PART ONE—
MAKING AND SENDING:
A VIEW FROM WITHIN THE LITERARY TEXTS

One—
Amar/Trobar :
The Vocabulary of Love and Poetics

A prerequisite to gathering and interpreting stylistic statements is the ability to recognize such statements when they appear. Having absorbed both the formal approach to trouvère and troubadour lyric and recent historical studies, scholars no longer assume that every troubadour worshiped a high-ranking lady who so stirred his emotions that they "overflowed" into verse. A few have begun to suspect that the Provençal poets' "lexicon of love" is in large part a lexicon of poetry—that the verbs amar and trobar are almost interchangeable. One of the best formulations of this hypothesis is by Edward I. Condren: "Because of what seems to have been a tacit understanding that the troubadour's love was almost always fictional, and because of the explicit belief that love was always the source of poetic power, it does not seem impossible that the word love and the entire vocabulary used to talk about it came to signify for some the creation of poetry" (1972, 191).

The circumspection of Condren's statement appears to have come from a wish to make only such claims as could be strongly substantiated in a brief article, not from lack of conviction. In fact, this view of love and poetry is ubiquitous in troubadour lyric; Condren finds it in songs by Bernart de Ventadorn, Jaufre Rudel, Peire Vidal, Gui d'Ussel, Guilhem de Montagnagol, and, most unmistakably, Raimbaut d'Aurenga. Condren clearly shows how at least one element of the troubadour's expression of love for "the implied lady"—his "desire" (talan )—can function within the lexical field of trobar: it is the poet's intention, his will to compose, his "desire" to make the best possible song. 

Paul Zumthor's quantitative analysis of the usage of the pair "to love/to sing" (aimer/chanter ) in a hundred trouvère songs identifies four ways in which these terms were used interchangeably by the northern French poets. Zumthor is much less reserved than Condren about his conclusion: "However one envisions this situation, it seems to me obvious that the connotative (if not the denotative) clusters represented by 'to sing' on the one hand, and 'to love' on the other, comprise a vast zone of intersection. I would scarcely hesitate to press the metaphor further and to announce bluntly that 'to love' 1 (referring to the subject of the song) is included in 'to sing' " (1971, 135). Zumthor cautiously includes the Occitan poet Bernart de Ventadorn, "whose songs more than the others' furnished the trouvères with the procedures of their art," among the poets conscious of "the circularity of the song," where "the poem is a mirror of the self . . . a mirror which is his eyes [i.e., the singer's eyes], which are Love, which is the Song" (p. 139).

One might object that the troubadours are exempt from Zumthor's conclusions because his concept of the "circular song" depends on the "impersonal I" of the trouvères: only when the "I" and the "you" of the poem recede can "it," the song, become its own self-fulfilling motivation. For the troubadours of around 1170, a highly individualized "I" dominates the song, and it is only the lady who vanishes. This objection can be met in at least two ways. First, "I sing" and "I love" can be equivalent or intersecting activities without weakening the identity of the speaker; on the contrary, the absence of an identified love object intensifies the song's reference to the performing self, and to what it performs. Second, the "individualism" of the troubadours serves their art: these poets have crafted for themselves stage personalities perhaps based on, but probably not identical with, their private characters (Sutherland 1965; Stevens 1978).

Although the intersection of amar and cantar cannot, without a great deal more and better evidence, settle the historical question of the origins of troubadour lyric,^[1] I am convinced that it can assist us in reading troubadour lyric. It can, for example, explain "the split role of the persona"—as lover and as singer—of the most famous (and most apparently sincere in his fin' amors ) of all the troubadours: Bernart de Ventadorn. In an analysis emphasizing the distinctness of the two roles rather than claiming identity for them, Mariann S. Regan (1974) arrives at surprising results: singer and lover receive unequal emphasis, because time and 

again in Bernart's songs the dissatisfactions of the unfulfilled lover are transcended and diminished by the artistic fulfillment of the singer.

For the most part, then, the argument that the vocabulary of amor intersects that of trobar has been restricted to the words amar/trobar/chantar themselves and their derivatives. Those who have carried the argument beyond this restriction have limited their claims to a single word (for example, as Condren explains talan, "desire" = "intention"), to a single poet's work, or to a single poem.^[2]

Nevertheless, the vocabulary of fin'amors (pure love) has been thoroughly studied and categorized, both in specialized articles and in comprehensive studies. In a review of such a work, Jean-Charles Payen (1978, of Cropp 1975) complains that "the ideological dimension" is lost when the poets' words are defined solely with reference to the experience of the lover and not of the whole man in his social and historical setting. The entire courtly vocabulary should be reinterpreted, Payen argues, with sensitivity to a broader range of connotation. I would add to this, since "the lover" is essentially a fictitious persona in this poetry, and since "the whole man" is not represented in the spectacle of the canso, that the "ideological dimension" through which the courtly vocabulary most needs reinterpretation is the artistic ideology of the cantador/trobador (singer/poet).

Before presenting details of this dual vocabulary, where words of love function to describe poetic activity, let me stress that in the chapters to follow I use this mode of interpretation sparingly and with caution. Just as one cannot assume that a poet is always singing about a lady, one cannot assume that he is always singing about his art. My overall purpose has not been to write the dictionary of poetic terms, but to discover how the medium of transmission—no less than the individual creators—made choices of style, form, and flexibility in troubadour song. It is helpful to be able to recognize the subtle affirmations of stylistic preference, and not only the explicit ones. To this end, I will briefly sketch some of the ways in which the troubadours use this ambivalent vocabulary to weave together their two dominant subjects.

We have already seen the most important field of terms that take both amor and trobar as their object: the poet's will, his motivation to attain his object.^[3] These terms have in the past been construed as always marking the influence of a lady. They are all words for intention. The most important of these are voler, talan, dezir (will, wish, desire), as well as cor  

and coratge (heart); the category includes words that indicate the attitude of volition, such as ardir (burn), s'eslaissar (rush [to]), and s'esforsar (force oneself [to]). Raimbaut d'Aurenga combines the poet's "desire" and the activity that fulfills it in one word: rima ("I burn" or "I rhyme").^[4]Joy is already well recognized as the creative impulse of poetry; few scholars still translate it jouissance .

The mental labor that precedes trobar is expressed by a group of words that have usually been taken as signs of the lover's pain, distress, and preoccupation with the lady.^[5] To the courtly definitions of these words one should add an artistic connotation, since the poet "meditates" on both love and song. Pessar and cossir (think, worry) direct themselves toward "creating the song" as often as toward "loving a lady." The selection of words and rhymes is governed by chausimen (discernment), the ability to triar son melhs . The first, chausimen, is easily confused with jauzimen (enjoyment);^[6] this word and triar (separate, discern) have both been read as referring exclusively to the selection of lovers or moral codes rather than of rhymes. Albir refers to a more decisive stage of thought (cf. mos albirs, "in my judgment"). A poet rapt in "thinking up" his song is "thoughtful": pensius, cossiros, or cabals .^[7]

Poetic control is generally expressed as a bond tied by Amors, by an unspecified feminine agent, or by the poet. Such verbs as lassar and liar (tie, bind), tener en fre (keep in rein), metre cadena (enchain), and serrar (constrain) take as their object sometimes motz or vers (word, poem) and sometimes the speaker, me, mos cors (my heart/body) or his volers (will, desire). Esmerar (to purify) is something the poet does either to his song or to himself; the same is true of melhurar (to improve). The song can also take the reflexive form of these verbs and "purify itself" or "improve itself." Both the poet and his poem "grow" (cresc ). Loss of poetic control, which the poet usually denies, participates in the same vocabulary as does discourtesy toward ladies: faillir/faillensa (fail, failure), franher/frachura (break, fracture), desmezura (breach of good measure). The usage of such terms suggests that at times the poet becomes identified with his poem: mos cors s'esmera (my heart/body purifies itself) makes very little sense if we imagine the poet taking a bath; he is purifying "himself," which is his "performing self," and thus his "performance":

Per que·us deu ben esser plus car
mas mos cors yes vos s'esmera
si que res no·i pot camjar.
           (R d'Aur, 26, 49–51) 

Therefore I should be more dear to you, since my being [ cor ]
refines itself toward you in such a way that nothing can be
changed in it.

The claim "nothing can be changed in it" is strongly associated with claims for the stability of the song. Thus, the poet's identification with his song can alert us to subtle declarations about his choice of style and his hopes for its success.

The poet's enemies—and he spends much of his time defaming them—are those who "speak ill," the lausengiers . Marcabru has also called them trobador bergau , "hornet troubadours." They are rivals, bent on destroying or stealing the poet's reputation, love, and songs. They do not "speak ill" merely by gossiping, but by singing badly: they twist one's words. Much like the comic figure of the blundering jongleur, and often indistinguishable from him as an avol chantador (bad singer), the lausengier makes songs pejurar (get worse), and instead of singing he shouts (cridar ), twitters (braire ), or bleats (bramar ). His evil speech (mal dich ) brings dan, mal trach , and trebalh —damage, abuse, and trouble. Some words for gossip serve also as words for publication: mazan (the noise of rumor) can mean applause; ressos (echo) can mean good fame or ill; espandre (spread) can refer to the circulation of a rumor or of a song.

The virtues making up the troubadours' "implied lady" are attributes of songs:

Ai, bon'amors encobida,
cors be faihz, delgatz e plas
frescha chara colorida
cui Deus formet ab sas mas!
           (B Vent 30, 50–53)

Ah, desired good love, body well made, delicate, and smooth,
fresh, colorful face, that God shaped with his hands!

If there is a lady here, she is an aspect of Amors (Love), as is the song; she has not quite materialized. Bon'amors conjures up a body and a face, both praised with the same words that poets use to boast about their songs: the body is "well made, delicate, and smooth" like a good melody, and the face is "fresh" and has "color," like a rhetorical trope. The greatest distinction between this being and a song is that she is God's handiwork, rather than Bernart de Ventadorn's. Yet since for us she exists only as this description, even she is Bernart's construct. Like the canso (love 

song), the implied lady is frequently described as doussa, clara, franca, and gaia —sweet, clear, free, and gay. It does not matter who she is, as long as she, like a song, is "refined and purified":

S'ieu lieys pert per son folhatge
ieu n'ay autra espiada
fina, esmerada e pura.
           (Mcb 28, 31–33)

If I lose her through her folly, I have glimpsed another one—
refined, purified, and spotless.

She is coind'e avinen, plazen, rizen (appealing and comely, pleasant, and laughing). So are many songs. If her love seems too steadfast ever to change, the speaker calls it veraia, certa, and segura —true, sure, and secure. So he may call his most stable songs. If she is "difficult," he complains that her behavior toward him is fer, greu, and escur —hard, difficult, and dark, like obscure poetry. He will not associate with a woman he considers vilana (lowly), or with obra vilana (lowly work).

Benefits conferred by the "implied lady" are implicitly verbal or musical as well as erotic. She "accords" favors or "accords with the speaker"—acort and acordar refer to "setting words to music." The benenansa and melhurament she offers "advance" the singer or "improve" the song. Solatz, plazer, and gaug —company, pleasure, and amusement—are to some extent conversational favors, bestowed in words. The speaker petitions for merce, the reward or prize for a good poem in the form of applause and money, of approving speech, or of a kiss.^[8] The addressee may "reward" the poet in other ways: by "understanding" him (entendre ), by "retaining" him (retener ), or by "understanding" and "retaining" the song, memorizing it (aprendre ), and "knowing" its contents (saber ).^[9]

Once having tried this approach, the reader will develop a sense of whether these meanings are applicable in a given context. I offer the method of "translation" from amor to trobar as a useful tool, capable of bringing out one of the "ideological dimensions" of troubadour language: the dimension of the poet's artistic intent. The troubadours are extremely self-conscious, and they are conscious of at least two and perhaps three "selves," for they portray the lover, the poet, and the song with which he identifies himself and which is the means of conveying all three portrayals. Their audience, moreover, was quite sophisticated by 

1170. Entendadors were amateurs who "understood" this art not only through practice in careful listening but in many cases through practice in singing and composing songs too. They were interested in the techniques of expression as much as in the thing expressed.

Reading what may appear to be a declaration about love as a declaration about the love song calls for a kind of interpretation with which the troubadours are quite familiar. Their idea of the poem as drogoman or messatge —as "translator" or "messenger," as diplomatic envoy—shows their awareness that the poem allows them to move between two worlds, giving form and sense to the shapeless "sighs and tears" of inarticulate emotion:

Ma chansos er drogomanz
lai on eu non aus anar
ni ab dretz oillz regardar,
tan sui conques e aclus.
           (R Berb 2, 45–48)

My song will be an interpreter, there where I dare not go nor
dare look with direct gaze because I am so vanquished and
downcast.

The idea of "the song as go-between" proposes a solution to the difficult double task of rendering the lover's "spontaneous emotion" believably in the highly artificial medium of rhymed verse with responding strophes. The song arbitrates between language and eros, imparting some of each of the other. Through its language, especially through the language of double reference, the poem can work out a treaty, act as diplomat, between separate "registers" of language; as ambassador, fluent in the idioms of both sender and receiver, it can "translate" from one code to another, addressing a feudal petition as an amorous one, interpreting the senses for the mind, making the language of poetic technique intelligible to "native speakers" of the language of social intrigue. The "game for one player" (whose object is artistic perfection) and the "game for two players" (whose object is fulfillment in love) share in language that applies to the larger "work" of the community—religion, economics, politics, as well as humbler activities like building and baking.

Bernart Marti does not despair over the incongruity of his two masks or over the need to end the lover's soliloquy with a request for praise of 

the poet's skill. He sees himself as a kind of civil servant, a political arbiter. The verb entrebescar, among other meanings, denotes the patterned intertwining of sound and sense. Bernart makes of it, and of Amors, a sort of whimsical proper name; he describes his own diplomacy as a poet who sings both of love and of poetry:

N'Eblon man yes Margarida
lo vers per un mesatgier,
qu'en lui es amor jauzida
de don'e de cavalier.
Et ieu soi sai ajustaire
de dos amicx d'un vejaire,
n'Aimes e·n L'Estrebesquiu.

N'Aimes e·n L'Estrebeschaire
son dui amic d'un vejaire
. . . ab l'entrebeschiu.
           (B Mar 7, 57–66)^[10]

To Sir Ebles near Margarida I send the verse by way of a mes-
senger, for through it love is enjoyed between ladies and 
knights. And, here, I am the adjustment officer between two 
friends of one mind, Sir Love and Sir Intertwine.

Sir Love and Sir Intertwine are two friends of one mind . . .
with intertwining.

For Bernart, as for many other troubadours, love is a matter of language. Except for the most intimate moments, it consists entirely of speeches, pleas, promises, agreements, avowals—in short, of words. Bernart does not confine this observation to the sublimated spiritual love that one often hears ascribed to the troubadours, those pleas and flatteries from afar. Even the kiss, the "reward" (merces ) of the petition for love, is for him a "linguistic act":

C'aisi vauc entrebescant
los motz e·l so afinant:
lengu'entrebescada
es en la baisada.
           (B Mar 4, 60–64)

Thus I intertwine the words and the music, refining [com-
pleting] them: tongue [language] intertwined is in the kiss. 

Bernart suggests that a good song provides its own "prize" for the cantador, if the amador 's prize is a lady's kiss. The "intertwining tongue" boasts simultaneously of the lover's success and of the poet's craft. Bernart shows us the one moment when lovers, after so much talking, are "tongue-tied."

Ultimately, eros is subsumed into language, and not the other way around. This book concentrates, therefore, on what the poets tell us about their words rather than about their loves. What was "the lyric" to them? What were its ideals, its limitations? The first step in understanding their conception of poetry is to understand their medium. 

Two—
Writing and Memory in the Creation and Transmission of Troubadour Poetry

No other body of poetry confronts its editors with so overgrown a wilderness of autonomous versions, nearly every surviving copy a unique variation, as does troubadour song. Some variants ornament the poem, others whitewash it; some betray a slip of the pen or of the eye, others record the ear's perception of two radically different meanings in approximately the same sound;^[1] still others, in a manner not explicable as the mechanical error or "vandalism" of scribes and compilers, evince a thorough recasting of the work. For example, a total of thirty-two distinct versions, distributed among six surviving songs of Jaufre Rudel, are now available in an edition of Jaufre's work by Rupert Pickens (1978). Readers accustomed to the monomorphic poem (either an editor's "best" choice or a composite) might hope that Jaufre's works were anomalous in their mobility. But it is not so: a study of 518 troubadour poems found "that nearly 40% of these songs show permutation in stanza order" (Paden 1979, 3). The percentage is much higher if we consider other kinds of variation: abridgments, alternative tornadas, and "fractured" stanzas.^[2] Further, the variability in form and content of each poem in the Jaufre canon, if it developed in transmission and not through authorial redaction alone, implies a set of distinctly unmodern assumptions that pervaded the entire transmitting culture from a very early stage through a very late one.

The notion of mouvance, of "the text in process of creating itself," is borne out by the chansonniers ' testimony to extensive change in most 

troubadour lyric. By comparison with Jaufre Rudel's, many of Bernart de Ventadorn's songs appear to be even more "fluid" because many more copies have survived. To take an extreme case, his "Non es meravelha s'eu chan / Melhs de nulh autre chantador" ("It is no wonder if I sing better than any other singer") comes down in nineteen manuscripts (excluding stray coblas ), and Carl Appel had to choose for his edition among eleven different arrangements of strophes (1915, 187). The thought that its self-congratulatory incipit made it a favorite among chantadors, who tailored it to suit themselves, is hard to resist. For the literary text, no less than for the musical notation, we must imagine either that the compilers of the chansonniers innovated drastically as often as they transcribed mechanically,^[3] or that most of these variations already figured in the "texts" before extant compilations were written.

The customary anonymity of medieval poets might be held responsible for textual instability in other medieval literary traditions. Since the songs of the troubadours purport not to be anonymous works, the question of the authenticity of many disparate versions must draw the serious reader, as well as the editor, into a study of the process of transmission itself. D'Arco Silvio Avalle faced this problem in his edition of Peire Vidal's poems (1960) and arrived at a system of authentication based on comparing the chansonniers ' selections and orderings of Peire's poems. Deducing an autograph collection—that is, an anthology written by the poet himself, the "libro del Vidal" of Avalle's introduction—the Italian scholar believed he could rate the extant manuscripts according to their stemmatic relationship to it. Pickens set out to follow this example in his edition of Jaufre Rudel, "by rigorous application of Lachmannian principles":

It soon became apparent, however, that not only can "authentic" texts not be discovered, much less "established" with a sufficient degree of certainty, but that, given the condition of the manuscripts and the esthetic principles revolving textual integrity affirmed by Jaufre himself as well as by his transmitters, the question of "authenticity," insofar as the meaning of the texts was concerned, was largely irrelevant. (Pickens 1978, 40)

What I intend to explore in this discussion of the transmission of troubadour poetry is the degree to which other poets and their transmitters shared the "esthetic principles involving textual integrity" that Pickens observes in the songs of Jaufre Rudel. There are two separate questions here: first, whether the troubadours did invite their transmitters to revise 

their poems; and second, whether the transmitters were inclined to do so even without an invitation.

A great deal depends on the breadth or narrowness of the "gap between creation and transmission"^[4] —that is, the degree to which, or the way in which, poets and transmitters agreed or were at variance on the subject of textual integrity. If the mutability of poems from one chansonnier to another reflected no more than one of the worst manuscript traditions in literary history, all because its scribes were so singularly and reliably erratic, then we would find no continuous lineage of "principles involving textual integrity." Rather, we would have to assume that the "chirographic folk" (Ong 1982) had disregarded the original poets' aesthetic principles in favor of their own and had allowed their pens to slip with extreme indulgence when it came to troubadour poetry.

One may overestimate the gap between creation and transmission, whether a gap of chronology or of mentality, if one forgets that a poem composed in 1150 must have been "transmitted" in one way or another before it was anthologized in 1254. True, many elements the poets valued in their songs had worn thin by the time the anthologies were written. The blank staves in the manuscripts where music should have been noted in; the bookish tracts from the period of compilation recommending a poetics so far removed from that of the classical troubadours that it used their works as a source of exemplary "vices"; the utter incomprehension of the late scribe of MS Q who filled all the blanks between cansos with fragments of miscollated tensos —all these instances point to more than the ordinary hazards of medieval dissemination.^[5]

Yet the inconsistency of extant written sources reflects health rather than sickness, insofar as it reflects an ongoing practice of singing troubadour lyric. Ong (1982, 81) points out the "startling paradox" that associates writing with death: writing destroys memory, and the letter kills. If there was a continuous performing tradition, perpetuating an aesthetic upheld by the troubadours themselves, then the diversity of current versions as manifested in writing, beginning around 1250, does not record the decadence of a dying art but the energy of a very lively one:

There is no reason to suppose that a nearly equal amount of diversity was not current before 1250. . . . Manifestations of texts are like bubbles accidentally rising to the surface which do little to evidence the currents and crosscurrents below. Thus, one cannot reasonably assume that a text by an author active around 1150 which is manifested in 1250 is any more "authentic" than one written down in 1350. Everything depends upon the 

quality of the traditions from which samples were selected by anthologizers. (Pickens 1978, 20–21)

"Authentication" may in some cases be available only in the poet's implicit consciousness and approval of the song's tendency to be essentially re-created in each act of its transmission.

Fortunately, the poets express themselves abundantly on the subject of the diffusion of their works: they speak of their intention to send their songs just as frequently as they announce the plan to compose a song. The place of honor given the topic of transmission within the poetry itself spotlights the very medium of the lyric as essential to its purpose and to the communication it contains. The exordium and the envoi often locate the addressee and the means of the song's conveyance inside the poem rather than outside it, just as the speaker himself is inside the song. Eloquence is the fabric of fin'amors (as well as of castimen in the sirventes ); in the medium of the poem, dichs e fachs (words and deeds) unite, for the speech itself is a deed in the negotiation it depicts. To "send the message" is to complete the song, entrusting it to the audience where it will do its persuasive work, give aesthetic pleasure, and live on. The fact of sending, then, has a legitimate place in the matter of the song: it is as well established a motif as praise of the lady and condemnation of those who speak falsely.

Medieval Editors: Jaufre de Foixa, Bernart Amoros, and Bernard de Clairvaux

Medieval compilers and theorists, when they expressed their principles of textual permanence, more than once cited the poets as the "wise men" whose example they have tried to follow. Jaufre de Foixa, in his Regles de trobar, appeals to authority (in this case, one who abdicates his authority) for a precedent to his own willingness to be corrected:

E si alcuna causa de repreniment hi ha ques eu non entenda, a mi platz fort que la puesquen esmenar segons rayso; car N'Aymerich de Peguilha m'o ensenya en una sua canço dient en axi:

Si eu en soy desmentitz
C'aysso no sia veritatz
No n'er om per mi blasmatz
Si per ver m'o contreditz;

Ans vey sos sabers plus grans
Si·m pot venser d'ayso segons rayso
Qu'eu non say ges tot lo sen Salamo.
                (Marshall 1972, 56)

And if there is any unintentional cause for reproach here, I would be very
pleased if they could emend it according to reason, for Aimeric de
Peguilhan teaches me this [attitude] in one of his songs, speaking thus: "If I
am deceived in this, so that this be not the truth, then no man shall be
criticized by me [on my account] if for the sake of truth he contradicts me;
   rather, I recognize his knowledge as greater if he can win out over me
with reason, for I by no means know all the wisdom of Solomon."

Only someone who "knows the truth" better than Aimeric, then, would venture to correct any false statements in his work; Aimeric authorizes such corrections, implying that criticism itself demonstrates superior knowledge. Jaufre de Foixa extends this principle, a matter of truth, to "any cause for reproach" that might require "emendation" in his treatise. He invites his readers to change his text if they find errors in it (again, the ability to criticize proves wisdom); more interesting, he believes that the troubadour Almeric extended the same invitation. The special privilege of the highly qualified reader here reaches its peak: permission to rewrite the book.

In similar fashion, the medieval compiler Bernart Amoros cites one of the poets, uns savis (a wise man), to confirm his idea that only the best qualified, those most nearly approaching the wisdom of Solomon, should undertake to emend a text: when the emendador "does not have understanding," then even the finest work is likely to be spoiled (see text below, points 7 and 8). In a sense, Bernart Amoros and Jaufre de Foixa share a tolerant view of intelligent, informed "improvements" to the text: Jaufre welcomes emendanon "segons rayso" (according to reason), and Bernart allows it, implicitly, wherever an "emender" has "ben aüt l'entendimen" (truly grasped the intended meaning). Each admits that his own version is not always letter-perfect.

In the light of current thinking about mouvance and transmission, Bernart Amoros's "curieuse note" takes on a significance hardly suspected since its publication by Ernest Stengel (1898, 350).^[6] Bernart's preface to a collection of troubadour songs survives in the paper manuscript a . As a first-person account of how a medieval compiler of troubadour poetry might operate, describing the choices open to him and the prin- 

ciples he followed, the passage from beginning to end offers rare and valuable information about medieval editing—an art that otherwise, in the chansonnier tradition, kept silent about itself and in its very anonymity disavowed its creativity as well as its conservatism, concealing its efforts behind fair copy under authors' names. Matfre Ermengaud—if (as Gustav Gröber believed) he did prepare chansonnier C for copying—prefaced it with no such frank statement of method.^[7] Because of the premises he takes for granted, as well as the ideas he presents as his own personal insight, Bernart Amoros's foreword is reproduced here in full.^[8]

Eu Bernartz Amoros clergues scriptors d'aquest libre si fui d'Alvergna don son estat maint bon trobador, e fui d'una villa que a nom Saint Flor de Planeza. E sui usatz luenc temps per Proenza per las encontradas on son mout de bonz trobadors, et ai vistas et auzidas maintas bonas chanzos. (2) Et ai apres tant en l'art de trobar q'eu sai cognoisser e devezir en rimas et en vulgar et en lati, per cas e per verbe, lo dreig trobar del fals. Per qu'eu dic qe en bona fe eu ai escrig en aqest libre drechamen lo miels q'reu ai sauput e pogut. (3) E si ai mout emendat d'aquo q'leu trobei en l'issemple, don eu o tiejn e bon e dreig, segon lo dreig lengatge. (4) Per q'ieu prec chascun que non s'entrameton de emendar e granren que si ben i trobes cots de penna en alcuna letra. (5) Chascuns horns, si truep pauc ne saubes, [no] pogra leumen aver drecha l'entencion. Et autre(s) fail non cuig qe i sia bonamen. (6) Que granz faillirs es d'ome que si fai emendador sitot ades non a l'entencion. (7) Qe maintas vetz, per frachura d'entendimen, venon afollat maint bon mot obrat primamen e d'avinen razo. (8) Si com dis uns savis:

Blasmat venon per frachura
D'entendimen obra pura
Maintas vetz de razon prima
Per maintz fols que-s tenon lima.

(9) Mas ieu m'en sui ben gardatz. Que maint luec son qu'eu non ai ben aüt
l'entendimen, per q'ieu non ai ren volgut mudar, per paor q'ieu non
peiures l'obra. Que truep volgra esser prims e sutils horn qi o pogues tot
entendre, specialmen de las chanzos d'En Giraut de Borneill lo maestre.^[9]

I, Bernart Amoros, cleric, writer of this book, came from Auvergne, from which many good troubadours have come; and I was from a town that has the name Saint Flor de Planeza. And I have spent a long time [traveling] through Provence and through regions where there are many good trou- 

badours, and I have seen and heard many good songs. (2) And I have learned so much in the art of poetic composition that I can recognize and distinguish, in rhymes both in the vernacular and in Latin, by case and by verb, the right trobar from the false. Therefore I say that in good faith I have written this book, correctly, to the best of my knowledge and ability. (3) And I have emended much of what I found in the exemplar, which I consider both good and proper to do, according to correct language. (4) Therefore I beg every man not to undertake to emend a great deal unless you truly find a slip of the pen in some letter. (5) Every man, if he knows too little, will not easily be able to get the intended meaning right. And I do not think that another man's error should properly be there. (6) For it is a great failing in a man who makes himself an emender unless he first has the intended meaning. (7) For many times, through a flaw in understanding, many good verses of the first workmanship and elegant reasoning have come to a bad end. (8) As a wise man says: "Through flaws in understanding of the first razo, pure works often come to be blamed, on account of many fools with erasers in their hands." But I took good care not to do this. For there are many passages where I did not really grasp the intention, and for this reason I did not wish to change anything, for fear that I might make the work worse. For a man will have to be extremely superior and subtle in order to understand everything, especially the songs of Giraut de Bornelh, the master.

In describing his background and travels, Bernart Amoros is not indulging in mere vanity: he is giving his credentials. By using the vida form for his autobiographical introduction, he presents himself as the peer of the troubadours, like them an initiate in the art of lyric composition. He counts on his reader's believing that the proximity of his birthplace to the origin of so many good troubadours, as well as his extensive travels in Provence and Auvergne, enhances his qualifications to distinguish "lo dreig trobar del fats." Certainly trobar was still a living art at the time of Bernart's visits, since it is clear from the passage that Bernart was not limited to traveling from library to library examining manuscripts: not only did he see many songs; he heard them also. He implies that the troubadours' native lands were a particularly abundant and reliable source of their songs; we may infer that each region not only preserved copies of its natives' works (the songs Bernart "saw") but also maintained them in a performing tradition (the songs he "heard").

When Bernart says that he has "learned so much in the art of composi- 

tion" that he can devezir (make decisions) both "en vulgar e en lati" (in the vernacular and in Latin), he does not necessarily mean only that he has become a good critic, an informed member of poetry's audience. He may also mean that he can versify. Five Latin hexameters bearing his name survive, at the end of a collection of Latin proverbs; they demonstrate that his enthusiasm extends to his having learned to create original lines in the style of what he edited:

Anno milleno ter centum ter quoque deno
Adjuncto terno complevit tempore verno
Dictus Amorosus Bernardus, in his studiosus,
Librum presentem, proverbia mille tenentem,
Milleque quingentos versus hic ordine junctos.^[10]

In the year one thousand three hundred and thirty and three
added, in the springtime, the man called Bernart Amoros, a
devoted student of these matters, completed the present book
containing a thousand proverbs, and a thousand five hundred
lines of verse here adjoined in order.

Bernart knew his contemporaries—both copyists and compilers—better than we know them today, and if he thought it necessary to make a special plea for careful, responsible copying, we can be sure that he expected emendadors by the dozen. He himself was one, and in good conscience. He begs that others correct only "slips of the pen," implying that writers of the day were in the habit of correcting more serious errors, perhaps even the kinds of errors (vicis ) for which the grammarians constantly reproach the anciens trobadors . What saves Bernart Amoros, at least in his own eyes, is that he made his emendations "in good faith," following both "correct language" and "the intended meaning." In his view, one must resort to mechanical transcription when one has not understood the passage. He admits that this has been his practice "especially with the songs of Sir Giraut de Bornelh the master." But how does Bernart proceed when he does understand the passage before him and finds it "flawed"? He emends "much." And it requires all the powers of this learned man, who has taken pains not only to learn the grammar but also to listen to as many songs as possible in their native lands, and who can write original verse if necessary, to make these emendations "rightly, to the best of my knowledge and ability."^[11]

Sylvia Huot, who has recently commented on Bernart Amoros's pref- 

ace, argues that this late-thirteenth-century text reflects an attitude that "can exist only within the framework of a written literary tradition; It is foreign to the semi-improvisational oral tradition." The case for this conclusion is well-stated in Huot's book: "For Bernart, vernacular and Latin poetry alike exist as a written tradition, governed by strict rules of poetic and grammatical form. His concern with textual emendations further reflects a consciousness of the poem as having a fixed form, composed by a gifted individual; the task of the copyist is to restore and preserve the work of the masters" (1987, 333). Yet Bernart's concern is not so much with preserving the literal utterances of the gifted individual (and Giraut de Bornelh is the only poet Bernart mentions as particularly deserving of non-emendation) but with preserving the "rightness" of the poem. As for the claim that Bernart's attitude belongs exclusively to a written tradition, we should not discount Bernart's "seeing and hearing" of many songs. It may be that in his "fieldwork" he conceived of himself as seeking the "right" version of any given song, assuming that the "right" version would of course coincide with the "authentic" or "authorial" version. Nonetheless, Bernart obviously was exposed to the "semi-improvisational oral tradition" and was keenly aware of the existence of variant versions.

It is unlikely that Bernart Amoros's plea for literal copying by the tasteless scribes of posterity reflects a sudden, late upsurge in zealous emendation. Even in the twelfth century we find much concern about the reproduction of musical texts. Bernard de Clairvaux addresses the preface of his treatise on song (De cantu ) both to future copyists and to future singers: "omnibus transcripturis hoc Antiphonarium, sive cantaturis in illo" (to all those who will transcribe this Antiphonary, or who will stag in it). St. Bernard apparently detects mouvance in the traditional antiphonary of the Cistercians and wishes to correct this situation, finding it unseemly for laudes Deo . Observing that the songs have long been entrusted to those who sang them,

Cantum quem Cisterciensis Ordinis ecclesiae cantare consueverant, licet
gravis et multiplex absurditas, diu tamen canentium commendavit
auctoritas

The song which the assemblies of the Cistercian order are accustomed to 
sing—though granted it is a serious and manifold absurdity—has never-
theless for a long time been entrusted by authority to those singing it, 

he urges that some regularity be adopted in singing by monks who, in all other respects, follow the Rule:

Dignum siquidem est, ut qui tenent Regulae veritatem, praetermissis al-
iorum dispensationibus, habeant etiam rectam canendi scientiam, re-
pudiatis eorum licentiis, qui similitudinem magis, quam naturam in can-
tibus attendentes, cohaerentia disjungunt, et conjungunt opposita; sicque 
omnia confundentes, cantum prout libet, non prout licet, incipiunt et ter-
minant, deponunt et elevant, componunt et ordinant. Unde nemo miretur
aut indignetur, si cantum aliter quam huc usque audierit, in plerisque mu-
tatum invenerit. (Migne 1844–1902, vol. 182, cols. 1121ff.)

If in fact it is proper that those who uphold the truth of the Rule, ignoring 
the directions of others, yet still have true knowledge of how to sing, refus-
ing the liberties taken by those men, who, paying more attention to simili-
tude than to nature in songs, disjoin coherences and conjoin opposites; 
and thus, confounding everything, as they will and not as they should, they 
begin and end the song, lower and raise it, put it together and put it in 
order. Thus, no one is astonished or indignant if he hears a song different 
from the way it was in the past, and finds changes in several places.

St. Bernard is speaking both of the melodies and of the words to these church songs; he finds an "excusatio facilis" (easy excuse) for "mutatio litterae" (change in the letter) in the fact that most of the repeated phrases in the antiphonary are nowhere to be found in the Scriptures. His desire to establish a fixed text for nonscriptural material, and the weight of the opposition to fixity in song that his preface is designed to overcome, show how deeply ingrained must have been the distraction between divinely inspired texts (to be reproduced prout licet ) and mere human, transitory utterances (to be reproduced prout libet ). Yet Bernard de Clairvaux seeks to give his own edition of the antiphonary some of the lustre and integrity of a sacred text. Like Bernart Amoros, the saint explains that he has made many necessary changes and now desires that no one after him undertake the same charge.

Raimon Vidal, in the Razos de trobar, observes that troubadour poetry has captured the imagination, and the memory, of the listening public. Audiences are not mere audiences—they participate. Everyone from the highest walk of life to the lowest has taken a daily interest in poetry, both composing and singing it: 

Totas genz cristianas, iusieuas et sarazinas, emperador, princeps, rei, duc,
conte, vesconte, contor, valvasor, clergue, borgues, vilans, paucs e granz,
meton totz iorns lor entendiment en trobar et en chantar, o q'en volon
trobar o q'en volon entendre o q'en volon dire o q'en volon auzir; qe greu
seres en loc negun tan privat ni tan sol, pos gens i a paucas o moutas, qe
ades non auias cantar un o autre o tot ensems, qe neis li pastor de la mon-
tagna lo maior sollatz qe ill aiant an de chantar. Et tuit li mal e·l ben del 
mont son mes en remembransa per trobadors. Et ia non trobares mot [ben] 
ni mal dig, po[s] trobaires l'a mes en rima, qe tot iorns [non sia] en re-
membranza, qar trobars et chantars son movemenz de totas galliardias.
                (cited in Poe 1984, 69)

All people—Christians, Jews, and Saracens, emperors, princes, kings,
dukes, counts, viscounts, contors,  vavasseurs, clerics, bourgeois, peasants,
small and great, every day apply their attention to poetry and song, either
that they want to compose it or they want to understand it or they want to
recite it or they want to listen to it; so that you could hardly find yourself
in any place so private or so isolated, be there few or many people, that
you would not hear singing one person or another or all of them at once,
for even the shepherds of the mountains, the greatest amusement that they
have is to sing. And all the goods and evils of the world have been placed in
remembrance by the troubadours. And never will you find anything, well
or badly said, once a troubadour has set it to rhyme, which will not forever
be in memory, for poetic composition and singing are movements of all
gladness.

But few of these amateurs know how to compose properly ("la drecha maniera de trobar"), so Raimon must take it upon himself to set them straight. As for his own book, Raimon is equally conservative and monumentalist: "Per qu'ieu vos dig qe en neguna ren, pos basta ni ben ista, no·n deu om ren ostar ni mais metre" (Therefore I tell you that in no detail, since it suffices and is good as it stands, should a man take anything out or put in anything more). Let the public wreak havoc on troubadour poetry, but let it leave untouched the manual for undoing that havoc! Raimon Vidal's Razos have been dated by Jeanroy at ca. 1200—half a century before the writing of the earliest extant troubadour songbook. We may notice that in Raimon's list of all the things people want to do to poetry—hear it, understand it, compose it, recite it—we do not see the words read and write . The song was a matter for remembranza —memory. 

If troubadour poetry was "a dying art" after 1254 in its native land, surely it continued to be a living art in its adoptive patriae —in Italy, Catalonia, northern France—in the regions, that is, where most of the early chansonniers were compiled. Ferrari de Ferrara, Bernart Amoros, and Miquel de la Tor were all, to some extent, practicing poets as well as makers of books. Ferrari de Ferrara

fo giullar et intendez meill detrobar proensal the negus om che fos mai en
lombardia e meill entendet la lenga proensal e sap molt be letras e scriuet
meil ch'om del mond e feis de molt bos libres e de beil.... Mas non fes
mais .II. cancos e una retruensa mais seruentes e coblas les el asai de las
meillor del mon e fe un estrat de tutas las cancos des bos trobador del mon.
                (Teulié and Rossi 1901–1902, 13:60–61)

was a longleur and understood better how to compose poetry in Proveçal
than any man who was still living Lombardy, and he better understood
the Proveçal language and knew letters very well and wrote better than
any man in the world and made some very good and beautiful books. . . .
But he only composed two love songs and one  retruensa,  but he composed
plenty of sirventes and coblas,  some of the best in the world, and he made a
selection of all the love songs of the good troubadours of the world.

The sixteenth-century scholar Giovanni Barbieri, according to his book on the origins of rhymed poetry, had the same text, with a few additions, on page 5 of his libro slegato (Barbieri 1790, 84). In the biography of Peire Cardenal (d. ca. 1272), Miquel de la Tor solemnly certifies that he personally wrote the sirventes of Peire following the vida, that he was in Nimes when he wrote it, and (as if he knew Peire) that the poet was more than a hundred years old when he died. Beyond that, the only trace of Miquel's work is preserved in Barbieri's testimony, and that only in a book published more than two hundred years after Barbieri wrote it.^[12] Barbieri quotes poetry from a book he calls the "Libro di Michele" and cites the following statements from it (Italian phrases are Barbieri's):

Maistre Miquel de la Tor de Clarmon del Vernhesi escrius aquest libre estant en Monpeslier &c.

E ne scrisse ancora delle sue in soggetto del suo amore, di cui dice in una Canzone:

En Narbone era plantatz
L'albre quem fara murir,

Et en Monpesher es cazatz
En molt bon luec se nes mentir.
                (Barbieri 1790, 120–121)

"Master Miquel de la Tor, from Clermont in Alvernhe, wrote this book
while he was in Montpellier etc." And he wrote more of his own [poetry]
on the subject of his own love, of which he says in a canso: "In Narbonne
was planted the tree which will make me die, and in Montpellier it [the
tree] fell, in many a good place and that's no lie."

Like Bernart Amoros, these men in their biographies evince the conviction that to compile and copy troubadour lyric poetry one should be able to compose it. Ferrari de Ferrara tried his hand at various genres but specialized in the sirventes (political satire)—yet his compilation is one of cansos . Miquel de la Tor, scholar though he was, adopts the ethos and idiom of the poet/lover ("que·m fara murir") with his boast that his song (the "tree," the body of knowledge he has mastered in Montpellier) has been well distributed ("es cazatz en molt bon luec"). Poets and compilers, these men's expertise in troubadour poetry came not only from reading but also from performing and listening to others perform.

For these reasons we must now challenge the accepted theory (of Gröber, Avalle, and Marshall) that all the poetry anthologized in the chansonniers descends exclusively from authorized Liederblätter (song sheets) distributed, multiplied, and modified only through repeated copying from copies .^[13] There is no proof that twelfth-century jongleurs habitually referred to written copies as they sang their renditions; indeed, there is evidence that many of these performers were illiterate (Paden 1979, 4–5; Paden 1984, 97–98). It is also, as Hendrik van der Werf points out, very likely that some copies were taken down from performances rather than from written texts. The act of transcription itself, given medieval methods of reading, may have tended to reenact performance. Van der Werf, in his study of variants in musical notation, observes that

there was no one prescribed way of performing a certain chanson, nor was there the uniformity in musical notation that we know now. Furthermore, we may conclude that the scribes did not copy at sight symbol for symbol. Instead, the differences between certain manuscripts suggest that a scribe may have sung to himself a section from the draft in front of him—not necessarily the melody of exactly one entire line—and then copied from memory what he had heard rather than what he had seen. Consequently he 

put himself in the position of a performer notating his own performance.
                (1972, 30)^[14]

Bernart Amoros implies strongly that the correctness of the text he writes owes as much to the fact that he has "seen and heard" many songs as to "so qu'ieu trobei en l'issemple" (that which I found in the exemplar). Might not a copyist, familiar with a song as song more than as tuneless poetry, also sing the words to himself as he wrote, copying what he sang rather than what he saw? Might not he sing, moreover, what he had heard in preference to what he had read? The adaptability of spelling to conform to the scribe's own pronunciation of the language would surely not detract from this point of view: one might cite, for example, the "nonidentical twin" manuscripts G and Q in their presentation of the poems of Rigaut de Berbezilh: the six songs are given in exactly the same order, with a line-for-line correspondence that suggests a case where both copies were made from the same original, yet there is almost no letter-for-letter correspondence in their spelling (Bertom 1912, 187–196; Bertoni 1905, 85–90).

Our modern literalism, influenced by the printing press, conceives of the "faithful copy" in quite a different sense from Bernart Amoros's transcription "en bona fe" (in good faith). It is true that "matters of editorial technique" need to come more to the attention of readers of troubadour poetry; it is also true that "at least in certain kinds of lyric poetry, the exact letter of the text matters a very great deal indeed" (Marshall 1975, 11). But do those "certain kinds of lyric poetry" to which Marshall refers properly include all troubadour songs, early and late, whether they forbid the jongleur to camjar lo ver or request that he improve it? This exactitude, this reverence for the "well-wrought Urn," for these "letters" and "texts," certainly is essential to our modern conception of lyric poetry. Nonetheless, we now have reason to doubt that the troubadours defined "lyric poetry" exactly as twentieth-century poets do. Marshall believes that written composition and purely written transmission allow us to reconstruct an authentic original text; underlying this belief are the assumption that such a text once existed and the hope that, like modern poets, the troubadours strove to perfect one original version and then, through written circulation, to claim it as inviolable literary property:

in so far as we can reconstruct the textual history of individual poets or songs, that history seems to be one of written texts, without interference (or with no demonstrable interference) from memorial transmission. So far 

as we can now ascertain, written copies were in circulation virtually from the moment of composition and formed ultimately the basis for the manuscript collections which have come down to us. (Marshall 1975, 14)

Since, as a matter of fact, "demonstrable interference from memorial transmission" is evident not only in the quality of variants in the chansonniers but also in the testimony of the troubadours themselves, we should reexamine this reconstructed "textual history . . . of written texts." A crucial question in evaluating the poets' own ideas of a "faithful copy" is the extent to which their own use of writing, and of other methods of promoting literalism and fixing their words, made the exact letter of the text matter to them. As we shall see, the twelfth-century poets know about writing but show little sign of using it either to compose or to fix their texts. They occasionally use "fixative' rhyme schemes, but only 10 to 20 percent of the time. They frequently speak to the addressee as to a future reciter of the song, and they occasionally ask for the kind of emendation "in good faith" that will, if necessary, "improve" their lyrics.

Images of the Writing Poet

Evidence that the troubadours used writing as an aid in composing, recording, and transmitting their songs often poses interpretive problems, for it is far rarer than the evidence for memorizing "careers." There remains the possibility that even the more formally complex poems could have been composed in the author's head and then noted down afterward. This may be difficult for us to accept, in an era when the lyric is considered a form that aspires to perfection "to the letter." Ezra Pound, a modern poet who knew the labors of composition in exacting forms of rhymed and metered verse, could not imagine Bertran de Born other than as a writing poet. He suggests that when fact runs out we "try fiction," and envisions Bertran scribbling at a table on strips of parchment, swearing as he revises, "testing his list of rhymes," scratching and erasing quarrelsome words. Pound's vision comes complete with green eyes and a "red straggling beard."^[15]

Ezra Pound's image of composition is borne out by some allusions in the works of the troubadours and undercut by others. Much of Gröber's (1877) evidence for the use of Liederblätter (individual song sheets) during composition comes from the time of Giraut Riquer, when written transmission had begun seriously to monumentalize in the permanent anthologies. The conclusion that pen and paper necessarily made it easier to 

"test one's list of rhymes" may owe its credibility to what "seems" true to the modern vision (which lacks some of the medieval acuity in audition) of lyric poetry. Pound's description of Bertran is a self-portrait, and the green eyes and red beard are not the only features borrowed from the modern poet to fill in what we do not know about the medieval one.^[16] As Gröber puts it, it seems inconceivable that paper and pen would not have "helped" in so complex an art: "The poetry and composition of the troubadours in no way seems to have been so easy that parchment and pen would not have provided a welcome aid to the fixing of their thoughts" (1877, 338).

Miniatures in the chansonniers often depict the Occitan poets writing, most often on a long scroll. The artist's conception does not distinguish between the poet composing verse on paper and the poet making a record of what he has composed. Sylvia Huot observes that in the trouvère MSS O, M, P, and W, "the scroll is an iconographic motif suggesting song as such—the lyric text, destined ultimately for oral performance. . . . The scroll as a visual image carried connotations of orality from its use as the medieval equivalent of the 'voice balloon': a figure held an unfurled scroll bearing the words that he or she was meant to be saying." The trouvère MS A, by contrast, depicts books (Huot 1987, 78–79). She also finds, in trouvère manuscripts, that "in every image of 'song making,' the song is represented by a scroll." While in one manuscript the scroll is inscribed with the opening line of the song being illustrated, in two others the figures holding the scrolls are not looking at them, and in two cases the scrolls are rolled up (p. 78). The manuscripts in Huot's study date from the end of the thirteenth century and later. Although these pictures of the scroll-using poet could have been influenced by the assumptions of the growing book culture for which they were painted, it is true that, executed so soon after the flourishing of the classical troubadour tradition, and many of them decades before its last revival, they may spring from factual knowledge of the poets' usage.^[17]

The image of the writing troubadour is also validated by several allusions in the works of the poets. Jaufre Rudel's famous lines remain ambiguous:

Senes breu de pargamina
Tramet lo vers en chantan.
         (J Rud 2/1, 29–30)

Without a letter of parchment I send the poem by singing. 

These lines could indicate either that the usual way to trametre un vers did involve a written letter^[18] or, on the contrary, that the lyric form, inscribed on the memory instead of on parchment, distinguishes itself from other kinds of permanent documents—deeds, contracts, or prose letters.

Guilhem IX complained that he had not seen any news:

De lai don plus m'es bon e bel
non vei mesager ni sagel.
        (Gm IX 10, 7–8)

From the place that most pleases me, I see neither a messenger
nor a seal.

Jeanroy, I think, is right in translating sagel as metonymous: "une lettre scellée," a sealed letter.

As late as the composition of some of the longer razos, the biographers still hesitate between depicting written exchange with literate ladies and reporting exchange via a human envoy. The razo of a poem by Guilhem de Balaun (P.-C. 208, 1) is recounted in two manuscripts, and there is a telling discrepancy between two versions of the role of the go-between, Bernart d'Anduza, as he "transmits" to Guilhem's lady the song "Lo vers mou mercejan vas vos."^[19] In H, Bernart carries a written copy of the poem: "Si·l portet lo vers escrit" (And he brought her the written poem). The equivalent action in R, however, runs to 150 words: Bernart rides to Balaun, where he interviews the poet; thence he goes to visit the lady "e comtet tota la razon de Guilhem a la dona" (and recounted the whole apology of Guilhem to the lady), whereupon he attempts to persuade her to pardon Guilhem, countering her objections. Both versions then proceed: "E la preget tan caramen . . ." (And he pleaded with her so preciously . . .). In H, "lo vers escrit" might itself be the prec, the plea; the biographer may even have considered it to be written in rimas caras (dear rhymes). In MS R, by contrast, the plea depends purely on the eloquence of Bernart d'Anduza, the go-between. The lady of R demands a song of apology from Guilhem, and the story requires that he deliver and perform it in person (Boutière and Schutz 1964, 328, no. 56). In this very vacillation we may find traces of the double transmission suggested by van der Werf—relying sometimes on parchment, sometimes on a memorizing carrier.^[20]

The question of how commonly writing was used as an aid to composition is far from settled. A poem perhaps by Bernart Marti (its authorship is disputed by Peire d'Alvernhe, Marcabru, and Bernart de Venzac)^[21 ]

mentions a peniers (Hoepffner: écritoire , writing box) in a context that suggests the joy of composition and the distraction from trouble afforded by poetry:

Non er mais drutz ni drutz no·m fenh;
Lo peniers ni jois no m'esjau.
("Belha m'es la flors d'aguilen,"
vv. 49–50, in Hoepffner 1929, 36)

I will nevermore be a lover, nor do I pretend to be one; neither
the writing box nor joy  gives me pleasure.

Since the tornada is one of the most frequent locations for commentary on the song—its success, its worth as a finished product, its value to the poet, its destination—the author is certainly using "the pen" in metonymy for poetic composition. If he were referring to clerical labors, he could scarcely expect esjauzimen (enjoyment) on a par with what joy, the source of inspiration that subsumes and surpasses all forms of gaudium, might give him (Camproux 1965). With songs occurring in few manuscripts, one must allow for the possibility that one or more tornadas have been edited out by the compilers of the chansonniers or by their performing predecessors; still, this tornada by itself gives the appearance of an anti-envoi, a defiant substitute for the increasingly usual address to a named, beloved addressee, to whom the poet expresses his modest hope that his own sense of joy measures the poem's worthiness of its recipient. The passage might be loosely rendered, "I have no one to send this to, so I do not enjoy composing; I don't even enjoy joy ."

Writing may not have been a necessary step in poetic creation, however. When Bernart de Ventadorn refers explicitly to sending a written copy of a poem for private reading, he implies that if he could have sent a messenger, he would not have bothered to write the words:

Pois messatger no·lh trametrai
Ni a me dire no·s cove
Negu cosselh de me no sai
Mais d'una re me conort be:
Ela sap letras et enten
Et agrada·m qu'eu escria
Los motz, e s'a leis plazia
Legis los al meu sauvamen.
        (B Vent 17, 49–56) 

Since I will not send her a messenger, and it is not fitting for me
to speak, I know no advice for myself. But of one thing I am
very glad: she knows letters and understands them, and I am
pleased that I might write the words, and if she likes, she may
read them for my salvation.

Bernart introduces the lady's literacy as if it were a great marvel, using a whole line just to create suspense for the news. We are not meant to be surprised that he can write, but only that her ability to read makes writing worthwhile. He implies that the written record will follow composition: the form of the verb (escria, subjunctive of escriure ) makes the act of writing potential rather than actual: it implies future possibility. There is no mention of noting down the tune (lo son ), but only the words—los motz being the standard phrase used to signify "the song I am now singing." Nevertheless one manuscript (G ) preserves a melody for the song, indicating that at some point it was "transmitted" by musical performance as well. Saber letras (to know letters), in this context, refers to literacy in the vernacular; that he speaks of writing down this very song and not a separate letter in Latin also finds confirmation in the fact that entendre nearly always has poetry as its object, whether used in the receiving sense (to comprehend) or in the creative sense (to design) (Schutz 1932). In a song whose subject is the lover's elaborate hesitation to communicate, the phrase trametre mesatge (send a message or messenger), taken together with the position of the stanza (preceding the tornada in all manuscripts that include it), spells out the motif of the envoi: "Go, my song."

Elsewhere Bernart speaks again of writing to the lady: "De l'aiga que dels olhs plor, / escriu salutz mais de cen, / que tramet a la gensor" (With the water that I weep from my eyes, I write greetings, more than a hundred, which I send to the most noble lady; 6, 49–51). The salut d'amor was at that period not an established lyric genre; Bernart calls his songs vers or chansos so this is not necessarily a case of sending off a song in writing. Nevertheless, the song is a sort of greeting. Although this is a metaphorical use of writing, "figures of speech" usually come about when the customs from which they derive are firmly enough established to be taken for granted. Writing to ladies would have to be customary in the first place before a topos about writing to them using tears for ink could become common.

Bernart treats writing, far from essential to his song, as a novelty; thus 

it becomes a clever solution to the problem of the untransmittable song first posed by Guilhem IX:

Ren per autrui non l'aus mandar
tal paor ai c'ades s'azir,
ni ieu mezei, tan tem faillir,
non l'aus m'amor fort asemblar.
        (Gm IX 9, 43–46)

I dare not send anything to her through another person, so
much I fear that she would immediately become angry, nor do
I myself dare to demonstrate my love strongly, for fear of
making a mistake.

One way or another, Guilhem's poem confesses (as its preservation attests) to its "publication," though the speaker has decided to keep his composition to himself.

The tornadas, for the most part, give the impression that songs are made to be sent off into the world and not hoarded. Giraut de Bornelh apparently views letters as the very antithesis of song; in despair, the speaker threatens to "turn to the profession of men of letters" if his poems continue to fail him in his personal life:

E no·m valran una mora
Sonet ni voltas ni lais
Ans me sui totz acordatz
Que viatz
Torn'al mester dels letratz
E·l chantar si'oblidatz.
        (Gr Bor 39, 65–70)

And they will not be worth a beet to me, melodies nor trills nor
lays, but rather I have completely resolved that I would quickly
turn to the occupation of lettered men, and that the song be
forgotten.

The sadness of the idea that "the song be forgotten" reflects an observable antipathy toward "letters," despite a thirteenth-century biographer's implication that Giraut owned a library (Boutière and Schutz 1964, 57). For Rigaut de Berbezilh, to quit singing meant annihilation ("per tos- 

temps lais mon chantar, / que de mi no·i a ren plus," I give up my singing forever, so that there is nothing left of me; R Berb 2, 14–15).

Arnaut Daniel shows signs of a very curious attitude toward writing. Here above all we might test the modern assumption that complex verse forms could not be composed without pen and ink—that is, we might test it if only the evidence were unambiguous. The vida (Provençal biography) states that Arnaut's famous style, his "maniera de trobar en caras rimas" (manner of composing in dear rhymes), is something he "took up" (pres ) only after he "abandoned letters." This might mean that he had merely given up Latin, the better to devote himself to the enrichment of his native language; the cultivation of "letters," however, was not an obstacle to trobar in the first place:

Et amparet ben letras e delectet se en trobar. Et abandonet las letras, e fet
se joglar, e pres una maniera de trobar en caras rimas.
        (Boutière and Schutz 1964, 59)

And he learned letters well and enjoyed himself by composing poetry. And
he abandoned letters, and became a jongleur, and took up a style of
composing in difficult rhymes.

However we may interpret this, Arnaut appears to have developed a reputation for giving up writing in favor of performance. The vida may represent an effort to synthesize two conflicting images of Arnaut—one of a brilliantly literate poet, the other of a poet so brilliant that he had no need of letters.

In one poem, Arnaut makes a statement that sounds very much like a defense of illiteracy, and may have seemed just that to his biographer:

Ben conosc ses art d'escriure
que es plan o que es comba.
        (Arn D 4, 41–42)

I know well without the art of writing what is flat and what is a
hill.

Since plan, in addition to its geographical sense (plain), is one of Arnaut's terms for a particular style he favors (it describes the refinement of words, their "planing" and polishing), this passage may have been understood as a boast about the poet's ability to make poetic distinctions, to triar los motz without writing them down. If this could be taken as an autobio- 

graphical statement (and I am not convinced that it can), then Arnaut styles himself as the exception to the rule, one who, like the lightning calculators of mathematics, dazzled his contemporaries by "doing it all in his head" while others labored with their pens.

The author of the famous anecdote introducing Arnaut's song "I've never had her but she has me" ("Anc yeu non l'aic") apparently wishes to give a similar impression. The entire razo creates an awesome idea of Arnaut's powers of memory: his own songs "non son leu ad aprendre" (are not easy to learn), but the song in question was made by a poet who claimed to compose "en rimas pus caras q'el" (in rhymes more precious than he [Arnaut used]; Boutière and Schutz 1964, 63). The narrator draws a humorous contrast between the difficulty experienced by the jongleur in memorizing his own song and the ease with which Arnaut "la va tota arretener, e·l son" (memorized it, including the melody).

As Arnaut's razo suggests, even the author of a song normally took considerable pains to memorize it, at least if he planned to perform it himself. Marcabru says, in what is probably an allusion to the poem-as-lawsuit (cf. R. H. Bloch 1977, 171–176), that he needs to practice his affar three times before making it public:

E dei me tres vetz doctrinar
Mon affar anz que si' auzit.
        (Mcb 8, 44–45)

And I have to teach myself my case three times before it may
be heard.

If the jongleur competing against Arnaut made a record of his song, he certainly did not intend to rely on it on the day the king would judge it. The razo affords a glimpse of this process of memorization:

Lo joglars fes son cantar leu e tost; et els non avian mas detz jorns d'espazi,
e devia·s jutgar per lo rey a cap de cinc jorns. Lo joglars demandet a·N
Arnaut si avia fag, e·N Arnautz respos que oc, passat a tres jorns; e no·n
avia pessat. E·l joglars cantava tota nueg sa canso, per so que be la saubes.
E·N Arnautz pesset co·l traysses isquern; tan que venc una nueg, e·l joglars
la cantava, e·N Arnautz la va tota arretener, e·l so.
        (Boutière and Schutz 1964, 62)

The jongleur made his song easily and quickly; and they had had only the
space of ten days, and [now] the king was to judge at the end of five days.
The jongleur asked Arnaut if he had finished, and Arnaut answered, "Yes,

three days ago"; and he had not thought of anything. And the jongleur
sang his song all night long so that he would know it well. Arnaut consid-
ered how be might mock him, until there came one night, and the jongleur 
sang it, and Arnaut memorized the whole thing, including the melody.

Of course, here the question is not one of the ordinary troubadour versus the superior one—Arnaut's challenger is depicted as very nearly his equal, someone whose song was good enough to become a permanent part of Arnaut's repertory. We are meant to be astounded by the jongleur's speed, and then still more in awe of Arnaut's. The whole ten days allowed by the king would not ordinarily have sufficed; the five days in which the jongleur composed his song seemed "easy and quick" work to the narrator; Arnaut escalates the boast by claiming that he composed his song in two days. The jongleur practiced for long hours, perhaps every night for the remaining five nights, to assure that he would not forget his own composition; Arnaut learns it perfectly (tota ) by ear in one night. If the jongleur's hare runs a mile in three minutes, Arnaut's ox runs it in thirty seconds.

I shall have more to say about this razo and its place in Gröber's theory of the autograph songbook. I believe that the razo makes a better case for oral, memorizing composition than for written composition. But before we address the question of Liederbücher, it would be well to learn as much as possible from the poets about their techniques for assuring faithful transmission without writing.

Memory and the Singer

A fascinating reference to the act of memorization occurs at the end of Peire d'Alvernhe's "Bel m'es quan la roza":

Chantador, lo vers vos fenis:
Aprendetz la comensansa.
        (P d'Alv 13, 36–37)

Singers, I end for you my song: learn the beginning!

These lines do more than affirm the moral value of the sententiae expressed in the first stanza. Whether Peire is addressing hired singers, like the two Giraut de Bornelh reportedly brought with him to assist in performance, or simply an audience of amateur singers, he appears to be asking that they practice singing the beginning of the song immediately after 

they hear the end. He wants them, in short, to "take it from the top." Set next to this passage, Bernart de Ventadorn's request that Garsio "sing and carry" his song appears less redundant than it otherwise might:

Garsio, ara·m chantat
ma chanso, e la·m portat
a mo Messager
        (B Vent 6, 61–63)

Garsio, now sing my song to me, and [then] carry it for me to
my Messenger

It is not at all surprising that a poet, before sending his song off to its destination by way of a jongleur, would like to know how it will sound when it gets there; he might even be expected to supervise the jongleur as he practiced until he got it right, as even Arnaut Daniel's talented competitor had to practice. The poet thus makes himself, for the first time, a part of the audience; he will hear his song in other voices many times afterward.

Peire d'Alvernhe's signal toward la comensansa (the beginning) may point to another key to memorization. The first stanza of a song carries with it a whole set of unspoken rules, a decorum, which the song establishes for itself and then follows. Rhyme, well known as a powerful mnemonic, makes its entrance here, and the poem's metric scheme also sets up in the exordium a pattern from which the rest, with greater or lesser determination, follows in a prescribed form. The dependence of the whole development of a song on its beginning was well known, for the troubadours readily applied the maxim "All's well that begins well" to their own work:

Ab joi mou lo vers e·l comens,
et ab joi, reman e fenis;
e sol que bona fos la fis,
bos tenh qu'er lo comensamens.
Per la bona comensansa
mi ve jois et alegransa
e per so dei la bona fi grazir,
car totz bos faihz vei lauzar al fenir.
        (B Vent 1, 1–8)

The poem begins with  joy  and begins it, and with  joy  it stays
and ends; and if only the end may be good, I hold that the be-

ginning will be good. Through the good beginning come  joy
and happiness to me, and therefore I should praise the good
end, for I see that all good actions are praised when they are
finished.

Eugen Cnyrim cites dozens of examples of this motif of continuity from "comensamens . . . al fenir" (1887, 35), and in most cases the more general meaning of the proverb can appropriately be understood as a comment on songs as well. A "recreant" from the verse form he has begun will author a defective poem:

Eu dic lo ver aissi cum dir lo solh
Qui ben comens e poissas s'en recre
Melhs li fora que non comenses re.
(Cnyrin 1887, no. 329 [Peire Vidal])

I speak the truth as I am accustomed to speak it: whoever be-
gins well and afterward becomes disloyal to it, it would be
better for him if he had begun nothing.

Theorists of the thirteenth and fourteenth centuries confirm the view that the comensamen was the determining element in the song, firmly establishing its decorum; however, they tend to play down the form and insist instead on thematic continuity:

E garda be que, en axi com començaras raho en amor, que en aquella
manera matexa la fins be e la seguesques.
        ("Doctrina de Compondre Dictatz," In Marshall 1972, 95)

And be careful that in the same way as you begin the statement about love,
in that same way you should finish it and follow it through.

E si ell comensa chanson
Deu continuar sa razon
En aysi con le comensa
Si ell no vol farfallensa.
Car may mi play e agrada
Razos ben continuada
Que mot qan alcus los entresca
Ab rimas e entrebesca.
("Doctrina d'Acort," vv. 737–
744, in Marshall 1972, 51) 

And if he [a poet] begins a song, he should continue his argu-
ment just as he began it, if he does not wish to make a mistake. 
For I am more pleased and delighted by an argument with 
good continuity than by stanzas when someone braids them 
together and interweaves them with rhymes.

The troubadours of the classic period (ca. 1170) are not especially cooperative in this rule of thematic continuity, and Bernart de Ventadorn provides Raimon Vidal with an example of "razons mal continuadas ni mai seguidas" (an argument with poor continuity and in poor sequence; Las razos de trobar, in Marshall 1972, 22).

Yet whether one emphasizes the form or the material, it is clear from all this that once one obeys the order "Aprendetz la comensansa" (Learn the beginning), the task of memorization is half done: "La meitat del fait tenc per faita / qui de be comensar se traita" (I hold that half the deed is done if a man takes care to begin well; Cnyrim 1887, no. 320). At least, as long as the cantador sings something in the same form as the first stanza, no one can tell whether he has forgotten the exact words he was taught, or whether the song is "improving" or "worsening" (si melhura/si pejura ).

The certainty that the rhyme scheme makes all the difference in the aptitude of a song for memorization can be illustrated with an extreme, almost contrary, case. Peire Vidal challenges Alfonso II of Aragon to memorize a poem with seventy-four rhymes in -atz:

Tant me platz jois e solatz
D'omes onratz, per qu'ieu fatz
Tal chanso viatz,
Bons reis, que prec qu'aprendatz.
E si·m demandatz:
"Tan soven per que chantatz?"—
"Quar es enuegz als malvatz
E gaugz a nos envezatz."
        (P Vid 4, 1–8)

I am so pleased with joy  and the conversation of landed men—
that is why I make such a fast song, good king, which I beg
you to memorize. And if you ask me, "Why do you sing so
often?"—"Because it is an annoyance to the wicked, and a
pleasure to us enviable people." 

Unless the king had a far more prodigious memory than his status as part-time troubadour would imply, Peire's request that he learn such a song might appear to be a joke. How could the king possibly remember in what order the stanzas came? Knowing well Alfonso's fondness for his songs, Peire pretends to anticipate annoyance on the king's part: "What! Are you singing again?" But the poet's reply forestalls any expression of displeasure with his tongue twister: "Only bad people dislike my songs."

Learning the comensansa may have been a great aid to memory, but sustaining the pattern it set, at least one poet admitted, could overtax the memories of those destined to learn it. Berenguer de Palazol, who comes close to belittling the musical and literary talents of those to whom he has dedicated his song, makes us appreciate by contrast the compliment Peire Vidal's poem in -atz paid to Alfonso's capacities. Berenguer takes it for granted that Count Jaufre and his wife will want to learn his song, and therefore he attempts to accommodate it to them, cutting it short before it becomes too cumbersome for them to learn:

Aissi fenira ma chanso
E no vuelh pus longa sia
que pus greu la·n apenria
mo senher, e siey companho
lo coms Jaufres, que Dieus ampar.
        (B Pal 6, 33–37)

Thus my song will end, and I do not want it to be longer, for
my lord would learn it with more difficulty, and [so would] her
companion the Count Jaufre—may God protect him.

The presence of the word greu (heavy, difficult) here—really far more frequent than clus (closed) as an opposite to leu (light, easy) in stylistic programs—may serve to remind us of the vital importance of audience, and of transmission by members of the audience, in shaping this body of poetry. The danger, all too evident in Berenguer's compliment to his sponsors, is that the poet may limit himself unreasonably, never allowing himself to surpass the amateur.

Berenguer was not proposing an unusual task to the Count Jaufre. It is fairly common, in the final lines of poems of the period, for poets to recommend that the addressee learn their songs—and they do so, for the most part, without exhibiting quite so awkwardly any fears about the re- 

cipient's power of memory. Peire Rogier makes the request explicitly in two of his eight preserved songs. In "Tan no plou ni venta," he imposes a time limit (like the ten-day limit imposed for Arnaut Daniel's contest), asking in two tornadas that the lady learn his poem before Christmas:

Peir Rogier per bona fe
tramet lo vers denant nadal
a sidons que·l fai vivre.
Clama li per gran merce
qu'aprenda·l vers denant nadal
s'ab joy de lui vol vivre.
        (P Rog 2, 64–69)

Peire Rogier in good faith sends the poem before Christmas to
his lady, who makes it live. He begs her as a great favor that
she learn the poem before Christmas if she wishes to live with
the joy  of it.

An addressee who could double as a performer, and hence as a retransmitter, would indeed "make a song live," since she would expand its renown by the number of her acquaintances, who might in turn learn the song from her.

In "Ges non puesc. . . ," Peire makes a slightly greater demand on the recipient's sense of responsibility toward the song: he asks that she first learn it and then send it on ("have it conveyed") to a second lady:

Mon Tort-n'avetz mant, s'a lieys platz
qu'aprenda lo vers, s'el es bos
e puois vol que sia trames
mon Dreit-n'avetz lai en Saves:
Dieu sal e guart lo cors de liey.
        (P Rog 6, 57–61)

To my You're-wrong I send instructions that if she likes [and] if
it is good, she might learn the song, and then I wish it to be sent
to my You're-right there in Savès: may God save and keep her.

Peire is not the only poet to cultivate the idea of transmission as a sort of relay, with the poem passed from hand to hand (or voice to voice) like the Olympic flame. Guilhem IX introduced the relay, as (at least for us) he introduced so many things: 

Fait at lo vers, no say de cui;
e trametrai lo a celui
que lo·m trametra per autrui
enves Peitau,
que·m tramezes del sieu estui
la contraclau.
        (Gm IX 4, 43–48)

I have made the song, about I don't know whom, and I will
send it to someone who will send it for me through another
person over to Poitou there, so that he can send me from his
box the counterkey.

The possibility that the addressee in Poitou (MS E, Anjou MS C ) in sending la contraclau would be sending another poem is supported by the fact that many of the senhals naming poems' recipients have been shown to designate other poets. A skilled poet, in fact, would be a likely choice for a retransmitter in the process of relay. Poems sent via a first jongleur to a recipient named Joglar (performer) or Messager (messenger) or Drogoman (interpreter) are almost certainly expected to be sung again, until the song is publicly known.

Invitations to learn songs by heart can often be detected even when they are not phrased quite so boldly as Peire Rogiers's. Rigaut de Berbezilh, famous for his shyness,^[22] softens the practical, professional request—that Miels-de-dompna memorize and perform "Lo nous mes d'abril"—by weaving it into a subtle comparison:

e li ausel son chantador,
qu'atendut an en parvensa
lo pascor.
Miels de dompna, atretal entendensa
aten de vos . . .
        (R Berb 6, 3–7)

and the birds are singers, for they have awaited, it seems, the
springtime. Miels-de-dompna, I await/expect from you a com-
parable understanding . . .

It is Rigaut's special invention, the forerunner of the Italian conceit, that he introduces the vehicle of his metaphor in an extensive, descriptive sweep and then modulates it. In some poems he gives the vehicle an un- 

expected twist, changing the proverbial to the original—like his lion, which, by its roar alone, wakes its cub from death. In the present case, he splits a single vehicle between two tenors: himself and the lady. The fact that the poet compares himself to the birds as well, using the same verb, atendre, does not interfere with the first comparison; instead, it gives the sequence of words atretal entendensa / aten de vos an uncanny compactness. The speaker waits, like the birds, for the lady to show entendensa like that of the birds. Other troubadours' frequent efforts to involve the addressee directly in the diffusion of their songs, some of which have been brought forward here, make it plausible that Rigaut might like to share the role of chantador with Miels-de-dompna. 

Three—
Song Sheets and Song Books

Arnaut, Gioglaret, and Individual Song Sheets

The troubadours' own comments have given us a picture of the first phase of transmission that differs sharply from Gustav Gröber's hypothesis that troubadour songs were first "published" in the form of Liederblätter —individual song sheets. In order to reevaluate this traditional theory of written transmission, I would like to return to the story (razo ) that recounts how Arnaut Daniel "got possession of" the song "Anc yeu non l'aic . . ." (see above, pp. 47–48) and to examine Gröber's analysis of it (1887, 337–338).

This razo has long been considered key evidence to be accounted for in any hypothesis on how the troubadours composed, learned, and performed their works. It is central to Gröber's theory of transmission, too, since for him it verifies the use of writing in composition and so substantiates his idea of autograph Liederblätter as the original components of later anthologies. Since this theory still shapes modern scholars' assumptions, both in editing troubadour "texts" and in interpreting them, his treatment of the razo is well worth examining in some depth.

Gröber concludes, in his discussion of the razo, that Arnaut and the jongleur were expected to write their poems on a carta . Weak links in his argument, however, reveal that the evidence in fact favors the opposite inference. First, Gröber asks what would be the good of locking the poets into separate rooms, if they meant to compose in their heads. More pertinent, I think, is to ask what would be the good of separate rooms if they meant to compose silently. One can conceal "little strips of parchment covered over" more easily than one can muffle a voice that tests its rhymes aloud. 

Second, Gröber claims that the song could not have been "given" to Arnaut, unless we assume that the king "gave" some tangible object, namely "a sheet of parchment" (em Pergamentblatt ). Yet the inclusion of the song in Arnaut's canon suggests that, for the author of the razo at least, Arnaut was awarded authorship of the poem and was not merely given a copy. He had no need of a parchment copy in any case, since he knew the song by heart already.

Third, and most important, we should reexamine the anonymous stanza Gröber adduces to show the troubadours' awareness of the usefulness of writing. Actually preserved in two stanzas, the song "Gioglaret, when you pass by" (P.-C. 461, 142) tends to sabotage rather than corroborate the idea of poets' reliance on written records. I cite both preserved coblas, which are found only in MS P:

Gioglaret quant passarez
Garda no moill ta cappa uerz
Qe fols fora si noi lai derz
Ceu darre un moi descle
Sen carta qen teregle
Poi scriver una tal cobla
Sun daqist moti non si dubla.

Ben es sauis e sel e serz
Qe son castel bast dinz e derz
Sqe dedinz nol prenda en grez
El sera belle el dintegle
Si qe nuils non lintegle
Qieu non prez una carobla
Terra qi dauol genz pobla.
        (Stengel 1872, 282)

Gioglaret, when you pass by, be careful not to get your green
cape wet, for you would be crazy not to elevate it in that place;
I would give a barrel [muid;  dry measure, 1,872 livres] of rye
if, on a page which in you I rule, I could write such a stanza
that not one of these rhymes is duplicated.

That man is very wise and discreet and certain who builds
his castle thick and strong so that the violent do not capture
him within, and [builds] the buttresses beautifully, and the bat-
tlements, so that no one may carve into it. For I do not think it

is worth a carob bean, a piece of land which is populated with
poor society.

The carta (page) mentioned here may have been brought into the literal from the metaphorical realm by Gröber's emendation alone. Line 5, "S'en carta q'en te regle," makes sense as it stands in the manuscript; Gröber's "S'en carta qu'eu te regle" (on the page which I rule for you ) implies that the jongleur, despite his difficulties in keeping his cape out of the mud, can carry safely with him on his travels a sheet of neatly lined parchment from which to read during his performances. I see no reason to reject the line in the manuscript, which means with its context, "if I can write, on a page which in you I mark with lines, such a cobla that not one of these rhymes will be duplicated." It makes sense as a conceit, the jongleur's memory being compared to a blank page, freshly ruled and ready for a neat "inscription." If writing alone sufficed to prevent "doubling," then why "such a cobla" (tal cobla )? It would make no difference qualis est, unless some danger that the jongleur might "double the rhymes" lay in certain kinds of composition. The monumentalist metaphor of the second stanza, then—describing the poem as a fortress—specifies a program for "tal cobla": he envisions a kind of verse that can, through inherently sturdy "construction," withstand the assaults of the common throng.

The poet's doubt of his ability to write such a cobla in the carta of the jongleur's mind may be justified by the topos of the jongleur's ineptitude as a diplomat, representing the poet: he gets his clothes dirty, abuses the host's hospitality by eating and drinking too much, and then, when the time comes for him to perform, he bungles his lines.^[1]Carta is also a measure of capacity, and the poet may be hinting at the jongleur's prodigious thirst, with all his talk of barrels and quarts (moi, Fr. muid, v. 3; carta, v. 4).

Neither the metaphorical carta of Gioglaret nor the "transfer of ownership" of a song to Arnaut Daniel provides very substantial evidence for the existence of the autograph parchment song sheets that Gröber supposed to be the original components from which derive all copies preserved in the anthologies. The next step in Gröber's theory maintains that some poets collected these Liederblätter containing their own individual poems and compiled autograph songbooks, the better to preserve their work for posterity. Again, the evidence for these individual authors' manuscripts should be reevaluated in the light of new knowledge about the transmission of troubadour poetry. 

At least one poet made a private collection of his own works. Giraut Riquer, significantly the "last of the troubadours," left a "book" of poems recorded in chronological order, noting the dates of their composition—perhaps even the explanatory notes are his own. Even in such a book, the written record bows to the sung performance: there are detailed instructions for joining the words to the tune when the poet's plan departs from normal practice.

Canson redonda & encadenada de motz e de son d'en Gr.' Riquier, facha
l'an m.cc.lxxx.ij. en abril. E·1 sos de la segonda cobla pren se el mieg de la
primeira e sec tro en la fin, pueys torna la comensamen de la primeira e
fenis en la mieija de la primeira aissi quon es senhat; pueys tota la cansos
canta se aissij: la primeira e la tersa e la quinta d'una maneira, e la quarta e
la sexta d'autra maneira. Ez  aquesta cansos es la xxiiij. (MS  C )^[2]

A song by En Giraut Riquer that is both round and enchained both in
words and in music, composed in the year 1282 in April. And the music of
the second stanza takes up in the middle of the first stanza and follows
through to the end, then returns to the beginning of the first and ends in
the middle of the first, just as it is notated; then the whole song is sung
thus: the first and the third and the fifth in one way, and the fourth and the
sixth another way. And this song is number twenty-four.

We also have evidence of poets making anthologies of the works of others—Ferrari de Ferrara, Uc de St-Circ, and Miquel de la Tor all at least tried their hand at composing original verse. But all of these writers lived in the second half of the thirteenth century, and efforts to trace "authors' manuscripts" for poets composing before 1250 have relied heavily on Gröber's theoretical model for the compilation of anthologies, for lack of more straightforward evidence.

Raimbaut d'Aurenga makes one allusion to "his book," assuring us at least that a collection of poetry like his was not inconceivable:

Mas aura ni plueja ni gel
no·m tengran plus que·l gen temps nou
s'auzes desplejar mos libres
de fagz d'amor ab digz escurs.
        (R d'Aur 10, 10–14)

But neither wind nor rain nor ice will hold me more than the
fine spring weather, if I dared unfold my book of deeds of love
in obscure language.^[3 ]

In the context of the poem, the passage serves as an announcement that Raimbaut does not intend to sing of love. To "unfold his book" would be to bring out his repertory of the topoi suited to love, the furled book holding in storage the resources of his artistic imagination. Since "digz escurs" is a stylistic term from a poetics Raimbaut sometimes supported, its presence lessens the likelihood that the poet meant us to imagine a "book of amorous deeds" in another genre—a romance, for example, or "the book that never lies," Ovid's Ars amatoria .^[4] Rather, in "desplejar mos libres / de fagz d'amor ab digz escurs " Raimbaut envisions an alternative, erotic development for this poem: its Natureingang, full of dreary weather, does not preclude a love poem. A libres would be a scroll, like the ones on which the poets wrote in the miniatures, and it is remarkable that Raimbaut conceives of the déroulement of a song as analogous to the unrolling of a scroll: the metaphor makes a striking identification of the verse itself and its parchment record. Nevertheless, Walter T. Pattison (1969), Raimbaut's modern editor, shows that even when one uses Gustav Gröber's method to analyze the sequence of Raimbaut's poetry in the chansonniers, one finds no trace of an author's collection.

Because it would be so exciting to discover traces of an "author's manuscript," more than one claim has been put forward that does not hold up. For example, François Zufferey (1987, 232–233) deflates the hypothesis offered by Martín de Riquer that chansonnier V included an authorial Liederbuch assembled by Pons de la Guardia. The Catalan manuscript contains a section copied by R. de Capellades, whose name designates a place only three kilometers from where Pons's patroness made her home at Cabrera.

Peire Vidal's "Book": Sequence by Chance or by Design?

D'Arco Silvio Avalle (1960) has concluded that for Peire Vidal's surviving works, one branch of the manuscript tradition derives from the author's own collection. To reach this theory, Avalle analyzes the relationships among manuscripts largely by the method invented by Gröber: by comparing the sequences of songs in the chansonniers, one finds series and pairs of songs grouped in similar or "scrambled" order in two or more manuscripts. Such similarities among pairs of manuscripts form networks of latent "similarities" among several manuscripts; on this basis, Gröber's method permits Avalle to speculate about the written descent of 

song collections and to reconstruct an "original" grouping of Peire's poems. Avalle then brings evidence from textual criticism, the comparison of variants, to support his reconstruction.

One problem with this method is that the groups of manuscripts found to present the poems in similar orders do not correspond to the stemmata constructed for individual poems; furthermore, the stemmata vary from poem to poem within a single poet's work.

If Avalle is right, then writing offered little safety from the transposition of stanzas in transmission: 56 percent of Peire Vidal's songs show stanzaic transposition. But the evidence for Peire's songbook is much more subtle than the evidence for Giraut Riquer's, and can be elicited only by means of very complicated editorial procedures.^[5] Traditional textual criticism presupposes a single original, which the editor must try to reconstruct as nearly as possible; in effect, the method itself assumes that an author's manuscript once existed.^[6] Since no geometry can prove its own postulates, we should examine the validity of using textual criticism to reconstruct and then attest the existence of "il libro del Vidal."

Available methods for identifying Liederbücher take too much for granted. To accept the use of Gröber's method to "reconstruct" an individual songbook, we must assume a high degree of literalism in copying. Surely a literalist principle strong enough to leave clear marks of manuscript filiation should have tended to stabilize the works of poets who compiled their own collections. Medieval scribes knew how to copy one word after the other; they generally maintained verbatim fidelity to sacred texts. If they could just as reverently preserve a sequence of vernacular poems, why not a sequence of stanzas or of lines?

The traces of an "authorized collection" discovered by Avalle in the works of Peire Vidal do not correspond to outstanding textual stability. If we assume a purely written transmission, so literal that nineteen of Peire's poems were copied mechanically enough to mark the lineage not only of individual texts but of sequences from an authorized canon as well, then how do we explain the high rate of version production in Peire's corpus?^[7] Why would the codice antico contain two poems not by Peire Vidal (P.-C. 344,4 and 70,3)?^[8] Would an "authorized" source of written descent give immunity from stanzaic transposition to the nineteen poems of the codice antico? If writing were viewed as a way to stabilize one's poems, why would Peire, seeing the opportunity to immortalize his words and desiring to do so, limit his selection to sixteen poems in the archetype? Could he, if he collected fewer than half of the surviving songs at- 

tributed to him, have intended to monumentalize one group of poems and still allow others to be transferred during his lifetime?

A count of the variations among manuscripts in their order of strophes, for the sixteen poems of the "archetipo" (i–xvi in Avalle's edition), shows that fewer than half resisted stanzaic transposition, and only one escaped both transposition and abridgment:

The "author's manuscript," if one existed, clearly failed to stabilize the order of stanzas. Still less did it guarantee that all of the stanzas of a given poem would be included in all manuscript versions (nor did it control which stanzas would be omitted in abridgments).

Several branches of Avalle's stemma must have gone back much earlier than the codice antico,^[9] and rival versions may have been known to Peire Vidal; the poet died sometime after 1204, his late travels taking him to some of the countries where manuscripts were made (i.e., Italy and Spain). Avalle's stemma postulates five written "editions" prior to MSS I and K, copied in the mid-thirteenth century.^[10] Yet there may not have been time for Peire's "authorized text" to erode gradually through the inadvertent errors of the scriptorium. We may conjecture about the time it took to prepare and execute an anthology (analogous to the modern book's time "in press") from the making of the Catalan chansonnier of Venice (V ): the original collection contains no poem composed much later than 1200, since the compiler includes nothing newer than Arnaut Daniel's works, and yet the explicit is dated May 31, 1268 (Frank 1949, 234–237). If the compiler had been interested in updating the texts of the poems he had selected, he would probably also have updated the selection itself, including more recent work. 

The five "manuscripts" predating I and K in Avalle's stemmata might easily have been people and not Liederblätter or Liedersammlungen: a jongleur with his repertory is much more mobile, more susceptible to local influence and style, freer to experiment with the interplay of words and music, than is the sedentary scribe with his quill and ink and issemple .

The system Avalle uses for tracking down "il libro del Vidal" places great faith in coincidence. When one compares even random selections of poems, the laws of probability strongly favor the occurrence of matching pairs, triplets, and longer series—particularly if series with gaps or substitutions are considered to "match."^[11] This fact tends to discount the method invented by Gustav Gröber for establishing genetic relationships among manuscripts that share such sequences: although two long, identical series of numbers will rarely be drawn by chance from a limited pool, the chances are extremely high for "scrambled series" or groups sharing several inverted pairs.

Avalle's application of textual criticism to describe the manuscript tradition of Peire Vidal's poetry and to guess its history is so cohesive and minutely detailed that it can be extremely convincing: the evidence from Gröber's method supports the evidence from Lachmann's, and vice versa. Still, even with such an impressive work of scholarship, we should be cautious. In order to demonstrate the applicability of this type of textual criticism, Avalle must apply it extensively, and along with this circular reasoning there is also great danger of selecting the evidence to fit the model.

For example, Avalle's principal reason for believing that Peire Vidal himself compiled the original book from which the archetype would derive is that the manuscripts' sequences tend to suggest a chronological order of presentation for poems in the archetype. Textual criticism had isolated several common errors which indicated that the archetype was copied from a copy "contraddistinto in alcum luoghi da un ductus fortemente individualizzato, non calligrafico" (counterdistinguished in certain places by a strongly individualized, noncalligraphic ductus; p. xxxvii). One may well ask what Avalle means by "a noncalligraphic ductus"—a written source such as a jongleur's copy, or "foul papers" of some sort, or an oral source? Avalle continues:

For this reason one may justly ask oneself if this copy were not in fact edited by the poet himself with the intention of grouping in chronological order a part of his production. How else can one explain the indisputably 

exceptional fact that a lover of Provençal poetry was capable of arranging, with such chronological exactness, such a substantial number of songs, especially when we consider that the medieval lyric anthologies were put together on the basis of criteria of a strictly formal type. (p. xxxvii)

Avalle's Table VI (p. xxxvi) shows that of the first sixteen poems in his edition, MS A presents ten in the roughly chronological order chosen by the editor—though not without intermediate poems that disrupt the chronology. If we discount the undatable poems from the sequence, A has only seven poems in this chronological order:^[12]

The main Peire Vidal collection in chansonnier A, nos. 268–292, could easily have placed certain of its poems in chronological order simply by chance. But Avalle sees strong circumstantial evidence: "Obvious is the conclusion that the results which obtain impose themselves, so to speak, in a peremptory manner: the original order in which these songs were arranged in the 'archetype' is that, well or badly preserved, of A " (pp. xxxvi–xxxvii). The conclusion that A reflects even dimly an original order is not so obvious if we consider not only the tenuous basis on which these poems are dated but also the fact that other chansonniers may give the poems in orders that, by coincidence, are also well within the bounds of establishable chronological order and yet contradict the chronology A would lead us to suppose. Compare, for example, the placement of poems viii and iv in A and in N:^[13]

We look for authenticity of all kinds in MS A because it is an "excellent" manuscript: it is old, neat, and ample, and Karl Bartsch (who first assigned the sigla ) had cause to give it primacy in his alphabetical rating. But is its chronology really more "exact" and "conspicuous" than that of N? We find coincidence where we seek it. The above table shows that N may be credited with having preserved a chronological sequence that could not have derived from the hypothetical codice antico . Since its sequence includes poems outside the sixteen singled out by Avalle as an "author's book," the chronology in N challenges two of Avalle's assumptions. First, it must make us doubt that only an author could be "capable of arranging, with such chronological exactness, such a substantial number of songs." Second, it must make us doubt that A 's superior chronology can lend support to Avalle's reconstruction of a manuscript lineage descending from an autograph copy that consisted only of poems numbered i–xvi in his edition.

The evidence for Thibaut de Champagne having compiled his own author's anthology includes the fact that his songs "appear in almost exactly the same order in nearly every manuscript" (Huot 1987, 66). This trouvère, as king of Navarre and a descendent of Guilhem IX, had both the resources and the motivation from "family pride" to preserve his poems in a book. We are a long way from this degree of stability with Peire Vidal.

The Songbook as Literary Property

When Walter T. Pattison determined that there was no trace of an autograph songbook in the manuscript sequences of Raimbaut d'Aurenga's poetry, he took Gröber's method as a point of departure. In his view, his findings support Gröber's theory of the individual songbook by helping to establish a terminus a quo for the presumed chronological trend toward writing. Pattison also suggests that Raimbaut's neglect of writing may reflect on the probable social standing and wealth, as well as on the literary self-consciousness and "carefulness," of poets who would be likely to collect their own works:

Gröber believes that the custom of preparing individual songbooks was prevalent toward the end of the period in which the troubadour art flourished, when the poets had become more self-conscious with respect to their fame. The first two generations of troubadours, especially those of noble 

rank, were much more careless about their literary reputation and the preservation of their works. The evidence from Raimbaut's works seems to prove that there was no individual songbook in his case. (1969, 232)

For several reasons, the default of evidence for Raimbaut's writing a songbook does not work constructively to show that later, poorer, or more "careful" poets did write songbooks. That prestige and wealth make people "careless," like F. Scott Fitzgerald's Daisy, is a novelistic rather than a historical principle. Nor can we safely conclude that the failure to make an autograph songbook, as evidenced by the diversity of the manuscript tradition, reflects disregard for "their literary reputation and the preservation of their works." Few poets were more self-conscious than Raimbaut, and he seems if anything to exercise closer control over the distribution of his songs than did his humbler contemporaries.^[14] Finally, as the evidence discussed in the next chapter will indicate, it is not true that the poets gradually became more successful in preserving their works.

To associate the author's proprietary interest in his creation with the manufacture of books is too modern an impulse: must Raimbaut, as a "man of property," therefore have had no need to secure "literary property"? Raimbaut's castles were almost all pawned, in any case. Most troubadours were not men of property (Köhler 1962); moreover, the persona of the poet/lover in troubadour poetry identifies itself strongly with the artisan rather than with the landlord.

And what might the troubadours' conception of literary property have been? Could the written document alone have constituted a "product" convertible into income for the literary artisan, as Gröber's analysis of Arnaut Daniel's razo suggests? We might also ask seriously to what type of landed property the troubadours might have compared "literary property," should the analogy occur to them at all: was a poem a fief or an allodium? Was it an inalienable possession or merely an honor entrusted to one's temporary care? The symbolic "territory" of authorship seems to have been as transferable as a fief, or even more so, since the troubadours witnessed a rapid expansion in the use of money, "the mobile form of property par excellence " (R. H. Bloch 1981, 956). The poets' comparison of poetry with currency becomes explicit in discussions of trobar clus (see Chapter 6).

Whether or not they made literary castles for lack of the real thing, poets of humble rank are even less likely than noblemen to have made songbooks; even having acquired the luxury of an alphabet and the lei- 

sure to apply it, poor poets would still face the high cost of parchment. Granted, among "the first two generations of troubadours," Marcabru did succeed in leaving a large body of work compared with the small poetic legacy of the Counts of Poitou and Blaye. Yet his contemporary, Cercamon, who must also have depended on his literary reputation to make a living (if we can trust the vida and speculate on his use of a professional name), seems to have been no more "careful" than titled poets: only seven poems remain to attest to any reputation at all.^[15]

Rigaut de Berbezilh had no noble title: he belonged to a family of overseers during the decline of its prestige and fortune (Lejeune 1957, 1962). If he tried to preserve his songs by making a book, then writing did him little good—six of his nine surviving songs show transposition of stanzas. Further, in the case of the song "Si co·l solilhs per sa nobla clardat . . ." (337,1), imitation shades into adaptation so subtly that one wonders whether a transfer of literary property took place. Attributed by C and a to Peire de Cols, and by f to Rigaut, the poem is so like Rigaut's work that the best argument against his authorship is the improbability of self-imitation. Várvaro speculates that perhaps a song of Rigaut's, "with a few modifications, might have been made the property of some jongleur, who might have been Peire de Cols" (1960, 244). It is conceivable that Rigaut made the song and then gave it away—perhaps not as finished verse but in its "skeletal form": a conception, a scheme, rhymed ideas and images recognizable as Rigaut's, a few perfect lines, the rest perfectible. The very idea is unsettling to modern readers accustomed to copyright and the sanctity of the poet's original text. But whether the song was given, sold, or stolen, the "deed and title" to it appear to have changed hands.

As I have shown, the theory of the writing troubadour has its firmest adherents among modern scholars. The troubadours themselves discuss composition and circulation as a matter for the memory, the intelligence, and the voice. When they speak of writing, they treat it as a curiosity or as a metaphor, not as a necessity for the transmission of their songs.

The evidence I have brought forth thus far, both from the transmitters and from the poets, depends on the interpretation of the poets' own words—texts made doubtful by the very process of transmission they describe. The poets' statements about poetry also have to be considered within their rhetorical contexts: when a poet commends his song to a jongleur, for example, there is always the possibility that said jongleur is there for fictive or conventional reasons and not because, at the historical moment of composition, the poet needed a jongleur. Bernart de Ven- 

tadorn's tornada beginning "Garsio, now sing my song for me . . ." could even, conceivably, have been the jongleur's own "signature" to his performance, rather than the poet's "original" instructions to him. The modern reader will feel the need for some way of gauging the degree to which the transmitters, singing or writing, changed songs. How much of a given song, as the text(s) come down to us, can be attributed to "the poet"? Unfortunately, performers and scribes masquerading as "the poet" can easily blur the distinction between their contributions and the "original." Even the painstaking discernment of various dialect traits in a work can be misleading: once we have discovered several contributors, which one is the poet? Is it not typical of medieval scribes (and perhaps of performers as well) to mix dialectal variants within a single work? By throwing out variants from dialects not the poet's own, we risk throwing out the poet's "original" words as preserved in another dialect in favor of "revisions" created by someone else in the poet's own dialect. Furthermore, the poets' own frequent acceptance of variants created by performers other than themselves helps to defeat, by rendering pointless within a twelfth-century context, the search for "the original."

In Part Two of this book I will approach the questions treated here from a very different perspective, to see what can be learned from the external "facts" about the way songs have actually come down to us. How many copies? Attributed to whom? In what metrical form? In what variety of stanzaic sequences? These are the physical results, the circumstantial evidence in which the mode of transmission has left its traces. The conclusions reached here—that twelfth-century troubadours did not produce "author's manuscripts"; that their songs were reproduced and circulated in performance rather than on "parchment leaves"; and that if some troubadours were more likely than others to regard their work as "literary property" or as "fixed texts," it was neither their social class nor their ability to write that made them so—can thus be tested against an objective analysis of the "end products" of transmission.