, the less often he is likely to use stanzaic linkage.^[12] As we shall see, though, combining both devices tended to help stabilize troubadour poetry.

Instability and Chronology

If it were widely agreed among the troubadours that jongleurs habitually scrambled stanzaic sequence, that complex rhyme schemes would help to combat this tendency, and that such combat was desirable, then we might expect to see overall "progress" toward textual stability and an increase, over time, in the use of devices that promote stability. That is, we should see a chronological trend. My study examines how the factor of time (using the assumed midpoint of each poet's lifespan) was related to the other two elements in question—sequential stability and poetic form. The results do not confirm the notion of a widely accepted need to "remedy" the vicissitudes of transmission, because they offer only qualified support to its underlying assumption that "progress was made."

Versions per Manuscript and Chronology

How can one suppose a chronological trend and still explain the remarkable stability of Raimbaut d'Aurenga's songs (only 16 percent show 

transposition), when his contemporary Bernart de Ventadorn has a high proportion of unstable songs (76 percent show transposition)? At least one measure of variation in the stability of twelfth-century poets' works—the percentage of poems showing stanzaic transpositions—follows no historical trend at all (Fig. A-6). It seems even less likely that the causes of this variation are related to chronological progress.

There was no chronological trend toward altering stanzaic sequence in fewer and fewer troubadour songs, or toward producing fewer versions per song. In my original survey, which included only one thirteenth-century poet, Raimon de Miraval, I found no chronological "progress" in any area, except toward an increase in stanza length. Yet when I extended the data, adding thirteenth-century poets up to Guilhem de Montanhagol, with an assumed midpoint of life in the year 1250, a weak chronological trend toward conservatism in version production came to light. This was not a trend to alter fewer songs, but to produce fewer versions per manuscript. Version production for both transpositions and abridgments correlated significantly with time.^[13] As time went on, it took more manuscript copies to produce either a rearrangement of stanzaic sequence or an abridgment.^[14] Either performances or copyings slowly and intermittently became more rigid. That this trend manifests itself only when thirteenth-century poets are included suggests that it may be related to the rise of the book culture and its accompanying model of the text as a fixed document, an entity capable of verbatim inscription whether encountered as a written object or not (Stock 1983). It may point toward the kinds of changes Paden suggests in his work on the jongleurs (1984).

Our broader definition of version production, including abridgments as well as stanzaic transposition, covaries more closely with time than does the strict definition by stanzaic transposition. In fact, more generally, the definition of "version" that includes abridgment follows all the same trends, but with greater predictability, than does the definition using stanzaic transposition as its criterion. This suggests that the two phenomena arose from the same causes and that "versions showing transposition" are actually a subset of a more general and predictable set of "versions." The broader criteria for "versions" (I refer to them below as "stanzaic arrays") thus seem most useful for discovering what element of transmission may have decreased "version production" over time.^[15]

Many will hasten to conclude, based on the clear trend toward fewer versions per manuscript among later poets, that it must have been an "improvement" in the poets' ability to write down their compositions that 

tended to stabilize their works. But we can explain most of the trend by examining measurable variables of poetic form, without resorting to speculation about the poets' ability to write.

If we partition the graph of stanzaic arrays per manuscript versus year (WM vs. Y; Fig. A-7) into quadrants, we begin to see within them meaningful clusters of poets. This schema illustrates what each cluster has in common:

I extract the following conclusions.

Mnemonic rhyme schemes did not make early poets stable . Early poets with a high ratio of versions to manuscripts (quadrant 1) tend to be low in stanza length (seven lines or fewer) and high in stanzaic linkage (30–50 percent). Most of these also have a numerically small body of MS copies (M = 53 or fewer, where the mean M of poets surveyed is 188).

Jaufre Rudel and Pons de la Guardia both invite their addressees to reperform the text. Bernart Marti speaks out against Peire d'Alvernhe's pretensions to textual integrity in his boasts of making vers entiers (see Chapter 7 for full discussion). Thus, these three early poets are openly uncommitted to the concept of the fixed text. Guilhem IX and Marcabru, though relatively "early and unstable" and therefore appearing in quadrant 1 of the schema, lie close to the trend line, so they are "average" in stability when their early dates are taken into account. They are stabler than the 

others in their quadrant. While they use stanzaic linkage less often than do Jaufre Rudel and Pons de la Guardia, their kind of "stylistic complexity" does include unusual or coined rhyme words. Marcabru also begins to use the longer stanzas that significantly improve stability. His statements on poetics suggest he believed in literary property and the original version, at least for his own works (see Chapter 6, pp. 150–60).

Bernart de Ventadorn is a special case, an early poet appearing relatively "stable" as regards stanzaic transposition but relatively "unstable" if abridgments as well as transpositions are considered.^[16] His overwhelming popularity surely contributed to the creation of new versions in his poetry.

Some earlier poets using mnemonic rhyme schemes were especially "stable ." Early poets with a low ratio of versions to manuscripts (quadrant 2) tend to have high stanzaic linkage (30–50 percent) and moderate or high stanza length. These are the best "survivors." Their works passed the maximum of time with the minimum of "damage" before being solidified by the book culture that brought us existing manuscripts.

For these poets, we can say that despite the mnemonic device supposedly offered by stanzaic linkage, their poems were "difficult." These three make use of such fancy devices as derivative rhymes, coined rhyme words, and fixed or pivotal "refrain words."^[17]

Abandoning stanzaic linkage made poets' stability reliant on stanza length . Of those "middle poets" around 1180–1196 who rejected stanzaic linkage—though all of them showed that they could do it, and brilliantly, if they wanted to—the best "survivors" (lowest in abridgment and transposition) were those who elected to use very long stanzas.

An exception, Arnaut de Mareuil, has stability comparable to Arnaut's and Giraut's, though he uses the shorter stanzas typical of less stable 

poets. For transposed versions per manuscript (VM), though, he is less stable than Arnaut Daniel and more stable than Peire Vidal.

Complex and sophisticated poetry without mnemonics produces frequent transposition . The later poets of quadrant 3, with high ratios of versions to manuscripts, have very low stanzaic linkage (less than 15 percent) and merely average stanza length (eight to nine lines per stanza):

Moderate complexity with some mnemonics and with temporal proximity to the advent of book culture promoted stability best . Later poets with low versions per manuscript (quadrant 4) tend to have moderately high stanzaic linkage and moderately high stanza length:

Thus, seeing the drop in version production over time, we need not assume it was caused because poets were slowly learning to write. Part of the trend can be ascribed to tendencies toward stabilizing formal properties in poetry. Two "unsuccessful" strategies, if the poets are assumed to be deliberately fighting change in stanzaic arrays, were (1) to put all one's energy into stanzaic linkage using short stanzas, as the early poets like Jaufre Rudel did, and (2) to abandon stanzaic linkage and use short stanzas, as some of the unstable middle poets did. Stanzaic linkage did help poets who participated in the movement toward longer stanzas. And it did help poets who already had "time on their side" because the chansonniers began to be compiled during or soon after their lifetimes. But it did not produce stability for early poets using short stanzas, whose work had to weather a hundred years or more before it was copied into MSS I, K, or V —and much longer before it reached late manuscripts such as a .

Time, although irrelevant to the number of versions per poem, was somewhat relevant to the number of versions per manuscript . This supports the theory that there was no movement toward authorial revision (which would have created more variation in "versions per poem" to start from and, hence, would have multiplied the variation in "versions per poem" in transmission). It also confirms that, except for stanza-linked 

poems, which defy the trend, the formal features of the poems themselves, or of particular poets and their reputations, only partly determined the likelihood of version production. Rather, there must have been change (though slight and gradual) in the practices of the transmitting culture—whether in its media or simply in its attitudes regarding textual integrity. As we reconstruct the forces for and against textual stability, we revisit a struggle between stabilizing formal properties or attitudes toward poetry, and the transmitters' steady, usually victorious, process of destabilization.

Rhyme As a Means of Fixing Stanzaic Sequence

One type of "stylistic difficulty" that significantly reduces transposition is the use of rhyme schemes that link one stanza to another or that by some system of alternation dictate the placement of each stanza within the whole poem. Because the motifs of fin'amors can follow in almost any order without disturbing convention, an abrupt contradiction is usually required to betray a frachura in the razo, a "breakage" in the "line of argument." When the versification is inconsistent, however, breakage becomes much more obvious. With coblas unissonans (uniform stanzas, each obeying the same structure and using the same rhyme sounds) and singulars (in which each stanza has an independent set of rhyme sounds), some errors within a single stanza can easily be detected: the mismatched rhyme, the incomplete stanza, or the line with too many or too few syllables. Such errors can signal either a failure of the poet's control or a deterioration through transmitters' imperfect grasp of the poem's requirements. The versification can also alert the hearer to displacement of entire stanzas, but only when it groups or links them in some way.

In my quantitative study of the effects of linking rhymes on the rate of transposition, I found it necessary to treat all types of "poems with linked stanzas" as a group because of the rarity of individual schemata other than the pairing of "twin" stanzas, coblas doblas . Under the heading of "poems with linked stanzas" came coblas doblas (stanzas matched in pairs), coblas ternas and quaternas (stanzas matched in groups of three and of four), coblas alternadas (alternating stanzas), coblas capcaudadas and capfinidas (stanzas linked by head-and-tail rhymes or repetitions), coblas redondas (stanzas paired by a ring structure), and songs with any system of rotating rhyme or alternating refrain words (see Appendix B). 

An illustrated discussion of each of these types of rhyme scheme can be found in Chapter 5. Only about 19 percent of the 552 songs surveyed used linked stanzas (104/552), and nearly half of these were composed in simple coblas doblas (46/104, or 44 percent).

As a group, poems with linked stanzas resisted transposition much better than did those composed with uniform or unmatched strophes in coblas unissonans and singulars; they were much more likely to be preserved in only one strophic sequence. Among poems preserved in more than one manuscript, resistance to transposition is strongly related to stanza linkage: 51 percent (198/390) of coblas unissonans and singulars show transposition; among poems with linked stanzas the rate drops to 23 percent, or almost one in four (21/94). In terms of version production, linked-stanza poems also fare well: while the average unlinked poem produces. 2.2 different stanzaic sequences (863/390), the average linked-stanza poem yields only 1.6 (146/94).^[18]

Poems with linked stanzas were not, overall, much more frequently copied than unlinked poems. Linked-stanza poems average 7.7 mss./ poem, whereas poems without stanzaic linkage average 7.9 mss./poem. Yet a poem with linked stanzas was slightly more likely, if preserved at all, to be preserved in more than one manuscript: linked-stanza poems make up 19 percent of all poems surveyed but only 15 percent of unique-manuscript songs.

These findings present a serious challenge to the theory that the troubadours wished to stabilize their "texts" and therefore invented complex verse forms, using them deliberately to inhibit change in transmission. If the troubadours developed these linked-stanza forms for the purpose of protecting their works from outside influence, why did they use them so seldom? Of all the linked-stanza forms, why did they rely most heavily on coblas doblas? (In view of the possibilities for rearrangement without disrupting the formal requirements, coblas doblas seem to provide the least sequential security of all the linked-stanza forms. Any poet who wanted to use linked-stanza forms to stabilize sequence should, in principle, have favored schemes with stronger interlocking of strophes. Conversely, coblas doblas are the simplest mnemonic device, with no potentially self-defeating complexity; thus, their prevalance could still be used to argue in favor of purposeful preventive use.) We cannot expect the troubadours to have foreseen how the centuries of transmission, both oral and written, would treat their works. But the theory that complex verse forms were 

intended as a "solution" to "problems" of transmission depends on the idea that they tried, to some extent, to second-guess the transmitter and to use forms that seemed sturdiest based on their own experience with jongleurs or with scribes.

Since poems with linked stanzas were comparatively stable, as a group, I expected to find that poets who used them most frequently would score lower in their rates of transposition. Not so: the percentage of stanza-linked songs (LK) among a poet's works did not correlate with his rate of transposition (TR). Frequency of stanzaic linkage as measured by LK+ (an index limited to poems in more than one manuscript) actually tends to increase slightly with versions per manuscript (Fig. A-12). Habitual use of linked stanzas therefore did not by itself persuade a poet's retransmitters to preserve the stanzaic arrangement in his whole body of poetry. In other words, it did not create "reputations for fixity" that particularly inspired reverence for a given poet's exact, "authorial" array of stanzas. One might expect the opposite in a written tradition: that the more usually a poet organized his rhyme scheme in ways that dictate stanzaic sequence, the more precision in sequential copying would seem to be called for in all his works.

Other indices also teach us something about the behavior in transmission of poems with linked stanzas: poems with unlinked stanzas behave much more predictably. When I subdivided the versions-per-manuscript index into separate categories of "linked" and "unlinked" poems, the songs with linked stanzas resisted nearly every trend. The very existence of these correlations signifies predictability, dependence on known factors in transmission or poetic form. Poems with linked stanzas generally refuse to fit neatly into the patterns obeyed by their unlinked counterparts.

Thus, poems with linked stanzas protected themselves from transposition, but the habit of using them did not particularly improve a given poet's chances of stabilizing his opus—except perhaps when combined with other kinds of "stylistic difficulty." So few poets use them more than 20 percent of the time that their variations in stability must be explained as resulting from other causes. The most telling factor was circulation. The poet's appeal, his success with audiences, encouraged transposition more consistently than other influences discouraged it.

Nevertheless, the finding that stanza length and stanzaic linkage affect stability proves that the formal properties of each song could influence whether or not performers and scribes rearranged the song. Although a 

preference for linked stanzas seems to be mildly associated with in stability (so that stanzaic linkage, if conceived as a purposeful "solution," sometimes backfired), poems with linked stanzas do fare better as a group than do songs constructed without a built-in mandate for stanzaic sequence.

The concept of the fixed text evidently did come to light in troubadour poetry. Many poets did connect textual fixity with stylistic complexity and elaborate rhyme schemes. But textual fixity was an idea before its time. The transmitting culture overwhelmed and appropriated whatever most appealed to it. It relentlessly altered whatever it chose to, no matter how strongly safeguarded. No doubt the troubadours observed the ultimate power of the transmitters within their own lifetimes. True, they seem to have envisioned an ideal of the fixed text, worked out grand ways and aesthetically impressive devices to safeguard it, and produced prototypes of the "incorruptible text." But the ideal of the fixed text was impractical within their transmitting culture. The same poets who most dramatically proved the possibility of fixing the text through monosequential rhyme schemes—Peire Vidal, Arnaut Daniel, Rigaut de Berbezilh, Giraut de Bornelh—were ultimately the poets who relied least on stanzaic linkage for their poems' stability. These poets earned their stability (or, in Peire Vidal's case, instability) through a combination of factors of "stylistic difficulty." Some of these factors, unmeasured here, could prove revealing in some future study: for example, lexical rarity might account for some variation. My investigation isolates only two factors of stylistic complexity—stanza length and stanzaic linkage—both of which contribute measurably to the stability of troubadour poetry.

Many poets who took part in exploring the concept of the fixed text seem, ultimately, uncommitted to its practice. Arnaut Daniel uses stanzaic linkage in only two of his eighteen or nineteen songs; he pokes fun at "the firm intention" (the ideal of incorruptibility) in his sestina. Giraut de Bornelh was said to have undergone a conversion from trobar clus to trobar leu . I suspect he tried the fixed text and then abandoned it, since he knew the only truly effective way to preserve textual integrity was to limit the song's circulation—and what good is an unchanged song if it is unheard?

The reasons for the unpredictability of songs with linked stanzas will be partly explained by a closer look at individual metrical forms: those for which the poets claimed superior stabilizing powers and those which 

we might expect to stabilize the sequence of stanzas. In the next chapter, the manuscript tradition of exemplary poems in each type of rhyme scheme will give us much information about the advantages and shortcomings of such inventions as coblas capfinidas as safeguards against amateur revision. At the same time, this study will allow us, in some cases, to compare the poet's apparent expectations regarding textual stability with the actual outcome of manuscript transmission. 

Five—
Rhyme and Razo :
Case Studies

Uniform Strophes: Coblas Unissonans

Poets' boasts that their songs cannot be dismantled often occur in songs that lack formal safeguards for preserving stanzaic sequence. That is, the notion of textual fixity precedes—anticipates, but does not by itself attain—the capability to fix texts. Peire d'Alvernhe's "Be m'es plazens" with its "locked and closed words" boasts of fixity in coblas singulars (an "early" type where each stanza has its own set of rhyme sounds) preserved with as many sequences as there are manuscripts. Marcabru connects his veto on altering his poem with the song's workmanship, but the stanzaic structure within which he claims to "lassar la razon e·l vers" provides no clear bond between one stanza and the next, or even between half-stanzas:

Aujatz de chan, com'enans'e meillura
E Marcabrus, segon s'entensa pura,
sap la razon     e·l vers lassar e faire
si que autr'om    no l'en pot un mot traire.^[1
]         (Roncaglia 1957, 23, vv. 1–4)

Hear how the song advances and improves; and Marcabru, in
accordance with his pure intention, knows how to bind up the
argument in the versification so that another man cannot re-
move one word from it.

Marcabru appears to believe he has devised a way to stabilize the poem, to ensure its integrity: the term lassar clearly refers to precautions he has 

taken to prevent others from damaging his song. The reference to "advancement" (s'enansa ) claims sequentiality, especially when combined with the notion that "intention" remains "pure."

Retrospect over the manuscript tradition, however, proves this boast to have been over optimistic. Marcabru may dream of a sequential razo, but here at least he does not compose accordingly. The relationships among the stanzas are still flexible, both in their development of an argument and in their fulfillment of the versification. AIK and E give the stanzas in two different sequences, and the version of AIK is two stanzas shorter. Furthermore, the two versions do not even give the same stanzas: the transmitters take advantage of the two-part stanzaic form (1oa' 1oa' / 4b 6c 4b 6c) to construct new stanzas from available parts. This "fracturing" occurs within nearly every stanza. Dejeanne's 1909 edition invented yet a third sequence, reshuffling half-strophes to create stanzas that existed in neither written tradition and suppressing stanzas that existed in both. Roncaglia points out that scholars have long considered it one of Marcabru's most obscure poems. The obscurity vanishes, he claims, when the proper sequence of stanzas is restored (1957, 20). I intend to argue that, though the 1909 version may be "obscure," clarity and logical continuity characterize the stanzaic sequences of both AIK and E .

If razo means "argument" or "line of thought" here, its "bonding" to the versification is not apparent. Nothing in the poem's construction dictates a single necessary, continuous sequence of reasoning: these are coblas unissonans (uniform strophes), meaning that each stanza repeats the same pattern of rhyme sounds throughout. Each of two sequentially distinct versions develops a defensible logical progression of its own.

The half-stanzas are syntactically independent: only in the exordium does the subject-verb connection bridge the natural breaking point at midstanza. This sirventes proceeds by amplification, such that sequential relations between motifs remain flexible. The first half-strophe makes an assertion related to the central idea, "Evil and dishonesty are on the rise; society is going to the dogs." The second half-strophe usually does one of three things: (1) provides a cause or reason to support the first half, (2) presents a descriptive appositive to the first half, or (3) offers a supporting, secondary assertion with its own causal clause. Because of the half-strophes' syntactic independence and because each half-strophe gives an amplification closely related to the poem's theme, both versions (each with its own sequence of half-strophes) maintain logical continuity. Let us look, for example, at both versions of stanza two: 

A:    Per so sospir, car mouta gens ahura
      de malvestat c'ades creis e pejura:
      so m'en somon     qu'ieu sia guerrejaire,
      c'a lieis sap bon     quan m'au cridar ni braire.

Therefore I sigh, because much of society augurs an evil that is
soon increasing and worsening. Thus they summon me to be a
warrior, for it pleases them when they hear me cry out and sing.

E:    Pero sospir, car tota gens aura
      de malvestat que creis e pejura:
      C'aquist baron     an comensat estraire
      e passat     per un pertus taraire.

But I sigh, because all society augurs an evil that is increasing
and worsening. For these barons have begun to take [instead
of give], and they've passed through a very small aperture.

In both versions, the poet begins with a statement that his complaining song ("sighing") was prompted by a forecast of the decline of morals in society; he then gives an explanatory clause. In A, the explanatory clause amplifies sospir (I sigh) as a vocalization, by expanding on the reasons for singing: "mouta gen," foretelling decline, has asked him to fight "the evil." In E, the second half-strophe amplifies instead the decline of morals, pointing accusingly at "these barons" and specifying that generosity, in particular, has lost ground. In A Marcabru supports an aristocratic class in its battle against moral decline; in E, "tota gen" includes specific barons in moral decline. A 's version thus shows more literary self-consciousness and awareness of the audience, while E 's version moves quickly to the object of critique rather than dwelling on the speaker or its audience. Each version combines two half-strophes to make a "logical" strophe, and each offers a "logical" sequel to the opening stanza.

Again, for their third and fourth stanzas, both versions can be said to "follow a reasoned progression." Having devoted two strophes to the poet's qualifications and reasons for singing, A now narrows down the general theme of "social decline" to "decline among the youth."

A 3:   No·i a     conort en joven, mas trop surra,
       ni contra mort ressort ni cobertura,
       pos ist baron     an comensat l'estraire
       e passat don     per pertuis de taraire. 

There is no hope among the youth—instead they become very
corrupt—nor is there any resource or shelter against death.
For these barons have begun to take [instead of give], and
they've passed "giving" through a very small aperture.

Both A and E follow this comment on the decline of generosity with another division of society into two classes, this time "the worst and the best," both greedy. The strophe is roughly the same in both versions, but in A it is the fourth and in E the third.^[2]

A 4/E 3:    Li sordeior ant del dar l'aventura
               e li meillor badon a la peintura:
               la retraisso·n     fatz trist e sospiraire,
               c'a rebuzon     fant li ric lor afaire.

The worst run the risk of giving, while the best merely ogle the
illusion: the story of it makes one sad and sighing, for the rich
do their business backwards.

The sequence of AIK now develops this criticism of "the rich" by contemplating the corrupting effects of wealth. This version personifies moral qualities, using in its fifth stanza the image of the embattled fortress. As in war, justice is abandoned when money conquers, "when, because of wealth, a child is made emperor." For its sixth stanza, the AIK version continues the list of virtues lost through avarice; he who loses his virtues in this way is no better than a beast or a thief:

AIK 5:    Proeza·is franh e avoleza·is mura
               e no vol gaug cuillir dinz sa clauzura;
               dreit ni razon     no vei mantener gaire
               que d'un garson     fai avers emperaire.

Courage shatters, and greed walls itself up and refuses to ac-
cept pleasure into its enclosure; neither right nor reason do I 
see maintained in the slightest, for wealth makes a boy the
emperor.

AIK 6:    Qui per aver pert vergonh'e mezura
               e giet' honor e valor a noncura
               segon faisson     es del semblan confraire
               a l'erisson     e al gos e al laire. 

He who, for money, loses his sense of shame and measure and
casts honor and worth into indifference is to all appearances
the brother of the hedgehog, the bitch, and the thief.

Meanwhile E, for its fourth and fifth stanzas, uses many of the same parts to construct a similar, but not identical poem. E returns its focus to "youth," which, like the rich, "does its business backwards" by falling prematurely under the shadow of death; E delays pursuing the motif of avarice until its fifth stanza, where it again ties together "greed and youth" through the image of the rich boy emperor.

E 4:   Non a conort en joven, mas trop fura,
        ni contra mort ressort ni cobertura,
        Qu'ist acrupit     l'an gitat de son aire
        e de cami,     per colpa de la maire.

There is no hope among the youth—instead they sweep it
quite away—nor is there any resource or shelter against death.
For these abased men^[3]  have cast it [youth] out from its lineage
and from the road, because of a blow struck by the mother.

E 5:   Qui per argen pert vergonh'e mezura
        e giet'honor e valor a noncura
        Pretz ni valor     no vezem tener
        quan per aver     es un gartz emperaire.

If someone for silver loses his sense of shame and measure, he
also casts honor and valor into indifference. We do not see
worth and valor upheld, when through wealth a boy is made
emperor.

With these two strophes the version of E, while still "logical," begins to show flaws. In the second half-strophe of each, this version strays from the required rhymes (-on and -aire ). Some of E 's lines are hypometric. In E 4, the proximity of the verbal adjective acrupit to the noun camin sounds, to my ear, so echoic of one of Marcabru's famous coinages, crupen-cami (croucher-in-the-road), that it suggests the work of a disciple. But poets sometimes do reuse their inventions.

The continuity remains smooth and thoughtful. E 's strophe 4 elaborates on "youth" to speak of family, lineage, and legitimacy (if the "blow struck by the mother" refers, as seems likely given Marcabrunian misog- 

yny, to a betrayal of the father leading to illegitimate offspring). While a "lost generation" of decaying youth is dispossessed in E 4, the "rich boy" triumphs in E 5.

Thus, it is a flaw of versification rather than of razo that announces alteration. The two versions equally possess a "logical progression," each with different emphases and continuities. Only when the transmitter allows formal requirements to lapse can we be sure, with Mölk, that "a public of literary experts could be expected to detect the mutilations of the avol chantador right away and denounce them as such" (1979, 5).

Do these poorly rhymed half-strophes mean we should reject the version of E as completely "inauthentic," while embracing AIK as the "authentic" version? If we do, we have to sacrifice more than two whole strophes, nine lines that address specific men in powerful political positions and that establish the poet's relationship with several addressees: Guilhem X, Duke of Aquitaine, perhaps Marcabru's patron (Roncaglia 1957, 44); Alfonse Jordan, Count of Toulouse; and Alfonso VII of Castile and Léon. The configuration of "names named" in E is consistent with composition in 1133–1134 by a poet staying at Toulouse in the court of Alfonse Jordan (Roncaglia 1957, 46–47)—in turn not in consistent with what we know of Marcabru. The lines about the boy who through wealth becomes emperor, present in both versions, allude to an actual event of June 4, 1133.^[4] Dating through the poem's political allusions thus rules out long-term wear and tear as an explanation for the corrupt rhyme scheme of E 's version. We can detect errors without being able to distinguish the errors of the original poet from the errors of a reviser. Who is to say that Marcabru never sang a false rhyme? Roncaglia accepts both versions as authentic: he is "surprised" to encounter "probable authors' variants: more exactly, residues of an earlier draft" (1957, 20–21). In the terms of my interpretation, these are "residues" of two performances, not necessarily both given by the original poet but both equally "authentic."

Marcabru may have believed that since coblas unissonans were difficult to compose, they would also be difficult to alter. To some extent, they can expose imposture, though they do not prevent rearrangement. The enlacement within the stanza does permit us to recognize "false rhymes." The scheme of repeated rhymes, more difficult to "find" with each stanza, might tend to prohibit adding to the poem, but it puts no barrier in the way of those who would like to subtract from it. Marcabru does not ask that the poem remain in a single order; he asks only that nothing be removed. Neither kind of stability ensued. 

With coblas unissonans, the security of the poet's original composition, at least in regard to stanzaic sequence, has nothing to bolster it unless it can depend on the coherence of a necessary sequence of ideas in the razo, the "argument," whether narrative or meditative. Given the circumstances of transmission, the troubadours could not expect to depend on continuity as part of the poem's entirety until they had developed the means to secure it. Bernart de Ventadorn's "When I See the Lark Move" would have failed through its very success if it had been planned as yielding its message, and its joi d'entendre, only to the keepers of the MSS Q and U . Yet this version has been defended as cohering in a necessary sequence of stanzas, on the assumption that Moshe Lazar's edition represented a fixed text, a "poem" in the modern sense. In fact, Lazar merely follows Appel's choice from among eleven different sequential arrangements. This underscores that one cannot safely perform the kind of literary analysis that relies on linear development without first ascertaining that a particular linear development is part of what constitutes "the text." Only the exceptional troubadour poem develops with an inevitable thematic continuity. After all, it would have been foolhardy for a poet to compose, in coblas unissonans as easy to fracture and rearrange as those in Marcabru's "Aujatz de chan," a song whose capacity to make sense would be destroyed if a few stanzas should come, in time, to be transposed.

Doubled and Alternated Rhyme Patterns: Coblas Doblas and Coblas Alternadas

The "problem" of the jongleurs' tendency to rearrange the order of stanzas (and not all poets viewed it as a "problem") was far from "solved" by the invention of systems for grouping stanzas, whether in pairs with coblas doblas and alternadas or in series of three or four with coblas ternas and quaternas . Numerous poems have survived, apparently, as seriously damaged coblas doblas. Peire Rogier's "Dous'amiga, no·n puesc mais," described by Nicholson as "six coblas singulars (plus three which are missing)" (1976, 125),^[5] could not have given a sure sign of deficiency without the fact that two of its stanzas match: either the poem was once entirely in coblas doblas, or else Peire Rogier committed the ineptitude of trying coblas doblas and then giving up. Many more songs survive in an indecisive form of coblas doblas, with one or more unmatched coblas intruding after the system of pairs has been established.

Raimon de Miraval's sirventes joglaresc "A Dieu mi coman, Bajona" 

(Topsfield 1971, 316–320, song 39) appears to have consisted, at one time, of paired strophes also linked by a sort of rotation in the rhyme:

The poem (preserved only in CR and presenting no alternate structures in these manuscripts) passes for coblas singulars (where each stanza has its own rhyme sounds), but stanzas ii and iv, iii and vi, and iv and v nearly follow the rules for constructing paired stanzas. Did Raimon de Miraval compose this poem with a system, only part of which is now visible? And if so, what sort of system was it? Did the paired stanzas follow each other consecutively, as in coblas doblas (as AABB etc.), or did they alternate (ABAB or ABCABC etc.)? The MSS C and R agree, as they often do, too closely in phrasing and strophic order to make us suppose that they represent separate traditions. Their agreement is more likely to affirm the vitality of a variant (flawed) version than to authenticate their version as "original."^[6]

A weakness of coblas doblas, as well as of coblas alternadas, is that one can adapt the same material to either system. In transmitting a poem of Guilhem de St-Didier (324, 5: "Bel m'es olmais qu'eu retraja"), in which the pairing of coblas alternadas is dictated by a system of derivative rhymes in retrograde order, the transmitter of C has inverted coblas ii and iii. As a result, the form of the poem appears in C as two sets of coblas doblas followed by one set of coblas alternadas and one set of tornadas .^[7] One of Jaufre Rudel's poems in coblas doblas, "Quan lo rius de la fontana," generated some confusion: what if the version known to the compiler does not have an even number of stanzas? The system of doubling breaks down. Some versions begin with two initial coblas whose first line rhymes on -ana, and then follow the pair with three more coblas beginning with a rhyme on -ina; version 2b settles for a compromise, with two coblas in -ana, two in -ina, and then one each in -ana and -ina .^[8] Although coblas in -ina appear to have been remembered in abundance, there were not enough coblas in -ana to make up a second double stanza. In version 2b, then, Jaufre's poem manifests the same hesitation we saw in 

MS C 's version of Guilhem de St-Didier's coblas alternadas . Likewise, for Marcabru's "Puois la fuoilla revirola," the manuscripts AEIKa give a version in coblas doblas with one unmatched stanza; the longer versions of C and R, however, arrange the stanzas as coblas alternadas .^[9]

Despite these failings, the use of coblas doblas did inhibit transposition, reducing it to 25–30 percent (compared with almost 50 percent transposition in poems with unlinked stanzas).^[10] Were it not for the transmitters' freedom to reverse the order of two matching stanzas, we might explain the success of coblas doblas very simply: they check stanzaic transposition by reducing the number of movable units.

Mathematically, the number of permutations increases with the number of stanzas at an alarming rate. Only seven stanzas could, in theory, produce more than five thousand different sequences. Consider the values of factorial n (i.e., n! ), representing the number of different possible stanzaic sequences, where n is the number of stanzas:

The first stanza of any troubadour poem is stable in its position, so in practice we have (n - 1)! as the number of possible sequences. If the linkage among stanzas effectively divides in half the number of movable units, with the first two stanzas stable in their position, then the number of permutations drops to ([n - 2] / 2)!, a substantial improvement. When we allow for inversion within pairs, then this figure is multiplied by the number of ways we can arrange the invertible pairs.^[11] Six stanzas in coblas doblas, then, could be arranged in only 2 different ways, assuming a rule that the pairs must not be inverted, or in 8 ways without that rule; the same number of coblas unissonans can be arranged in 120 different ways, given a stable first stanza.

Coblas ternas and quaternas should, in theory, work the same way by reducing the value of n !. These forms, however, are rare, and the sequences are not as stable as one might expect: as the movable unit lengthens, it becomes increasingly difficult to control inversion within the unit. 

Paired coblas (i.e., doblas or alternadas ) with controls built into the rhyme scheme to prevent reversion of pairs should, then, be nearly invulnerable to permutation. Guilhem de St-Didier's "Bel m'es oimais qu'eu retraja," mentioned earlier (at note 7), simultaneously arranges its strophes in contiguous pairs and discourages their inversion. It does this through a system of derivative rhymes in inverse order (rimas retrogradadas ). Two strophes, in this system, comprise as close-knit a unit as a single strophe:

I. Bel m'es oimais qu'eu retraja
   Ab leugieira razon  plana
   Tal chanson que cil l'entenda
   A cui totz mos cors s' aclina
   Q'en la soa desmesura
   Mi part de si e·m des loigna;
   Tant m'es de merce  estraigna
   Que no·l platz que Jois m'en  veigna

II. Non sai si·m muor o viu o  veing
    o vau, c'a mal seignor  estraing
    serv que no·m met neus terme  loing
    Que ja jorn vas me s'a mesur .
    Et on ieu plus ll'estau cap cli .
    Negun de mos precs non  enten .
    Ans sai qe m'aucirant de  pla
    Li ben c'om de lieis mi  retrai

III. Trop si feing vas mi  veraja .
     Car una promessa  vana
     No·m fai . . .
        (Gm St-D 4, 1–19)^[12]

I. It pleases me henceforth to  repeat  a song with such an
easy flat  argument that she toward whom my whole heart
bends  may understand  it. For in her intemperance  she divides
herself from me and distances  me. She is such a stranger  to
mercy on me that she is not pleased when joy  comes  to me
from it.

II. I know not whether I am dying or living or  coming  or
going, for I serve a strange  lord who does not set me a date,

even a distant  one, when she might someday treat me  tem-
perately . And the more I bend  my head, the more she  under-
stands  not one of my pleas. But instead, I know I will be killed 
flat out  by the good things repeated  to me about her.

III. She pretends to be very  true  to me, for she does not
make me an empty  promise . . .

Out of ten manuscripts, only C disrupts the sequence of paired stanzas by inverting strophes 2 and 3. The poem thus "shows permutation" but still protects itself very well from sequential change. Guilhem's poem carries further the schematic idea introduced in Raimbaut d'Aurenga's "Cars, douz e fehnz," which uses derivative rhymes in inverse order within each of its coblas unissonans . Raimbaut's rhyme scheme echoes the content, which "comes back on itself": the male wren's low song makes the singer arise; the cricket's voice comes light as cork from the heavy stone wall and becomes, like the square stones in the wall, aligned and "squared." The singer in turn is raised high, rivaled in exaltation only by the cricket and the female wren. Ascribing to the voice such properties as weight and geometry, while at the same time vividly painting the scene of a summer's evening when the poet is wonder-struck by affinities between nature's singing and his own, Raimbaut "squares off" the masonry of his rhyme scheme. Each stanza is indivisible within itself because of its self-reinforcing construction:

Cars, douz e fenhz      del  bederesc
M'es sos bas chans,      per cui m' aerc ;
Cab joi s'espan      viu e noire
El tems que·lh grill      pres del  siure
Chantan el mur      jos lo  caire;
Que·s compassa e s'escaira
Sa vos, qu'a plus leu de  siura
E ja uns non s'i aderga
Mas grils e la bederesca .
        (R d'Aur 1,1–18)

The low song of the wren, on account of which I am exalted, is
dear, sweet, and fictitious to me, for it spreads abroad, exists,
and grows with joy at the time when the crickets, near the cork
tree, sing in the wall under the block of stone; so that their

voices, which are lighter than cork, are aligned and squared;
and let no one exalt himself so high except the cricket and the
female wren.
        (trans. Pattison 1952, 65)

Raimbaut's system—with each stanza part of a coblas unissonans scheme—produces extremely self-contained, autonomous coblas : solid "bricks" in a well-aligned wall. Guilhem's, going further, makes each cobla dependent on the next. In regard to transmission, we might say that "Cars, douz e fehnz" protects itself only against fracturing and transposition within the stanza (in six manuscripts it produced only one sequence), while "Bel m'es oimais . . ." protects itself from transposition within each pair of stanzas. One could also say that Guilhem is merely working with a longer (double-length) stanza; however, the melody generally marks the beginning of a new stanza.^[13] Guilhem's method, then, bridges two repetitions of the melody.

Head-and-Tail Linked Stanzas: Coblas Capcaudadas and Coblas Capfinidas

An easier way to forestall reversion in coblas doblas might be to join them by repeating the final word of each stanza in the first line of the next, thereby making them coblas capfinidas . This would strengthen the hearer's sense of continuity, both in sound and in meaning. Pons de la Guardia's "Be·m cujava que no chantes oguan" (Frank 1949, no. 7) uses these two techniques together. The poet claims to have composed the song expressly for Marqueza (de Cabrera, according to Frank) and her entourage, hoping that through them it "will be sung and learned in many a good place." Perhaps Pons believed that a song with its stanzas both paired and linked head-to-tail would be leu ad aprendre (easy to learn) and so used this form in deference to noble, but amateurish, transmitters. The manuscripts offer no evidence of either difficulty or ease for transmitters, however, since such information cannot be drawn from agreement between C and R, the only manuscripts in which the song survives.

Supposing that one has chosen coblas capfinidas as one's method for stabilizing coblas doblas, one might decide to discard the system of doubling as a redundant safeguard. Coblas capfinidas dictate the sequence of stanzas intended by the poet and might be expected to remain stable 

without the aid of other devices. An overview of the transmission of coblas capfinidas does show that the device was extremely effective in preventing transposition but that it was rarely used. The surviving works of the twenty-three poets in my survey (552 songs in total) include only eleven songs in coblas capfinidas, four of them by Guilhem de St-Didier. Only two have transposed stanzas: Arnaut Daniel's "Chansson do·ill mot son plan e prim," and Peire Vidal's "Nulhs hom non pot d'amor gandir"—the latter barely qualifying as capfinidas, with only four links extant among eight stanzas. Defective coblas capfinidas, however, are likely either to become unrecognizable as such or to lose their attribution to famous poets such as those in my survey.

Close scrutiny of the manuscript variants in Arnaut's "Chansson do·ill mot son plan e prim" will demonstrate both the value and the potential fallibility of "redundant safeguards" on stanzaic sequence. The song illustrates how the technique of coblas capfinidas, even when reinforced by a system of pairing and a system of rotation among the rhymes, can fail as a guard for stanzaic order.^[14]

I.    Chansson  do·ill  mot son plan e prim              1
      farai puois que botono·ill vim
      e l'aussor cim
      son de color
      de mainta flor                                                5
      e verdeia la fuoilla,
      chan e·il braill
      son a l'ombraill
      dels auzels per la  bruoilla

II.   Per bruoill  aug lo chan e·l refrim                10
      e per q'om no m'en fassa crim
      obre e lim
      motz de valor
      ab art d'Amor
      don non ai cor qem tuoilla;                         15
      ans si be·m faill
      la sec a traill
      on plus vas mi s' orguoilla .

III.  Val orguoill  petit d'amador
      que leu trabucha son seignor                       20
      del luoc aussor

      ius al teraill
      per tal trebaill
      que de ioi lo despuoilla;
      dreitz es lagrim                                            25
      et ard' e rim
      qi·n contr'amor  ianguoilla .

IV.  Per ianguoill  ges no·m vir aillor,
      bona dompna, yes cui ador;
      mas per paor                                                30
      del devinaill,
      don iois trassaill,
      fatz semblan que no·us vuoilla;
      c'anc nos gauzim
      de lor noirim:                                              35
      mal m'es que lor  acuoilla .

V.   Si ben m'acuoill  tot a esdaill
      mos pessamens lai vos assaill;
      q'ieu chant e vaill
      pel ioi qe·ns fim                                           40
      lai o·ns partim;
      dont sovens l'uoills mi muoilla
      d'ir' e de plor
      e de doussor
      car per ioi ai qe·m  duoilla .                           45

VI.  Ges no·m  duoill  d'amor don badaill
      ni non sec mesura ni taill;
      sol m'o egaill
      que anc no vim
      del temps Caym                                            50
      amador meins acuoilla
      cor trichador
      ni bauzador,
      per que mos iois  capduoilla .

      Bella, qui qe·is destuoilla                              55
      Arnautz dreich cor
      lai o·us honor
      car vostre pretz  capduoilla . 

I. A song whose words are smooth and supreme I will
make now that the willows [osiers ] are budding, and the high-
est peaks are the color of many flowers, and the leaf turns
green, and the songs and birds' cries are in the shadow  across
the woods .

II. Through the woods  I hear the song and re-echo it, and
so that no one may make this a crime by me, I work and polish
words of worth with the art of love, from which I have no will
to part; rather, although I slip, I follow its path, all the more
though it raises its pride  against me.

III. The pride  of a lover is worth little, for it easily trips up
its master from the highest place down to the ground, through
such trouble that it strips him of joy; it is right that he
should weep and burn and rhyme [burn], whoever  chatters
against love.

IV. I do not turn elsewhere for the sake of  chattering,  good
lady toward whom I make adoration; but for fear of riddling
which makes joy tremble, I make the appearance of not want-
ing you; for we never rejoiced from their lineage; it is unlucky
that I met  them.

V. Though it  comes to meet  me all haphazardly, my
thought attacks you there, that I may sing and have worth
through the joy we made there where we separated; because of
which my eye is often wet with anger and with tears and with
sweetness, for through joy I have that which  grieves  me.

VI. I do not grieve  for the love I yearn for, nor do I follow
moderation or measure; I alone am equal to this, for never
have I seen since the time of Cain a lover who less accepted a
trickster or capricious heart, and therefore my joy  reaches its
crown .

Fair one, whoever [else] might turn away, Arnaut runs
straight to where he can honor you, for your worth  reaches its
crown .

For this song, incontestably, there is one and only one "correct" stanzaic sequence. Arnaut has created, by building sequence into this rhyme scheme, the capability to incorporate a linear movement, a progression in the razo, into the "integrity" of his song. In this case, linearity is part of the theme as well as of the formal structure. 

Like Raimbaut d'Aurenga, Arnaut uses geometric figures as imagery—and here, the straight-line path represents the "right road" for the faithful lover and the faithful poet. From the first stanza, the poet makes us conscious of the extremes; he moves our line of vision from the "aussor cim" (highest peaks) to the shadowy forest floor, where vision is restricted and sounds, instead, mark the way. With a hunting metaphor he speaks of following the track of love along the slippery path, which he does not wish to leave, though he does lose his footing, and where the pride of lovers makes them fall also: "del luoc aussor / ius al teraill" (from the highest place down to the ground). He passes judgment in terms of what is "straight" or "right": "dreitz es lagrim . . . qi·n contr'amor ianguoilla" (it is right that he should weep . . . who chatters against love). He does not turn from this path ("ges no·m vir aillor"), though gossip makes joi unsteady ("don iois trassaill"). Although he follows the path of love, he does not follow that of moderation, "ni non sec mesura ni taill," for he recognizes no counterpart, as a lover unreceptive to trickery, since biblical times. The idea of linearity and "straight" progression follows through all the way to the tornada, where the speaker contrasts himself with those who swerve: "qui qe·is destuoilla / Arnautz drelch cor" (Whoever else might turn away, Arnaut runs straight).

Yet in eighteen manuscripts, only half preserve the correct order of strophes (ABGIKN² QSg and c ), while the other nine invert strophes 5 and 6 (CDEHLNPRS ). The thematic progression, the sequence of ideas, did not hold together by itself: strophe 5 presents a memory of past joi that makes the speaker weep, while strophe 6 describes the lover's immoderate fidelity by way of the rather sinister biblical figure of Cain. It may be that some performers or copyists chose to end the song not with Cain but with the remembrance of joy and the lover's tears: nine of the manuscripts make the poignant note of past joy their concluding stanza, and four of those nine end on that note by omitting the tornada . Arnaut evidently intended to contrast the "trembling" of joy in strophe 4, where it was shown as vulnerable to gossip, with the powerful joy of the speaker's memory (str. 5). With this "correct, original" sequence, the strong memory of joy brings tears (str. 5), but it also leads to a boast of the lover's immoderate fidelity (str. 6), in turn leading into the tornada, which signs the name Arnaut to a second boast of faithfulness.

Why was the stylistic device of coblas capfinidas not strong enough to prevent this transposition? Apparently what happened is that the linking mechanism, which should derive the cap of one stanza from the fin of the 

preceding stanza, began to weaken after the third cobla . Thirteen manuscripts invert lines 28–29, exclaiming "Bona dompna!" at the head of the fourth stanza and thereby dropping its formal linkage (with ianguoilla/ianguoill ) to the third stanza. At that point, the "redundant safeguard" of coblas doblas probably helped prevent transposition: the form dictated a matching stanza with the same rhymes as those preceding.

The device of the linking word weakens as the song progresses. Lines 36–37, the meeting of stanzas 4 and 5, appear in Toja's edition as follows:

mal m'es que lor  acuoilla .
Si ben m'acuoill  tot a esdaill

Woe is me that I met them. Although it meets me all hap-
hazardly. . . .

Although no manuscript inverts the stanzas at this point, memory does weaken here. R lacks lines 37–45 entirely, and even the manuscripts that do follow strophe 5 with strophe 6 lack the linking rhyme word in -uoill: the manuscripts instead give "Si be·m vau (vai)" or "Si tot val(s)," and it was only a lucky conjecture of Lavaud that restored the linking word acuoill in line 37.^[15]

At the next bridge between stanzas, the loss of linking words leads to an actual alteration of stanza order. Where in Toja's edition stanzas 5 and 6 are joined with duoilla/duoill (vv. 45–46: "car per ioi ai qe·m duoilla./Ges no·m duoill d'amor don badaill," Through joy I have that which grieves me. I do not grieve for the love I await), the transmitters again have difficulty in "remembering" the grammatical derivative of the final rhyme in -uoilla . Only three manuscripts retain a rhyme in -oill in the first half of line 46: P has doill, while LS have toill . A remarkable thing thus occurs in the manuscript S, which displaces lines 37–45: strophes 4 and 6 become properly linked coblas capfinidas:

S 36:              Cor ai queu lor o  toilla

S "37"(46):     Ges no·m  toill  d'amor dun badaill

I have desire to take  it away  from them. I do not take  myself
away  from love for which I wait in vain.

All other manuscripts abandon the idea of coblas capfinidas at line 46, where a bridge with duoill is needed: 

ABCDEGHIKNN² QSgc: Er ai fam  d'amor don badaill (Now I am
hungry for the love I long await)
R:  Er ai fag  d'amor un badaill (Now I have made of love a long awaiting)

In general, scribes or performers "remembered" the linking word well enough in its position as the final rhyme word in a stanza but easily "forgot" it in its unrhymed position at the beginning of the next line. Without these linking words, stanzas 5 and 6 become free-floating coblas doblas, prone to inversion within the pair. The poem's system of rotating rhymes also proves useless in preventing the inversion of coblas doblas, since it dictates only the sequence of pairs and not the sequence of stanzas. The combination of devices was probably responsible for delaying until late in the poem the disruption of sequence and for minimizing the production of versions. The beginning of a song is the most stable, and its stability weakens progressively. By the time the weakening linkage finally broke, only two different sequences were possible: 123456 and 123465.

Transmitters were often unobservant of coblas capfinidas, and only rarely did they attempt to restore an apparently lost set of linking words. A song of Rigaut de Berbezilh, "Tot atressi con la clartatz del dia" (Várvaro 1960, song 8), comes down in nine copies (HIKLL'NTa¹ d) as imperfect coblas capfinidas: of the four "links" needed to join five strophes, only the first two survive, and not a trace of lost linking words persists among the variants. It would have been easy for a writing redactor to create such bridges between the strophes, since the final rhyme of the cauda is the most common verb ending, -ar . Here is the way Rigaut's pattern works in the first two "bridging rhymes":

strs. 1–2:  desviar/desvia
strs. 2–3:  servir esforzar/servia

One could easily continue this pattern by deriving common verbs from the initial lines of strophes 4 and 5:

str. 4, v. 28: Domn'es de mi, qu'eu non auz  dir amia
str. 5, v. 37: Mas ieu consir si  merces m'en valria

to compose stanzaic links for lines 27 and 36:

str. 3, v. 27: Per far de mi zo c'om del sieu deu far
hypothetical redaction: ^*  Per far de mi zo c'om  disses amar
str. 4, v. 36: De que ia·m puosca Amors ochaizonar
hypothetical redaction: ^*  De que Amors ia·m puosca  mercejar . 

But no copyist, no redactor, attempts this in nine manuscripts. Perhaps they were accustomed to unfinished coblas capfinidas and could untroubled allow such inconsistencies to pass; or perhaps, like Bernart Amoros, they had scruples about emending passages they did not understand. In any case, this poem does not show permutation. Its first three strophes are fixed in their position by the surviving linkage, and only two remain vulnerable to transposition. Thus, only two different sequences are possible (12345 and 12354).

Similar to coblas capfinidas, and about equally rare, is the device of coblas capcaudadas . Here the rhyme sound rather than a refrain word links the end of one stanza to the beginning of the next. Only twelve poems in my sample used this system of linkage, and it was almost always supported by some form of alternation or rotation among the rhymes—partly because the technique itself assumes that a single rhyme will shift in position from one stanza to the next. Thus, the apparent success of coblas capcaudadas (of eleven songs in more than one manuscript, only three show stanzaic transposition) may be attributable rather to the accompanying patterns of shifting rhyme than to the rhyming "heads" and "tails" of adjacent stanzas.

Raimbaut d'Aurenga's "Joglar, fe qed eu dei" (Pattison 1952, song 33) illustrates the simplest form of coblas capcaudadas: each stanza takes up a different monorhyme, except in its last line, which foretells the rhyme of the next stanza: aaaaab / bbbbbc / cccccd and so on. Since linkage is confined to a single line, a slight error could lead to transposition; moreover, a writing transmitter (working without the aid of the melody), might easily mistake such a form for coblas singulars . Because the poem is preserved in only one manuscript, though, the stability of its form remains untested.

The next more complicated use of coblas capcaudadas is in conjunction with coblas alternadas: the two poems in my sample that use this method alternate only the "head" and "tail" rhymes (producing rimas dissolutas ). The manuscripts' treatment of Bernart de Ventadorn's "Non es meravelha s'eu chan" (Appel 1915, song 31) creates the impression that transmitters eagerly took advantage of every flexibility in the poem to individualize their versions, within the bounds of acceptable form. Each, evidently, wanted to sing it "melhs de nul autre chantador." The song was extremely popular; excluding the fragmentary version of F, it is preserved in nineteen manuscripts, with eleven different sequential arrangements. Despite this amazing variety, almost all manuscripts pre- 

serve the alternation of odd and even stanzas and thus preserve the rhyme link of coblas capcaudadas between adjacent stanzas. The sequences of strophes, as summarized by Appel (1915, 187), illustrate the remarkable strength of the "head-and-tail rhyme," despite transposition:

There are only a few deviations from the pattern that alternates odd and even stanzas. Manuscript R presents the poem as coblas doblas (two odd, two even, two odd); so does N, which leaves the first and last strophes unmatched (one odd, two even, two odd, one odd, one even). V begins with coblas alternadas but ends up with an extra odd strophe. The only others to disrupt the pattern of coblas alternadas e capcaudadas are G and Q, the "nonidentical twins," both of which place strophe 7 after another odd strophe. In G, however, strophe 7 and the tornada (half-strophe 8) follow in a logical sequence: having offered the bona domna his services, Bernart allays his fears about such submission by contrasting her with ferocious carnivores: "ors ni lion non etz vos ges, / que·m aucizatz, s'a vos me ren" (you're not at all a lion or a bear, who would kill me if I yielded myself to you; vv. 55–56). In G, thereupon, Bernart sends his poem "A ma tortre" (To my Turtledove) (in C, "A Mo Cortes," To My Courtly One). The lady is not a lion or a bear, but a turtledove; the connection of the first two animals with the tornada, as G has it, might have made it undesirable to omit strophe 7 even though it does add an extra odd stanza.

Let us stand back for a moment from the abstract evaluation of particular rhyme schemes to see how the meaning of a song like "Non es mera- 

velha" is affected by sequential changes in transmission. It is interesting to see to what extent the mode of composition gives the poem flexibility and actually accommodates sequential changes. In order to make this kind of analysis practicable, I can cite the poem only once in a complete form, and I will subsequently work with synopses of each stanza to show how the rearrangement of topoi affects meaning. First, here is the song as edited by Appel (1915, song 31):

I.    Non es meravelha s'eu chan                  1
      melhs de nul autre chantador,
      que plus me tra·l cors vas amor
      e melhs sui faihz a so coman.
      cor e cors e saber e sen                         5
      e fors'e poder i ai mes;
      si·m tira vas amor lo fres
      que vas autra part no·m aten.

II.   Ben es mortz qui d'amor no sen
      al cor cal que dousa sabor;                   10
      e que val viure ses valor
      mas per enoi far a la gen?
      ja Domnedeus no·m azir tan
      qu'eu ja pois viva jorn ni mes,
      pois que d'enoi serai mespres                15
      ni d'amor non aurai talan.

III.  Per bona fe e ses enjan
      am la plus bel'e la melhor
      del cor sospir e dels olhs plor,
      car tan l'am eu, per que i ai dan.            20
      eu que·n posc mais, s'Amors me pren
      e las charcers en que m'a mes
      no pot claus obrir mas merces,
      e de merce no·i trop nien.

IV.  Aquest'amors me fer tan gen                  25
      al cor d'una dousa sabor:
      cen vetz mor lo jorn de dolor
      e reviu de joi autras cen.
      ben es mos mals de bel semblan,

      que mais val mos mals qu'autre bes;          30
      e pois mos mals aitan bos m'es,
      bos er lo bes apres l'afan.

V.   Ai Deus! car se fosson trian
      d'entrels faus li fin amador
      e·lh lauzenger e·lh trichador                      35
      portesson corns el fron denan!
      tot l'aur del mon e tot l'argen
      i volgr'aver dat, s'eu l'agues
      sol que ma domna conogues
      aissi com eu l'am finamen.                         40

VI.  Cant eu la vei, be m'es parven
      als olhs, al vis, a la color,
      car aissi tremble de paor
      com fa la folha contra·l ven.
      non ai de sen per un enfan,                       45
      aissi sui d'amor entrepres;
      e d'ome qu'es aissi conques,
      pot domn'aver almorna gran.

VII.  Bona domna, re no·us deman
      mas que·m prendatz per servidor,              50
      qu'e·us servirai com bo senhor,
      cossi que del gazardo m'an.
      ve·us m'al vostre comandamen,
      francs cors umils, gais e cortes!
      ors ni leos non etz vos ges,                        55
      que·m aucizatz, s'a vos me ren.

VIII.  A Mo Cortes, lai on ilh es,
      tramet lo vers, e ja no·lh pes
      car n'ai estat tan lonjamen.

I. It is no wonder if I sing better than any other singer, for
the more my heart/body draws me toward love, the better I am
suited to its bidding. Heart and body and knowledge and sense
and strength and power I have devoted to it, and the rein draws
me so hard toward love, that I pay attention to nothing else.

II. He is truly dead who does not feel from love some sweet
taste at the heart, and what value has living without value, ex-

cept to annoy people. May God never hate me so much as to
let me live a day or a month longer, if ever I should be guilty of
annoyance or have no wish for love.

III. In good faith and without trickery, I love the fairest
and best woman. From the heart I sigh and from the eyes I
weep, because I love her so much, and on this account I suffer
harm. What more can I do if love takes me, and the prison in
which she has placed me no key can open but mercy, and I find
no mercy there?

IV. This love strikes me so softly at the heart with a sweet
taste: a hundred times a day I die of pain, and I revive again
from joy another hundred times. My ills are truly of fair ap-
pearance, so that my ills are worth more than another's good,
and since my ills are so good to me, the good will indeed be
good after the pain is gone.

V. Ah, God, if only true lovers were distinguished from the
false, and slanderers and tricksters wore horns on their fore-
heads! All the gold in the world, and all the silver, I would like 
to have given, if I had had it, if only my lady could have known 
how perfectly I love her.

VI. When I see her, it is very apparent in my eyes, my face,
and my complexion, for I tremble with fear, just as the leaf
trembles against the wind. I have not enough sense for a child,
I am so encumbered by love; and a lady may have great pity for
a man who is thus conquered.

VII. Good lady, I ask nothing of you but that you take me
as servant, for I will serve you as a good lord, just as if recom-
pense were coming to me. Here I am at your command, you 
honest and modest person, so gay and courtly! You are neither 
a bear nor a lion who would kill me if I offer myself to you.

VIII. To My Courtly One, there where she is, I send the
poem, and let it not trouble her because I have not been [there]
for such a long time.

The greatest number of manuscripts, ADIKO, supports the sequence just given, but without the tornada . The opening sequence of stanzas 1 and 2, where the second stanza justifies devoting oneself completely to love, by equating love with life and nonlove with death, is used by sixteen manuscripts: (1) "I sing better than other singers because I put all my energy 

into love"; (2) "And what is the good of living without love?" As a second stanza, (2) sounds like idealism, but as a final stanza it would sound like an admission of defeat, or even a suicide threat. Eleven of these sixteen follow with (3) "I love the best woman in the world; I am in a prison, merce is the only key and there is no merce ." The sequence 1-2-3, then, moves from the singer's claim of superiority (1) justified by his absolute placement of value in love as life (2) to more conventional topoi of the lover's preferring death to loss of love (2) and the motif of love as a prison (3). The prison of love and the "pleasant pain of love" are closely associated, so ADIKO and U follow with (4) "This love wounds me sweetly; I die and revive daily; this harm is better than another good."

At this point ADIKO move to a topic that is characteristically Bernart's, as well as a typically Ventadornian way of envisioning an impossible world: (5) "If only true lovers were distinguished from false! If only liars end cheats had to wear horns on their heads! I'd give all the gold in the world if she could know how well I love her." True love, then, does not give a visible outward sign, or at least not a sign blatant enough to let his lady understand him. He is misunderstood, then. He does not have all the gold in the world, nor do tricksters wear horns, so it seems unlikely that the lady will understand his love. In this medial position, the stanza (5) does not contain a strong self-reference to the canso itself.

Yet his love is visible, and does carry outward signs—how could the lady not have noticed? (6) "Whenever I see her, I show physical symptoms of love; I tremble like a leaf and have less sense than a child—a lady can pity a man so overtaken." Following on stanza 5, the argument is that though cheats do not wear horns, true lovers visibly love, and in general such a case calls for pity. The sixth stanza, then, in ADIKO serves as a generalized prelude to the lover's personal request: (7) "I ask only that you take me for your servant: don't be cruel like a bear or lion if I give myself to you." Thus ADIKO end with the lover offering himself up to what may be a hungry lion. The speaker hopes she is not, but the final lines of the poem (in this version) voice that possibility.

The sequence of Q is just like that of ADIKO except that it places strophe 7 in the middle (1237456) as it to avoid ending on the line "Que·m aucizatz, s'a vos me ren" (That you might kill me if I give myself to you). Or perhaps Q merely wants to run through the conventional topoi: why not follow "the prison of love" (3) with "the vassalage of love" (7) and then the "pleasant pain of love" (4)? Q disrupts the alternation of coblas alternadas in presenting this sequence. It ends with the pair 5–6, 

which first voice the wish that lovers and nonlovers were clearly marked, then points out that this lover is openly symptomatic and asks, indirectly, for pity.

Manuscript U (1234765) also follows the mainstream up to a point. After stanza 4 and "the pleasant pain of love," however, it makes the lover surrender: the speaker offers himself to the lady as her servant (7), though he fears her so much that he has to remind her (and himself) that she is not a lion or a bear. His subsequent descriptions of his love symptoms (6) seem to stem from this fear; even though his physical and mental weakness are quite outwardly visible signs of love (6), he still fears misunderstanding (5), wishing that false lovers were obliged to wear horns.

At midpoem, stanza 5 does not insist on the poem's self-reference. "Would that false lovers and true were clearly marked—would that tricksters wore horns on their heads. I would give all the gold and silver in the world if my lady could know how well I love her." When (5) appears as a final stanza, however (as it does in eight manuscripts), it draws from the convention of poetic self-reference in the tornada and closing stanzas and from the expectation that the poem will create a bridge between the speaker and the addressee. By giving persuasive testimony to the speaker's fidelity, it does "let her know how well he loves her," and it is therefore perhaps worth "all the gold in the world." As a final and self-reflecting stanza, then, stanza 5 is a request for payment to the performer. Six manuscripts have this as a final stanza (CMRUVa ), and three (RUV ) have it as their ending.

The sequence in CMa (12367458t) sharply challenges the kindness and goodness of the lady herself. We cannot say it is "unlike Bernart" to be so critical, since he is forever using the courtly conventions to undercut one another. The first two stanzas establish the superior devotion of the lover; the third claims superiority for the lady; yet meanwhile he is in a hopeless prison of love and there is no merce . By moving directly to (6), the symptoms of love and the generalization that such a man should be pitied, the poet raises the question, If she is the best of ladies, why isn't she unlocking his prisonhouse with her merce ? This question is reinforced when the speaker offers himself as her vassal (7) and then has to plead with her not to act like a lion or a bear. In making such a bond of fealty, the lord should know how to do his part as a civilized human being, without having to be told by his prospective vassal. Thus far the poet has kept the question of the lady's cruelty or kindness uppermost in our mind. Now when he speaks of the sweet wounds of love (4) he lets the 

hearer imagine the savage lady-lion-bear inflicting them. By using (5) as a final stanza, the speaker drives home the idea that, although he loves perfectly, he is after all misunderstood. It would take all the gold and silver in the world to make her understand: she is not only cruel, but insensitive as well. The tornada of C addressed "To My Cortes," makes one wonder if "My Cortes" can possibly be the lady in question; if it is, it is no wonder that he has "been away so long."

Manuscript G (1436578t), which like C defuses the danger of the lady, this time by addressing the song "To My Turtledove," follows the singer's initial boast with "the sweet pains of love." The sequence 1–4 highlights the verbal play in (4), where the meanings of the words good and bad are altered through love: now both refer to "good." This stresses the power of song to "enchant" reality, as much as the power of love. Stanza 2, the "suicide threat," is completely absent from this version: nothing so realistic as death intrudes on this world made of words. Next comes "the prison of love" (3) and the "visible symptoms of love" (6), both of which end with a request for pity (merce, almornas ). Having already stated the visibility of his own love symptoms, the singer now wishes that false lovers were more clearly marked (5). To clearly express his love would be worth all the gold in the world—gold that he pointedly does not have, and expression which he manifestly has accomplished. Thus in G, the singer (7) offers himself as the lady's servant, "as if he were to be recompensed." Since "Ma Tortre" is gentler than the lion or the bear, perhaps she will also be more generous.

The preceding sketch of a multiple reading of one poem in its many sequential versions has demonstrated that sequence remains essential to a given performance's meaning: every moment depends on moments that have gone before. At the same time, songs constructed with the stanza as the movable unit lend themselves well to a number of different sequences. Just as all the topoi of canso convention are interrelated, just so any given sequence of the topoi will yield a network of interrelations. Each stanza is syntactically autonomous, yet each is potentially linked to the others by a common ground in fin'amors convention. The integrity of the song, "It is no wonder if I sing . . . ,"is not harmed by alteration; on the contrary, through the many versions it realizes more fully its potential as a multifaceted song. The song both celebrates the performer's uniqueness and allows him artistic leeway in presentation. By contrast, Arnaut Daniel's "Song whose words are polished . . ." claimed a particular sequence, a linear progression in time, as part of its integrity. The manuscript tradi- 

tion of Arnaut's song shows that performers who had lost parts of the "original" entity attempted nonetheless (by forging new "bridges" between linked stanzas) to reinstate the continuity and sequentiality of the form. The contrast between their two apparent claims to stability is much sharper than the contrast between the two rhyme schemes. All else being equal, Arnaut's coblas capfinidas and Bernart's coblas alternadas e capcaudadas could have had comparable stabilizing effects.

Far less variety is possible when, with coblas capcaudadas, the system of rotating rhyme groups the stanzas in threes instead of in pairs. Bernart de Ventadorn's "Can vei la flor, l'erba vert e la folha" ("When I see the flower, green grass, and leaf," Appel 1915, song 42) suffers transposition in only two (CQ ) out of thirteen manuscripts. The rhyme scheme dictates the sequence of strophes, provided that the transmitter can distinguish between introductory strophes and concluding strophes, and can discern the pattern in the first three:

Manuscript C, with its usual waywardness, reorganizes the strophes into a ring structure: 12365478t. The only disturbances this might create for the listener would be the loss of coblas capcaudadas at the juncture of strophes 3 and 6 and the crowding of strophes in -olha near the end. The sequence of Q, 123475, suggests poor transmission (where stanza 6 was not available) rather than deliberate revision. This version weakens toward the end.

Although in his extant poems of secure attribution Bernart de Ventadorn seems to rely mainly on simple coblas doblas for stanzaic linkage, the compilers of chansonniers apparently characterized him as a composer of coblas capcaudadas and capfinidas . Among the uncertain or erroneous attributions to Bernart edited by Appel, nearly half have linked stanzas—there are four with coblas capcaudadas, two with coblas capfinidas, and two with coblas doblas (one of them incorporating rotating rhyme).^[16]

One poet whose extant work is dominated by coblas capcaudadas is Pons de la Guardia: three of his nine songs take this form. All employ some form of simple rotation in the rhyme, and none shows transposi- 

tion. Unfortunately, each of these songs has come down in only two manuscripts (CV, EV, and CR ), so very little can be concluded about their stability. The agreement of V with other manuscripts is, however, more meaningful than agreement between C and R: V is early, somewhat isolated, and (like Pons himself) of Catalan origin.

Systems of Rotating Rhyme

Marcabru seems to have developed simple forms of rotation in the rhyme scheme as a variation on coblas alternadas . Instead of two rhyme sounds alternating in every other strophe, song 2, "A l'alena del vent doussa," has three rhyme sounds that shift their position in each stanza, returning to their original positions in every third stanza. If we compare the schema for this song as reconstructed in István Frank's Répertoire métrique (1953, 733: 1) with other types of coblas alternadas, we can see a progression in the complexity of stanzaic interdependence, and in the interlocking of lines:

As if to tantalize us, "A l'alena del vent doussa" comes down only in manuscript C . The transmitters have not preserved any of the presumed rhymes in -or but have replaced them all with rhymes in -um, -an, and -on . This shows that C, at least, reflects a certain resistance to the increase in complexity brought on by the addition of a third element to coblas alternadas:  

We have no way of knowing how the rhyme scheme might have protected this song in other manuscript families. The version in C, however, suggests that early corruption in the rhyme scheme might have discouraged other compilers from including this song in their collections.

The trio of rotating rhymes is by far the most common kind of rotation. Sometimes it appears as coblas capfinidas or capcaudadas; it stabilizes songs by linking stanzas in groups of three, each group having its sequence fixed by some pattern of progression.

Bernart de Ventadorn's "Tant ai mon cor ple de joya" (song 44) displays a simple yet effective system of rotation, combined with a somewhat longer stanza than is usual for Bernart. Only two rhymes change; the rest remain unissonans:

Each stanza converts the old "b" rhyme into the new "a" rhyme and introduces a new "b" rhyme that will in turn carry over into the next stanza. If the singer knows the stanzas, he knows in what sequence they are intended to be sung.

In eleven manuscripts, the song undergoes no stanzaic transposition, although as Appel's edition shows, there is a great deal of transposition of individual words in unrhymed positions:

Some variants offer a similar idea in different words:

Others change the meaning entirely, yet remain true to the rhyme sound:

Substantial variation of this kind is the rule rather than the exception in very popular troubadour songs, even those that do not show transposition in the stanza order.

An interesting feature of this song, a device not uncommon in the songs surveyed, is the fixed word amor that appears at the end of each stanza's ninth line. While other rhymes are shifting and changing one word stays in its place, serving as a steady axis around which other rhyme sounds pivot. The fixed word gives unity (as it did in Marcabru's "Vers del lavador"); it also doubtless compounded the difficulty of composition.

Some poets increased the number of these fixed rhymes. Raimbaut d'Aurenga's "Ar resplan la flor enversa," which uses eight rhyme words and their derivatives alternately in retrograde order, seems to reach the extreme limits of this technique and to proclaim itself the ultimate—and thus the last—of its kind. The song (which, in effect, is linked only as coblas alternadas ) survives in eight manuscripts and in only one sequence. Its tornada suggests that Raimbaut did want the song to be protected from outside influence. He has in mind a particular addressee, whose musical ability qualifies her as a worthy retransmitter:

A midons lo chant e·l siscle
clar, qu'el cor l'en entro·l giscle,
selh que sap gen chantar ab joy
que no tanh a chantador croy.
        (R d'Aur 39, 45–48)

To my lady (so that the shoots of it will enter her heart) let him
sing and warble it clearly who knows how to sing well and
joyfully, for it is not fitting for a bad singer.
        (trans. Pattison 1952)

As the "shoots" of the song "enter the lady's heart," she learns it par coeur in her turn. Raimbaut specifies that it should enter the heart of a competent singer, "que sap gen chantar ab joy," and not a bad one, "chantador croy." He then sends the song to two people: "doussa dona" and "Jocglar"—the latter perhaps the trobairitz Azalais de Portcairagues, who had the skill to maintain or improve the song rather than damage it (Sakari 1949).

The ultimate in difficulty is to integrate these mots-refrain into a pattern of rotating rhyme. Giraut de Bornelh does this so sparingly that it is 

hardly noticeable: in two songs, he uses one or two refrain words that shift their positions in alternate stanzas.

At the opposite extreme is Arnaut Daniel's sestina, with six refrain words "rotating" through six stanzas. Each stanza dictates the sequence of rhymes in the next. The song opens with a boast of superior textual stability:

Lo ferm volers q'el cor m'intra
no·m pot ies becs escoissendre ni ongla
de lausengier, qui pert per mal dir s'arma.
        (Arn D 18, 1–3)

The firm intention which enters my heart, neither the beak nor
the claw of an evil-speaker can break into fragments—an evil-
speaker who by speaking ill loses his soul.

Neither "orally" nor "manually" (not by beak or by nail) can an ill-speaking lausengier "fragment" Arnaut's "firm retention." It was true: although Toja calls the manuscript tradition of the sestina "confusa e contaminata" (1960, 378), the strophic order remains stable in all twenty-four manuscripts. Only one (V ) omits the tornada . The sestina attributed to Bertran de Born (P.-C. 233,2), preserved in only three manuscripts, suffers no transposition; however, two of the three (D^a H) give only the first five stanzas (Paden, Sankovitch, and Stäblein 1986, 403).

The sestina is the most famous scheme with rotating refrain words, but it is neither the first nor the most elaborate. Rigaut de Berbezilh's "Pauc sap d'amor qui merce non aten" (song 7) both alternates and reverts its refrain rhymes to create an extremely complex rotation using eight shifting refrain words:

Pauc sap d'Amor qui merce non  aten,
des qu'el consen
qu'om suffra e atenda,
qu'en pauc d'ora restaura et  esmenda
totz los mals traigz  qu'a faigz lonc temps suffrir;
per qu'eu voll mais ab fin'amor  murir
que senz amor aver lo cor  iauzen,
qu'aissi·m fadet Amors  primeiramen .
        (R Berb 7, 1–8)

Little he knows of love who does not expect grace, after he
consents to suffer and wait, for in a short time she restores and

amends all the ill treatments she has made him suffer for a long
time; that is why I would rather die of perfect love than to have
a joyous heart without love, for in this way love enchanted me
at the start.

The thematic words of the song create a balance of sadness and happiness, completion and expectation, harm and restoration. The collocation mal traig appears in stanzas 1, 2, 5, and the tornada, so it functions as a sort of sporadic internal refrain. It may well be that the range of meaning in the pair mal traig/esmenda includes the idea of transmitters' errors and emendation. Translating the terms of love into the terms of poetry, one finds here a song that complains bitterly of "the mutilations of the avol chantador, " accepts them as inseparable from poetry, and begs for "amendment" that will restore the joi of the poem. The tornada asks the lady to "know" the poet's account of the ill treatment that comes to one who waits (or "expects it"):

Miels de domna, no mi laissatz murir
quar mais non es maltraigz mas de suffrir,
per qu'eu volgra, s'Amors vos o cossen
que'n saubeses qual mal trai qui aten.
        (R Berb 7, 41–44)

Better-than-Lady, do not let me die, for there is no ill treatment
worse than to endure; therefore I wish, if Love permits you
this, that you should know what pain he who waits endures.

"Pauc sap d'amor" is preserved only in AIKa, and thus their agreement on the order of stanzas probably owes as much to close manuscript filianon as to the song's built-in controls. Although he anticipates the sestina, and even outdoes it in the sheer number of recurring rhyme words, Rigaut still did not carry the system to its limits. Peire Vidal won that distinction.

As the ne plus ultra of "rotating rhyme," Peire Vidal's "Mout m'es bon e bel" (P.-C. 364,29), with sixteen rhyme words rotating systematically in eight 12-line stanzas, cinches up its already incorruptible sequence by linking the stanzas head-and-tail (for coblas capcaudadas ). As a further complication, in strophes 2–7 some of the refrain rhymes move in pairs: even if the rhymes were not refrain words, we would have two sets of four stanzas with rotation where, from one stanza to the next,

Although the rhyme scheme nearly requires advanced mathematics to schemanze, the positional shifting of refrain words is so intricately regular that the pattern of the first strophe would predictably reappear, having come full circle, in a hypothetical seventeenth strophe (Carroll 1970, 345; see also Chambers 1985, 144–145). At that point, the sequence of positions (indicating the line in which a particular rhyme word appears as it moves from stanza to stanza), and the sequence of rhyme words (indicating which of the sixteen rhyme words appears in the first line of the stanza) coincide, returning simultaneously to "1."

The poem can be read as the elaborate offering of a speaker who proves his devotion as lover by his achievement as poet. He identifies himself with the song, which "renews itself" (by its principles of rhyme renewal) through the joi de trobar:

Qu'ab joi lonjamen
viu e renovel
Co·l fruitz el ramel
Quan chanton l'auzel.

Because for a long time, with  joi,  I live in renewal [or: I live
and renew] like the fruit in the branches when the birds sing.

This renewal's association with bird song affirms its reference to the pattern of rhyme. In this context, the "wholeness" of the speaker's enjoyment, and the assertion that "everything seems easy," recalls the vocabulary of textual integrity and of "difficult poetry":

Qu'e mon cor ai folh' e flor
Que·m ten tot l'an en verdor
Et en gaug entier, per qu'eu
No sen re que·m sia greu.

For in my heart I have leaves and flowers that keep me in
greenery all year, and in entire enjoyment; therefore I feel
nothing that might be hard for me.

The speaker's insensitivity to physical reality, achieved through his absorption in the poem ("en gaug entier," where "integrity," "wholeness," characterizes the pleasure of composition), is associated with difficult poetry, with trobar natural, and thus with the primacy of the original version (see Chapter 6). It is the lady's sen, her understanding of social and 

poetic conventions, that explains her ability to appropriate something that was once difficult (greu ) for her. If the poet speaks for the poem, then her native rank and intelligence have fitted her to overcome the problems of learning a complex poem and to "make it her own":

Quora que·lh fos greu,
Ara·m te per seu
La genser sotz Deu
E del melhor sen.

No matter when I was difficult for her, now she keeps me for
her own, the most noble woman under God, and of the best
intelligence.

Several times he subordinates his love for the bird song to his fulfillment of the lady's will:

Mes ai lonjamen
Mon cor e mon sen
En far son talen
Plus qu'en chan d'auzel.

I have for a long time set my heart and my intelligence on
doing her will, more than on bird song.

Clearly, he has made a serious pursuit, both intellectual (sen ) and emotional (cor ) of bird song—that is, of his poetry. The claim, then, is that he has applied himself more assiduously to this song than to the songs of those others, the birds, whose music is not dedicated especially to his addressee.

The poet's talen (motivation) to make a song merges with the recipient's talen (will), so that the words realize a mutual desire. The personal will of both parties is transformed into an intent to perfect the song: the poet's steadiness is the song's steadiness; its unswerving fulfillment of its own inner mandate—insofar as it embodies a speaking amador —demonstrates and in fact fully constitutes the speaker's unswerving obedience. Twice he boasts that he is "hard to change," and this applies both to talen (desire) and to the self as amador:

Qu'anc non camjai mon talen
Ni anc amador
No vitz qui·s camjes plus greu. 

For I never changed my intention, nor have you ever seen a
lover who changes with more difficulty.

The boast of superlative stability applies not only to the speaker as lover but to the poem itself; greu (heavy, difficult), by itself a dark attribute on the continuum from sadness to happiness, is transformed in each of its eight systematic repetitions into a benefit: the poem's very difficulty ensures its (and, depicted within it, the lover's) constancy.

If maltrag does refer to the song's potential mistreatment by singers—the structural isomorph of the lover's ill treatment by the lady—then the last four lines of the exordium can be read as a protest against the usual course of transmission. That this protest is an active one is confirmed by the song's very form:

Amaire sui e drutz sui ieu,
Mas tan son li maltrag grieu
qu'ieu n'ai suffert longamen,
qu'a pauc n'ai camjat mon sen.

A man who loves and a lover am I, but so severe are the mis-
treatments that I have suffered from it a long time, that I have
all but changed my sense.

With maltrag the lover almost "goes mad," the poet almost changes his "meaning." To literary maltrag, moreover, this poem is almost immune: it cannot change its sen (direction). Yet its means of preventing transposition of lines or stanzas, through schematic intricacy, has one drawback: it also cannot be duplicated. Peire has guaranteed the uniqueness of his song forever. Its form solves the problem of tampering for itself as a single song, but it does nothing to solve the problem for the rest of troubadour lyric.^[17]

Conclusions to Part Two: Sequence and Razo

The outward features of troubadour song, as they come down to us written in the chansonniers, have provided new and specific information about the conditions of performance and transmission, both oral and written. This is particularly true for the ways in which the technical form of the rhyme scheme could act—and sometimes conspicuously did not act—to stabilize stanzaic sequence.

Since stanzaic transposition is a phenomenon of the transmitting cul- 

ture, whereas rhyme scheme is a matter of the poet's particular art, we can observe in poems where transposed stanzas clearly violate the rhyme scheme an indirect clash between poet and later performer (or scribe). On the one hand the poet asserts a specific strophic order (or a limited range of acceptable orders); on the other the transmitter persists in habits of revision and rearrangement that had caused no trouble when applied to poems in which each strophe was an independent unit.

Judging by the claims Marcabru and Peire d'Alvernhe made for coblas unissonans, the poets could not fully anticipate which verse forms would prevent transmitters from "taking words out." Many of them seem to have begun to form the ideal of "textual integrity," without actually having considered that the sequence of stanzas might be a desirable thing to protect. A few, who may have tried to work with a logically sequential razo (tropeic development [Ghil 19791]), looked to linked-stanza forms for its protection only sporadically. Poems like Arnaut Daniel's sestina, Rigaut de Berbezilh's "Pauc sap d'amor qui merce non aten," and Peire Vidal's "Mout m'es bon e bel" seem to reduce to absurdity the idea of protecting a song's sequence by interlocking the rhymes. Such intricate and muscle-bound schemata produce their own peculiar harmony, but despite their individual success they could never have become standard for all poets who wished to compose fixed texts.

If the "fixed text" did become an ideal for certain troubadours, then that ideal came before its time. It was at least impracticable.

Stanzaic transposition brings a crisis in the literary interpretation of troubadour poetry. One cannot deny mouvance, so the question is, What does one make of it? The Gandensian formalist response to "mouvance in the manuscripts" and stanzaic transposition potentially disenfranchises literary interpretation by making the lyric not a work of art, not "a text in the modern sense":

Can the song be a unique work of art, a text in the modern sense? An important objection has been raised by Dragonetti and Zumthor: that in the manuscripts we find versions of the same piece in which stanzas appear in different and varying order. According to these scholars, 1) the order in which the strophes are arranged is of no esthetic importance; 2) the song is made up of an arbitrary juxtaposition of strophes, without continuity, without a sense of narrative or duration; therefore 3) the stanza itself and not the poem plays a central role in poetic composition, is the basic unit, the kernel so to speak, of courtly discourse. (Calin 1983, 76) 

Seeking to refute these three points—which, stated this way, bereave traditional literary criticism of its dearest "fundamentals of poetry" (from the French explication de texte, which requires the student to identify the "movement" or progress in a poem, to New Criticism, which desires continuity, unity, and the possibility of justifying the poet's choice and positioning of each word within his text)—Calin argues from modern reception theory: "When these poems were sung aloud by a trouvère or minstrel to a court public, the order in which the strophes were sung is crucial to our own and the audience's interpretation of the text" (ibid.). Extending Zumthor's quasi-mathematical formulation of stanzaic sequence, Calin argues:

In a series "n" plus "x," the audience's interpretation of "x" will be shaped by having heard "n" previously. . . . Each stanza should unleash an esthetic reaction, at its best a shock that would enlarge the horizon left by the preceding stanza. Therefore, the reception of the song, the audience perception, recreation, and response to it, will be totally different depending upon whether "n" comes before "x" or "x" before "n." (p. 77)

Clearly, a song expressing suicidal despair in strophe 2 and hope in strophe 6, with courtly topoi intervening, will impress the audience as gradually mending the speaker's state of mind, while in the same poem, performed with the hopeful stanza first and the suicide threat at the end, the intervening courtly topoi (the sickness of love, the wish to be a bird, the lady's silence, etc.) will have seemed gradually to destroy the speaker's strength and sanity. A competent performer could decide which of two such versions best suited his audience and his own disposition.

Even given my findings about the likelihood of transposition and the poets' means to control it, I would be the last to argue that "the order in which the strophes are arranged" is "of no aesthetic importance." The juxtaposition of strophes is not "arbitrary," not "without a sense of narrative or duration." Obviously, from the point of view of performance and reception, sequence is central to the character of each realization of a particular song. Yet sequence is "arbitrary" in the sense that the particular path and direction "of narrative or duration" is left to the libre arbitre of the performer or scribe. As I have shown in my study of their manuscript transmission, poems in coblas unissonans and coblas singulars come equipped with a high degree of flexibility as to their realization.

This flexibility, problematic and challenging as it is for the modern reader of troubadour verse, has proven indispensable to our purposes of 

conceptualizing medieval poetic creation. Given that, however, we need not abandon sequentiality as an "essential element" in interpreting troubadour lyric. The sequence of stanzas in a particular version will still constitute the "movement" of that version; the critic should merely take the additional step of distinguishing between "a version" and "a song" as poetic entities—bearing in mind that the song itself may embody plural movements. Finally—and Part Three is largely given over to detailed studies of this alternative—the "flexible poem" may not have been the only model for lyric known to the troubadours. Even if they could not successfully "swim against the tide" of transmitters' insistent habits, some poets apparently did conceive of the "firm lyric" as a desirable product of their poetic creation.

In Part Three, I will examine some of the stylistic ideals debated by the troubadours circa 1170 and relevant to the idea of a fixed text. Trobar clus will be the first considered (Chapter 6); the poets' discussion of other elements of "textual integrity" will be treated in Chapter 7.