Introduction to the musical edition

Sam Barrett

Any musical edition that includes text is strictly speaking a double edition but a publication in which a musical edition appears alongside a textual one requires more than most that its twofold criteria be made explicit. The parameters of this edition are determined in the first place by a restriction to rhythmical verse texts (ritmi) composed before c. AD 900. Operating alongside this textual definition is a musical one: the texts selected are distinguished by the presence of musical notation in at least one manuscript. The second definition restricts the scope of this publication to a certain type of material as defined by transmission. Notated hymnals survive only from the eleventh century onwards and the vast majority of notated ritmi composed before c. AD 900 survive in non-liturgical sources copied between the ninth and eleventh centuries. What has been collected together may therefore be initially described as rhythmical verse composed before c. AD 900 circulating with notation in non-liturgical manuscripts between c. AD 900 and c. AD 11001.

While texts have been selected on internal stylistic grounds, musical selection on the basis of surviving notation might give the impression that the status of music in this edition is accidental. It was, after all, the opinion of Dag Norberg that «La poésie rythmique était en général destinée à être chantée et non à être lue»2, and if this view is accepted then the survival of musical notation for certain ritmi is an indication of not so much a distinct body of sung texts as a sub-group of a larger body of sung texts. Yet even if what is collected here is not a closed corpus of material, it is still of exceptional importance for it affords a rare glimpse into the ways in which certain texts were understood and articulated when projected in performance.

The disjunction between the dates of textual composition and the dates of the surviving musical notations requires further explanation. The reason for the later initial date of the musical notations is that neumatic notations only survive in number from the late ninth century onwards3. The later boundary of c. AD 1100 is justified by the introduction of new Latin song styles at this time, for it was towards the last quarter of the eleventh century that older Latin songs of the type represented by the ritmi gathered here began to be replaced by newer styles of song4. One implication of this new departure in terms of style and recorded repertory is that the repertory of Latin song before c. AD 1100 can in retrospect be regarded as in some respects continuous, with many texts remaining in circulation from the seventh century through to the later eleventh century5.

While the musical component of this publication follows the contours laid out above, the boundaries are inevitably blurred on occasion. Later notations in liturgical manuscripts for verses initially circulating in nonliturgical manuscripts have been included as potentially valuable witnesses to the earlier performance of non-liturgical verse. Two categories of these later notations may be distinguished. First, a handful of verses were adapted for use in the Mass at an early date6. All the relevant liturgical sources copied before c. AD 1100 have been included for these ritmi as well as occasional later medieval notations. Second, one ritmus entered the later medieval hymn repertory7. Any attempt to include all the later notations for this ritmus would result in a spiralling number of sources; only representatives of later melodic traditions that display some degree of proximity to the notations recorded in non-liturgical manuscripts have been reproduced here8. A further class of exceptions also deserves mention. At least one rhythmical hymn is transmitted with notation in a non-liturgical manuscript9, yet exceptional survivals of notated rhythmical hymns have not been included since hymns follow a different pattern of transmission from the material gathered together here. By contrast, a ritmus about the antichrist (Quique cupitis audire) was added to the Moissac hymnal, and has been included here since the text is not a hymn and is not suited for use in the standard liturgical cursus. What all these qualifications demonstrate is that the criterion of transmission in non-liturgical manuscripts, although a useful guide to identifying non-liturgical song, needs to be interpreted with latitude due to fluidity in recording practices and the capacity of a handful of texts to be adapted to serve more than one function.


The musical component of this edition is equivalent to the textual component in providing pictures, transcription and an analytical database of the surviving source material. Where the musical edition differs is in provision of commentaries and parallel transcriptions, the need for which stems from the fact that the majority of musical notations survive in forms that cannot be converted directly into modern musical notation. Verbal commentaries are necessary to elucidate salient features of the notations, while parallel transcriptions allow information transmitted by individual notations to be supplemented through comparison. By way of introduction to the musical component of the edition and database, the discussion of editorial principles below follows the subdivisions used in arranging the material.


    The commentary on the notation presents information concerning the approximate date of the notation, its type and characteristic scribal features. Since established palaeographical criteria for dating neumatic scripts remain scarce, dates have mostly been suggested on the basis of the text hand where both text and neumes appear to have been copied by the same scribe. Notations have usually been classified on the basis of regional styles, for which standard categories have been employed, although it should be noted that the term 'Lotharingian' has been preferred to 'Messine' and 'French' to 'Northern French'10.

    Scribal features noted include the complexity of the notational script, its formality of presentation and traits of disposition. Cross-references to further rhythmical verse notated by the same hand have been provided where appropriate. In cases where more than one scribe added notation to a rhythmical poem, the two notations have been described separately.

      Picture (sequenza di immagini)
    • The principles followed in presenting pictures from the manuscripts that transmit musical notations are identical to those used in the textual part of this edition.

    • Database (trascrizione alfanumerica)

      Neumatic notation transmits information about melodic profile, relative rhythmic duration and vocal production through a system of graphic signs or neumes. These neumes, which refer to the performance of one or more pitches, are customarily classified along three axes: simple signs, signs with qualification, and signs of modification. The only viable way to classify this multi-layered signification of individual neumes is to create a symbolic system with a similar structure. The criteria for the symbolic system used in the database are laid out below, but it should be noted at the outset that a translation of neumatic information into a symbol system does not exhaust the modus significandi of a neumatic script, which also transmits information through its disposition of signs.

      For the purpose of creating a database, each neume has been classified according to three criteria: melodic movement conveyed by the basic sign, melodic contraction indicated by signs with qualification, and information about vocal delivery conveyed by modification of the sign. The system for indicating melodic movement has been laid out in Table 1. In this table, the numbers on the vertical axis refer to the number of distinct pitches signalled by the neume: thus 1 for a neume indicating one note, 2 for a neume indicating two notes and so on. The letters on the horizontal axis refer to melodic direction. These letters are grouped in pairs: 'a' represents a rising melodic gesture and 'b' represents its corresponding falling gesture; 'c' represents a gesture that initially rises but falls at the end and 'd' represents the mirror image of this process (a melodic gesture that initially falls, but rises at the end). This pattern continues with odd-numbered letters of the alphabet representing melodic gestures that initially rise and adjacent even-numbered letters representing mirror images of these melodic gestures. Where the indicated melodic movement includes repetition, either an 'x' is used, for pitches that consist solely of repetition, or capital letters are used according to the principle of pairing just outlined. Any melodic movement can be systematically assigned a letter and number following this method of classification. It should be noted, however, that single-note signs are a special case in so far as they do not represent melodic gestures that in themselves rise or fall. These single-note signs, listed by name along the horizontal axis of Table 1, are presented in the order conventionally found in tables of neumes11.

      The attribution of a single number and letter to each neume, although sufficient to encode melodic movement as represented by melodic notation, cannot represent the information transmitted by signs with qualification. Two types of signs with qualification are encountered in neumatic scripts. The first is «signs with oriscus ». The precise meaning of the oriscus sign has not been determined: the sign may indicate a non-diatonic pitch, or a note that directs rhythmic weight to a succeeding note that is often lower. For the purposes of classification, these signs have been distinguished by placing a dash (') after the number but before the letter in the code used to indicate melodic movement described above. The number of dashes indicates the position of the oriscus within the melodic gesture: thus when the oriscus occurs first within a rising two-note gesture it is signified by 2'a, when it occurs second in the same gesture it is signified by 2''a.

      The second type of signs with qualification is «signs with quilisma ». The precise meaning of the quilisma sign has also provoked debate. Suggestions as to its melodic interpretation have ranged from a portamento delivery, to a melodic turn, to a simple indication of the position of the semitone step within a modal scale. It is most likely that the quilisma represents some form of light passing note. «Signs with quilisma » have been distinguished by placing a dash after both the number and letter in the standard database code. The position of the quilisma within the melodic gesture is indicated by the same means as the position of the oriscus ; that is, by the number of dash signs placed after the standard indication of melodic movement, so that for a rising three-note gesture a quilisma on the first pitch is indicated by 3a', whereas a quilisma on the second pitch is indicated by 3a''.

      A further layer of information is transmitted by neumes with modifications. These modifications transmit information relating to three different categories: rhythm, pitch and vocal delivery. It has not proved possible to encode the range of rhythmical information transmitted by the addition of significative letters, by additions to basic signs (in the form of an episema ), and by the way in which melodies are parsed into separate signs (the socalled system of 'coupure'); the rhythmical nuance transmitted by early St Gall scripts in particular is simply too rich to reduce to a simple code. The multifarious ways in which information about relative pitch is transmitted by early St Gall scripts are similarly too complex to reduce to an easily graspable formula12. With this said, modifications of basic signs relating to vocal production are signalled in the database. Most of these modifications relate to liquescence and appear in the main to indicate some form of singing through consonants that can be vocalised (such as l, m, n or r). Liquescence is commonly understood to entail an addition of a light passing note (usually lower by a tone) in performance. The question therefore arises whether a two-note gesture that is liquescent should be classified as an augmented two-note neume (two notes plus a liquescence) or a diminished three-note neume (three notes with the third note liquescent). Studies of the use of the liquescent sign in neume scripts used for chant have shown that notators were themselves unsystematic in their use of liquescent signs: the same melodic figure is sometimes indicated by a liquescent twonote figure and sometimes by a liquescent three-note figure. In view of the demand for systematic classification imposed by a database, the practice adopted here in cases of ambiguity is to classify liquescences as diminished signs. The way that liquescence is signalled in the database is through quotation marks around the basic indication of melodic movement: a single pitch that is liquescent is signalled as “2b'', indicating the addition of a second lower pitch that is lighter and altered in vocal delivery.

      A threefold system of classification has therefore been used: numbers and letters describing the basic signs, dashes qualifying simple signs, and quotation marks signalling liquescent modifications of basic signs. In principle, all neumatic signs may be systematically categorised and their codes entered into a database. In practice, signs of ambiguous meaning pose a problem for systematic classification. The virga strata , for example, may signify either two repeated notes, or a rising melodic movement equivalent to a pes. In cases where there exists no means of establishing which melodic gesture is signified by the virga strata , a distinct code has been assigned to this neume, 2V. The entries in the database are laid out by line. The ends of lines are signalled by | and the end of strophes by ||. Where notation breaks off before beginning again over later lines, a double line is also used in the database. Where neumes have been omitted over certain syllables, this is signified within the line by square brackets. In cases where neumes have not been copied over the end of a line, no entry has been made in the database.

    • The primary aim of the diplomatic transcription of the musical notations is to ease legibility. Since neumes are often minute, lightly executed and rich in calligraphic nuance, these transcriptions, which have been checked against the original manuscripts, serve as aids to reading the photographic reproductions. The diplomatic transcriptions also facilitate comparison between individual lines and strophes through the systematisation of layout: the layout of the texts is by line with strophes set apart through additional spacing.

      The criteria used for textual transcription follow those adopted elsewhere in the edition. In cases of erasure or emendation, full information is not always given in the transcription since to do so would distort the spacing of the neumes above. For definitive textual transcriptions, the reader is referred to the textual part of the edition. All the notation added to any given text is included in the transcription. Where notation breaks off and is added again to later strophes, lines numbers are added in the margin in lower case Roman numerals.

    • Verbal Commentary

      The commentary on music and verse provides two types of information. Information that may be recovered from the notation alone is presented first, which in most cases consists of a commentary upon melodic profile, organisation and complexity. Information that may be recovered from the notation when read against textual properties follows; in other words, any discernible relations between melodic features (such as gradations of pitch, rhythm and delivery) and the prosody of the text (accent patterns, caesuras, features of syntax etc.) are recorded. Where patterns of repetition have been discerned in the musical notation, letters are used to label segments, with lower-case letters used to indicate half-lines and upper-case letters used to indicate full lines.


    The parallel transcriptions for the most part comprise synoptic tables of notations for the same texts. In one case a melodic concordance for a different text has been identified (Gratuletur omnis/Alma vera). The notations are aligned in the parallel transcriptions by syllable, which leads to a certain loss of fluency in the neume script but eases comparison. In general, only notations for first strophes are copied. Notations for subsequent strophes are included in cases where notation added beyond the first strophe differs substantially from that of the opening strophe. A reading of the relation between the notations is provided in a verbal commentary as a supplementary aid to understanding the parallel transcriptions.


    Creative reconstructions of the neumatic notations proposed by a variety of scholars have been reproduced both in visual form and in the form of an audio recording.

Alphanumeric System for Neume Classification
Alphanumeric System for Neume Classification


Neumatic notation provides a graphic image of the singing voice. The fundamental unit of this notation is the neume, a unit of melodic motion that may consist of either a single note or multiple pitches joined together. The forms assumed by neumes, their manner of disposition and their mode of signification vary from script to script, from scribe to scribe, and even from notation to notation. Such variation makes generalisation hazardous, but certain general, regional and individual traits may be observed. In the most general terms, it can be said that the majority of neumatic notations copied before the eleventh century are not securely convertible into modern notation without some prior knowledge of the melodic information conveyed. This leads to the assumption that neumatic notation functioned as an 'aide-memoire', reminding the reader of melodies learned primarily through an oral tradition. A broad agreement as to the basic shapes used in the various notations to represent the same melodic profiles can also be observed, which in turn suggests that a fundamental agreement as to the way in which graphic signs might represent melodic contour underlies otherwise diverse neumatic scripts.

By way of introduction to both the basic principles of neumatic notation and the main regional families of neumatic notation found in this edition, examples of four different scripts for a single melody are aligned in Example 1. The melody given in modern notation is that proposed in the Graduale Triplex for the Maundy Thursday mandatum antiphon Dominus Jesus postquam cenavit13. As for the neume scripts, Ein 121 is a particularly refined German script copied at St Gall in the eleventh century, while La 239 is a Lotharingian script of the early tenth century. The two remaining scripts were copied in France: PaM 384 is a French script copied at St Denis in the early eleventh century and Pa 903 is an Aquitanian script of the first half of the eleventh century14.

Single Note Signs

Several distinct signs for single notes can be seen in Example 1.

The term punctum is used for a sign that indicates a single note without conveying any further melodic or rhythmic information. The punctum can be seen in all scripts except Ein 121 at the word cum. In most scripts, the punctum is shaped as a point; in the Lotharingian script, however, a sign like a reversed '7' is routinely used.

The name virga is given to a sign that indicates in most cases a note that is higher than the note that precedes or follows. This sign and its function can be clearly seen in Ein 121 and PaM 384 at (post)quam ce(navit). The basic shape of the virga is a short rising stroke. The tractulus is also used in Ein 121: examples can be seen at cum and (discipu)lis. In this script a tractulus sign is used instead of a punctum and implies a note lower than one either preceding or following. A leaning form of the tractulus, known as a gravis, is also used in the St Gall script at post(quam). This sign indicates a note lower than expected.


The pes indicates two notes, the second of which is higher than the first. The various signs used for the pes can be seen at Ie(sus). The basic shape common to the pes in all four scripts is a rising movement of the pen from a lower to higher position on the parchment.


The term clivis is given to the sign for two notes, of which the second is lower than the first. The clivis signs can be seen at (disci)pu(lis). The basic shape is a movement of the pen from a higher to lower position on the parchment; in La 239 and Ein 121 the movement is preceded by an initial stroke. Since the clivis does not appear in the incipit copied in PaM 384, its form has been reproduced in the table of the neume scripts under discussion here (see Example 2).


The torculus is a three note turn, in which the middle pitch is higher than the outer two. Examples of torculus signs can be seen at (ce)na(vit). The motion of the pen common to all scripts is a rise followed by a fall.


The term porrectus is given to the inverse motion of the torculus i.e. a three note turn, in which the middle pitch is lower than the outer two. The various regional porrectus signs can be seen at (cena)vit. In all scripts there is a fall followed by a rise, although in all scripts beside the Aquitanian this is preceded by an initial stroke.


Climacus is the name assigned to signs representing a melodic descent over three or more pitches. The differing regional signs for the climacus can be observed at (Do)mi(nus). In all scripts at least three distinct marks are made by the pen as it falls from a higher to lower position on the parchment.


The name scandicus is given to signs that indicate a melodic ascent over three or more pitches. The scandicus does not appear in Example 1, but the various forms it takes in the four different regional scripts discussed here can be seen in the neume table below (Example 2). The scandicus is composed of at least three distinct marks made by the pen as it rises from a lower to higher position on the parchment.

Signs indicating some kind of qualification are also used:


An oriscus is added to a virga at (su)is: the resultant neume is known as a virga strata . The oriscus is usually found in combination with other elements and often repeats the previous pitch. Some form of special vocal delivery seems to be implied, but its precise meaning remains obscure. The virga strata is commonly found between conjunct phrases and has been interpreted as placing rhythmic emphasis on the following note.


The term quilisma is given to a sign that is often found as the middle note in an ascending minor third as at (Domi)nus in all scripts except PaM 384. The implications of quilisma signs have been debated without any clear answer emerging as to its significance: answers have ranged from a portamento delivery, to a marking of the semitone step, to some kind of ornament. The quilisma is most often performed as a lighter and shorter note.

Significative letters

The letters l, e, i, t and s are used in the St Gall script, with the letter t also found in the Lotharingian script. The letter t stands for tenete or 'hold', thus indicating slower movement on the marked note. The remaining letters provide cautionary information about relative pitch: s for sursum and l for levate mark out a significant elevation in pitch; e for equaliter clarifies that the note concerned is at the same pitch as the previous one; i for inferius or iusum indicates a descent to a lower note than might be expected. Further letters encountered in the notations in this edition include a for altius indicating ascent to a relatively high pitch, m for mediocriter, an adverbial qualitication signifying that whatever other indication is given through letters is only 'moderately' the case, c for celeriter indicating a quick execution and stm for statim indicating that the pitches concerned are to be sung immediately or without any break.


Liquescences are not indicated in Example 1 and an attempt to introduce even some of the forms used lies beyond the scope of this introduction. The principles behind liquescent neumes are nevertheless easy to grasp. Modifications of the basic signs outlined above are introduced in conjunction with certain sets of letters in the text. These modifications are most commonly introduced when the letters l, m, n or r are found towards the end of syllables. The standard assumption is that the modified sign indicates that such letters were themselves vocalized by singing through the consonant. Diphthongs are also often marked by liquescent signs, as sometimes are cases that are harder to imagine as sung through, such as the word et. There is some debate as to how liquescent neumes should be realized in performance. The most common solution, and one that has a certain amount of support from comparative manuscript studies, is to introduce a single lower pitch as an augmentation of the basic neume on which the relevant consonant is vocalized.

Other modifications

Other modifications to the basic signs can indicate further rhythmic nuance or impart information about relative pitch. A full introduction to this practice cannot be given here, but one clear case in Example 1 may be noted. A dash has been added in Ein 121 to the standard virga form at the climacus on the second syllable. This dash or episema indicates that the pitch concerned is to be lengthened.

Besides the variety in individual neumatic forms, a further distinguishing feature of the four neumatic scripts is their disposition of signs. One aspect of this disposition is the way in which the signs are heighted in relation to each other. In all four scripts there is a degree of pitch differentiation in the relative heighting of signs. Ein 121 is the least demonstrative in indicating pitch height through sign disposition, although it should be noted that it makes the most use of distinctions between signs (virga-punctum) and significative letters to transmit information about relative heighting. The heighting of the signs over the words postquam cenavit also clearly traces the melodic contour of the phrase. La 239 makes more use of relative heighting throughout: it is most consistently applied over the words discipulis suis, whereas at the opening relative heighting is observed at (Domi)nus Iesus but the relation between the first three neumes is less precisely measured. The heighting of the signs in PaM 384 is more obviously differentiated according to pitch height, although pitches could still not be assigned from the notation alone with any confidence. While the heighting of the signs over postquam cenavit, for example, traces the overall melodic contour, the difference of a tone between post- and -quam and a minor third between -quam and ce- is not evident. The most systematic method of disposing the signs in accordance with pitch height is found in Pa 903, where a ruler could be placed across the page and the pitch height of individual neumes compared. Indeed such a process was used to align the neumes, as they are carefully regulated in their disposition in relation to a dry-point line. The so-called absolute diastematy or geometrically measured heighting used in Pa 903 allows pitches to be securely recovered without prior knowledge of the melody if the modal assignment and a first pitch are assumed.

A second aspect of the disposition of signs concerns the overall axis of the scripts. The script copied at St Gall (Ein 121) uses a gentle axis of ascent and descent as witnessed by the climacus at (Do)mi(nus), whose initial virga rises at the diagonal while the following puncta fall at a corresponding diagonal. The axis of the Lotharingian script (La 239) is more varied, but there is a basic principle of a slanting rise, as at the quilisma on the third syllable, and a vertical descent, as at the climacus on the second syllable. The French notation (PaM 384) has a characteristically erect appearance due to its near vertical rise, which is offset by a gentle descending axis. The Aquitanian script (Pa 903) has a relatively sharp rise, as at the third syllable quilisma , and vertical descent, as at the second syllable climacus. These differences in script axis are, in addition to the repertory of forms used, one of the key features in distinguishing regional neume scripts.

Neume Table
Neume Table

1 For a full account of how the notated sources were identified and a preliminary list of manuscripts, see S. Barrett, «Ritmi ad cantandum»: Some Preliminary Editorial Considerations, in Poesia dell’alto medioevo Europeo: Manoscritti, lingua e musica dei ritmi latini (ed. F. Stella), Firenze, 2000, pp. 397-419. The research presented in this article was drawn from S. Barrett, Notated Verse in Ninthand Tenth-Century Poetic Collections, unpublished PhD thesis, University of Cambridge, 2000.
2 D. Norberg, Introduction à l’étude de la versification latine médiévale, Stockholm, 1958, p. 136.
3 For a summary of the current state of research into the question of the earliest neumatic notations, see J. A. Emerson (with J. Bellingham and D. Hiley), Plainchant 2. (History to the 10th century III) The Origins of Plainchant Notation, in Grove Music Online, ed. L. Macy (accessed 26/09/11),
4 For an introduction to the new song styles of the late eleventh century onwards, see W. Arlt, Das eine Lied und die vielen Lieder, in Festschrift Rudolf Bockholdt zum 60. Geburtstag (ed. N. Dubowy and S. Meyer-Eller), Pfaffenhofen, 1990, pp. 113-27.
5 For discussion of the characteristic features of the Old Song, see S. Barrett, Stylistic Layers in Early Medieval «Ritmi ad cantandum», in Poesía Latina Medieval (Siglos V-XV): Actas del IV Congreso del «Internationales Mittellateinerkomitee», Santiago de Compostela, 12-15 de septiembre de 2002 (ed. M. C. Díaz y Díaz and J. M. Díaz de Bustamante), Firenze, 2005, pp. 525-38.
6 Three verses were adapted for use in the Mass: Gloriam Deo, Homo quidam and Tertio in flore (the latter on the basis of its appearance in Paris Bibliothèque Nationale de France latin 1121). The verse Arbor natus was also adapted for use as an introductory Alleluia trope. In all cases, earlier transmission of longer versions of the texts suggests that the liturgical versions were later adaptations.
7 The ritmus in question is Gratuletur omnis caro.
8 The identification of later medieval melodic traditions is dependent upon the material collected in B. Stäblein (ed.), Monumenta monodica medii aevi I Hymnen, Kassel and Basel, 1956.
9 The hymn Nocte surgentes is found with neumatic notation on fol. 13r of Bern Burgerbibliothek 455. The manuscript is an exceptional case since it contains not only part of a hymnal, but also both late antique and early medieval non-liturgical verse.
10 For a brief and accessible introduction to regional styles in neumatic notation and their associated terminology, see D. Hiley, Western Plainchant: A Handbook, Oxford, 1993, pp. 346-56.
11 The earliest tables of neumes date from the end of the eleventh century: for full details, see M. Huglo, Les noms de neumes et leur origine, in «Études Grégoriennes», 1 (1954), pp. 53-67. The gravis, signifying a lower single pitch, is not included in the earliest neume tables and is often of ambiguous status in early neumatic notations since it is difficult to isolate as a separate sign (as opposed to one among any different angles at which a single stroke was written). The approach taken here has been to distinguish a gravis only in scripts in which it can be consistently identified as a separate graphic sign.
12 Experiments were made with a system in which the rhythmic and melodic nuances encoded in individual neumes were translated into a symbolic system but the resulting codes proved formidably complex and unsuited to data processing. For one version of this earlier and more complex system, see S. Barrett, On Editing Neumatic Notations of Rhythmical Verse, in Poetry of the Early Medieval Europe: Manuscripts, Language and Music of the Rhythmical Latin Texts (ed. by E. D’Angelo and F. Stella), Firenze, 2003, pp. 149-69.
13 Graduale Triplex, Solesmes, 1979, p. 165.
14 The manuscripts indicated by the sigla are as follows: Ein 121, Einsiedeln, Stiftsbibliothek 121; La 239, Laon, Bibliothèque Municipale 239; PaM 384, Paris, Bibliothèque Mazarine 384; and Pa 903, Paris, Bibliothèque nationale de France latin 903.