Royal Holloway Graduate Music Research
Internet Journal written and edited by students of the Department of Music

Text, Music, and Compositional Approach in
Deux Paroles de Jacques Prévert (2000-2001)


Blank manuscript paper is a daunting sight for any composer, and a point of crippling crisis for some. As a composer, sounds flow constantly through ones mind. But this does not solve the problem of how to go about composing. What shall the first note be? How might one organise this ephemeral frenzy of half-heard chords and rhythms? For that, when all post-Romantic, romanticised notions of inspiration and genius are stripped away, is what composers do: they organise; they make decisions; they choose one note over all other possibilities, and one moment, out of all the available moments, in which to place that note. They calculate. Even the shortest piece of music represents many thousands of decisions, for the decision to prohibit is as difficult as – maybe, more difficult than – the decision to permit; and within a reduced framework, each single decision assumes greater importance.

What demon incessantly impels composers towards literature? asks Pierre Boulez. Are musicians so strongly attracted to poetry that they are unable, at some point of their evolution, to do without a text around which to crystallize their music?1 I suggest that the demon in question is this dread of infinite choice, this uncertainty as to how to select that elusive first note: how, in short, to begin. To choose a text – literary or otherwise – as the basis for a musical composition is to choose a framework, a set of boundaries, and a more manageable set of possibilities. To choose to set a literary text in a piece of music is also to determine some aspect of the sonorous qualities of that music (questions of opera in translation aside).

I began to compose my Deux Paroles de Jacques Prévert at a point of crisis.2 This is not overly to dramatise the situation, but rather to say that I was unsure of how to proceed not simply with this work but, rather, in general. Much – if not, all – of my previous music had concerned itself with texture: either in terms of polyphony as a means of generating texture, or in terms of a striated texture in which different kinds of material were placed in dynamic relationships – slow- and fast-moving material played simultaneously, for example – or in sectionalised temporal juxtaposition. The results, which I found engaging and interesting, were often complex.

The cause of this crisis, indeed, was my close encounter with Préverts texts, which I had first discovered in 1996, and had ever since wanted to set to music. Nonetheless, it was not until 2000 that I felt able to approach the texts with the serious intention of setting them; until this time, I could not find the music to express the texts. (I say express the texts purposefully, for that was my aim.) These strikingly direct texts neither provoked nor required the music of the complexity that I was interested in producing. Two further, more liberating, encounters helped me find my way out of this crisis.

The first was with the writings of Pierre Boulez, whose discussion of the relationships between text and music – drawn from his experience of composing such seminal works as Le marteau sans maître (1952-4) and Pli selon pli (1957-62) – prompted a reconsideration of my views on the purpose of setting a text to music.3 Second was a rediscovery of the music of Anton Webern which, it seemed to me, was music of the kind that Préverts texts needed, and also to some extent the compositional equivalent of Préverts creative approach. I ought to qualify this statement somewhat, for it might appear at first to miss the point: Webern is most noted, at least among composers, for his clever structural conceptions, and such sophistication seems to be entirely absent from Préverts work (and, as far as I can tell, actually is). But the similarity to my mind between the two lies in the fact that, for all the astonishing intellectual virtuosity that lies hidden in Weberns music, it does remain hidden. Analysts discover it through study; audiences (at least, I should argue, in the majority of cases) do not hear it in performance. Rather, one hears subtle, striking gestures, often separated by spans of stasis or silence. Herein lies the connection: it requires tremendous courage on the part of the composer to rely on simple means, regardless of the presence or otherwise of complex structural underpinning.

In this article, I shall explain how I set about organising the musical settings of these texts, relating this process to some of Boulezs germinal observations as I proceed. It is beyond the confines of this essay to present full-scale analyses of each movement; rather, I shall select salient points from each setting to illustrate the techniques that I have applied more broadly throughout the work. Then, by way of conclusion, I shall suggest some speculative routes that further exploration of these techniques might follow in future works.

First, I shall selectively quote the passage from Boulez that first prompted me to examine more deeply my views on the music-text amalgam. He observes:

There are several levels of language and meaning at which this transfusion of poetry into music may take place. The first that comes to mind is that of description or expression. This represents the vaguest and most diffuse kind of correspondence, the elementary stage of common perception which is still incapable of envisaging the means, properly so called, of achieving any profound contact. … Communication is in fact established by means of structure, whether it be aesthetic or grammatical. … [A]s I see it, it is only structure that can form the foundation of any valuable and lasting conjunction. … It does this by making use of the criteria common to both, such as time (rhythmic quantity and vocal technique, or prosody in the widest sense) and form (reciprocal structures in determining duration, phonetic regulation and the ordered disposition of the different formal components).4
This idea of reciprocal structures appealed greatly to me, for I had long sensed a dichotomy between my approach to instrumental music (heavily planned, serially structured, motivically organised) and my vocal music (essentially Romantic in outlook, strophically conceived, illustrative in nature). My primary aim in composing these texts, then, was to find a deeper level of musical-textual transfusion than that of my previous works, and thereby to bring my vocal music into closer alignment with the rest of my œuvre.

In dealing with each movement, I began by measuring the divisions inherent in the textual structure. In the case of LAutomne this was straightforward, as the visual structure of the single-stanza text on the page made such divisions clearly apparent. The terms of textual-musical translation here are clear: the proportional relationship between line and stanza, measured in terms of syllabic length, is rendered musically in terms of the relationship between the corresponding structural section and the entire movement, measured in terms of real time. The four lines of this text have the syllabic distribution 12, 7, 6, 6, giving the line-stanza proportion-sequence 12/31, 7/31, 6/31, and 6/31. Immediately, then, the outline of the musical structure is determined: four sections, respectively comprising 39%, 23%, 19%, and 19% of the overall duration of the movement.

It must be stated that this has no bearing at all on the distribution of the sung text within this structural pattern. Of course, it is possible to divide the duration of the movement by the number of textual syllables and attach a note of the resulting duration to each syllable of the text, thus filling the entire duration of the movement with vocal declamation; accordingly, musical structural divisions and textual line-ends would coincide. This approach, however, would be crude and unimaginative to say the very least, except perhaps in some rare circumstances. Prosodically and semantically, too, this would be spurious, for textual lines do not necessarily define grammatically self-contained units, nor do they always contain discrete, concrete ideas.

Having chosen that this work would last approximately twenty minutes in total, and bearing in mind the relative sizes of each text, I decided that this first movement would last approximately five minutes. According to my proportion-sequence, this required that the four sections last 1'57", 1'09", 0'57" and 0'57" respectively. I then decided that the vocal textual declamation would be distributed in the two middle sections only, so divided this portion of time (2'06") again by the textual proportions, giving the sub-sectional durations 0'49", 0'30", 0'23.5", 0'23.5" (though in practice these final two proportions were rounded to 0'24" and 0'23"). This sub-structural division indeed does determine to some extent the distribution of the text, in that the beginning and end of each sub-section form boundaries within which the text in question must be declaimed. Significantly, however, the structural and sub-structural divisions do not coincide. The result is a striated texture wherein the vocal textual declamation and its instrumental setting progress at independent rates or, rather, the vocal line and its instrumental setting project the dimensions of the textual structure at different magnifications (see Table 1). This raises a significant point for any composer whose structural thinking is largely sectional. Sectional boundaries need not be made explicit, nor must sectional divisions mark points of obvious (i.e. audible) change. In conventional vocal music, this has more often than not been the case: textual lines correspond to musical phrases, and textual stanzas to musical sections, whose boundaries are clearly defined in terms of audible phraseological and harmonic (i.e. tonal) completion.

This historically conventional approach is clearly connected with the communication of textual meaning, which is a problematic policy. As Boulez observes, from the point of view of simple comprehension, [a kind of reading in/with music of the poem] will never replace a reading without music, which certainly remains the best means of imparting information about a poems content.5 Music acts as a distorting lens, through which is reflected not the poem itself, but rather the composers reading of it. Rather than to attempt to obfuscate this inevitable transmutation, it seems to me that a surer way of proceeding is openly to take advantage of this situation, that is, to use the musical setting as a means of exploring ones personal reaction to the poems content. From the point of view of our compositional dilemma, of course, this further helps us to find our way towards the creation of autonomous music. Freed from having explicitly to follow the structural outlines defined by the meaning of the poem, one is able to reflect the visual structure of the text, which is usually significant, in the durational structure of the setting.

So much for overall form. Boulez further notes that the textual structure can also regulate the ordered disposition of the different formal components, that is, govern the material contained by the form. I derived the pitch-content of each movement by applying the textual proportion-sequence to the number of semitones within one octave (twelve), rounded up or down to the nearest semitone. In LAutomne, the textual proportions yield the interval-sequence 5, 3, 2, 2, (expressed in numbers of semitones), i.e., a perfect fourth, a minor third, and two whole-tones. Working upwards from A, the pitch-classes selected by this means (D, F, G, and A) were omitted from the chromatic scale, and the remaining eight pitch-classes were arranged to form a modal row (see Ex. 1). Six forms of this row – prime, directional inversion, intervallic inversion, retrograde, retrograde directional and intervallic inversions – were then transposed by the row, as in twelve-tone serial technique, in its first and eighth transpositions. Any occurrences of the four excluded pitch-classes were then removed from the resulting transpositions. We are left with a pitch-sequence that preserves the integrity of the original untransposed mode and which, ultimately, is derived from the proportion-sequence inherent in the text (see Ex. 2).

In Place du Carrousel, these generative procedures are slightly more complex, as the poem seems to move, despite its visual continuity, in discrete stanzas, indicated by the poets use of upper-case characters at the beginning of lines 7, 11, 14, and 20, and by the separation of the final nine lines into an envoi, the first and last lines of which are also capitalised (lines 28 and 36). Having divided the text, according to these criteria, into five stanzas (A-E), the same generative techniques were applied to each corresponding musical section (with the exception of E, as we shall later see). Durationally, as each section was treated as a proportion of the whole, and each line was treated as a proportion of the section, the effect is ultimately no different from simply treating each line as a proportional duration of the whole. In terms of pitch-structure, however, the effect is very different, for the numerical sequence of each stanza is different, so different pitch-classes will feature to varying degrees in each section of the work. Furthermore, as I chose for this movement an approximate duration of ten minutes – that is, roughly twice the duration of LAutomne for almost seven times as many textual syllables – it follows that the ratio of vocal to non-vocal music could be correspondingly narrower.

Having explained how I arrived at my pitch-sequences, it is now necessary to examine how they were organised into the musical work. It is at this stage of the process that what one might call modal units – that is, the mode in its various forms and transpositions – ceases to have any relevance for the compositional process. In LAutomne, for instance, the pitch-sequence is divided, again according to the textual structural proportions, between the sections of the work. In some cases, this division cannot be exact. The proportional values of the four sections (39%, 23%, 19%, and 19%) bring the appropriate number of pitches from the overall sequence, which comprises 275 pitches, to 107.25, 63.25, 52.25, and 52.25. This obviously being impossible, the final two sections have 52 and 53 notes respectively. Ex. 3, which shows the opening of LAutomne in short score in relation to its pitch-sequence, serves as an example of how I proceeded. As this figure is rather self-explanatory, it would be superfluous for me to expand further upon the distribution of pitch-classes throughout the movement. That notwithstanding, there are certain procedural features to which I should draw attention. Firstly, any pitch-class may be repeated, as may any group of pitch-classes sounded simultaneously; subsequent pitch-classes may be sounded before this repetition is terminated or has occurred. Notes of different duration may be applied to pitch-classes, with the result that some pitch-classes continue to sound after subsequent pitch-classes have expired. The order in the pitch-sequence has no bearing on the vertical distribution of pitches within a chord.

It will be apparent from Ex. 3 that the musical style of these settings is extremely gestural. Pulse as such only exists in the far background; the rhythmic character of most gestures is ametrical, transitory, elusive. It is at this level that semantic representation becomes important as a means of determining the character of a factor that I shall call gesture-types. This is not a question of hypertyposis as in madrigalism, or even the picture-painting we associate with, for example, the accompaniments of certain celebrated Schubert songs, but rather concerns the generation of autonomous musical events that reflect my perception of the emotional content of the text in question. It follows that these gesture-types occupy the musical space without respect of their relation to the semantic content of the vocal line at any specific point of occurrence.

This representation of mood is obviously subjective in quality, and generalised in its structural position. Certain textual images might, nevertheless, generate more specific musical conceits. In LAutomne, for instance, the music at bars 35-46 (Ex. 4a) is quite clearly a response to line 2 of the text – the descending pizzicato figures literally illustrate the falling leaves – while the tremolandi at bars 50-53 (Ex. 4b) reflect line 3, our love shivers. However, use of these musical devices is not restricted to these points of textual occurrence; rather, similar gestures take place before and after the declamation of the textual lines that they may be seen to illustrate. Interpretation of these devices as illustrative must, therefore, be retrospective: the musical setting seems to anticipate the text, to remove the linearity of textual declamation, to superimpose juxtaposed elements of the text.

The improvisational rhythmic quality of much of the instrumental music in this work notwithstanding, there are several points where the number-sequence governs rhythmic organisation. To take an example, the second section of Place du Carrousel (see Ex. 5) uses the number sequence to determine the duration of chords, expressed firstly in terms of crotchets, then in terms of quavers (the second passage being treated more freely than the first).

Ex. 6 shows the third section of Place du Carrousel in relation to its pitch-sequence. This section also exemplifies direct compositorial intervention in the results of schematic application. The recurring rhythmic pattern minim-crotchet-minim, which occurs in two-voice canon at a crotchets distance, is formed of the number-sequence for the first half of the textual section – that is, 4, 2, 4 – expressed in quaver-units. The clarinet and bassoon, doubled by held notes in the second violin and viola, apply the first pairs of pitches to the minim-crotchet-minim pattern, each pitch being repeated three times in total. At F, the pitches that form the oboes, flutes, and horns interjections come from a later point in the pitch-sequence. The rhythmic pattern continues in the same instrumentation, but the three pitch-classes from immediately before the oboe material are inserted into the pitch-sequence (the F sharp doubled in both clarinet and bassoon) between the fifth and sixth pairs of pitches. At bar 59, the pitch-sequence is resumed as before, with the canonic entries reversed, the next four pairs and remaining F sharp (again doubled in both instruments) filling the remaining time. Meanwhile, the horn and flute interjections are repeated, again outside the pitch-sequence. In this case, I intervened in the working of the scheme to generate a clearer articulation of local form. The wordless four-note baritone figure (bar 58) alludes to those used throughout to set the words immobile, inutile, debout, and later the phrases il était là and il attendait. At the conclusion of this section, the figure occurs again, and, as at its wordless occurrence, is preceded immediately by the flute and horn, all accompanied by the ongoing rhythmic pattern. The first time, the pitch-sequence is F sharp, C natural-B flat; the second time, C natural-B flat, F sharp, resulting in a mock imperfect and perfect cadence-pair. At G, the pitch-sequence continues uninterrupted, while the rhythmic pattern is gradually developed.

The figure with which Place du Carrousel opens (Ex. 7a), for example, recurs at A, in a slightly different guise. The statement of the motif, forte growing to fortissimo, just before the opening three lines of text (2-4), encapsulate the shock of this initial image. At A, the altered figure, again forte to fortissimo, double-takes this shock as the image is expanded (lines 4-5 of the text). This motif recurs twice at K, both times piano and with varied pitch-content. In enclosing the lines il était là and il attendait (bb. 108-9 and b. 110) with this motif, I aim to mirror the poets juxtaposition of the horror of the horses accident with the animals quiet patience. As I see it, however, the direction of this text is completely towards its envoi, which expresses passionately and contemplatively by turns the poets reaction to the scene described in lines 1-27. For this reason, the final recurrence of this motif follows the completion of the sung text immediately, at M, molto fortissimo, in the entire ensemble.

This ongoing motivic development is an important device for formal articulation. Similarly, ongoing rhythmic development features in Place du Carrousel as an important formal device. The motif just discussed is preceded by a single note, struck first on the piano and gradually picked up by other instruments. This same pitch, also on the piano, marks the beginning of both halves of the second section (B and D); when heard on bassoon, this pitch marks the beginning of the third section (E). The final statement of the pitch comes just before L, first in the piano, as a single note, but immediately picked up by the clarinet in the minim-crotchet-minim pattern, pointing to the affinity between the two gesture-types throughout, which is mainly found in the doubling of the rhythm-pattern with held notes, so that the rhythm becomes a pulse in the held note.

Likewise, rhythmic patterns are restated with varying pitch-content in LAutomne. The opening clarinet phrase (bars 2-5, Ex. 7b) is aped first by the flute (bars 17-19 and 21-22), then by the oboe (bars 66-7). One of my principal concerns as a composer lies in finding the means to articulate musical form in the absence of an acoustically structuring phenomenon, such as functional tonality. Such recurrence of pitch-classes, figures, and gesture-types seems to be one of the most effective ways by which this is achieved.

Outside its relationship to the instrumental ensemble, I have hitherto said very little about the voice. From those examples that feature the vocal line, it will be apparent that prosody, in the conventional or traditional sense, is not my prime concern. The music has now been liberated from its rôle as bearer of the textual meaning, so while I do not aim deliberately to obscure the text, its audibility and comprehension are secondary concerns to its purely musical interest. Admittedly, the setting tends to work with self-contained grammatical units; this is especially true of LAutomne. The melodic character of Place du Carrousel is more fragmentary, the grammatical relationships between words and lines being obfuscated by their separation in time. However, as Boulez notes: …any use of the voice soon leads to articulation, since vocalization alone soon becomes wearisome and makes the impression, even if it is only an unconscious one, of being a rather restricted use of the vocal mechanism.6 As the opposite characteristic, continuous articulation, could become similarly wearing, relief is occasionally provided by melismatic passages, and bocca chiusa gestures, which usually allude to one of the recurring vocal motifs, and serve almost as an aide-mémoire for the listener in more extended instrumental passages, recalling the text previously set to these notes. The vocal line is the prime means for drawing links between textual parallelisms, be they structural or semantic.

It will be obvious from Exx. 3 and 6 that certain events lie outside the structural-generative scheme outlined in this article. In LAutomne, the vocal line lies entirely outside the pitch-sequence, though three of the four lines of text are sung to four forms of the row (see Ex. 8a). Certain instrumental passages also fall outside the scheme, largely because they imitate or quote the vocal line: the flute, bars 32-3, whose final note is doubled by the strings (Ex. 8b); the strings and woodwind, at bar 57, who pick up and – in the case of the latter – decorate the E from the baritone, all outside the scheme. At some points, allusion is made from within the scheme: the oboe line in bars 59 and 61-2 follows the baritone line in bars 50-51 and 56 (Ex. 8c); the clarinet continues this allusion in 63-4, aping the baritones aussi from bars 57-8 (Ex. 8d).

In Place du Carrousel, as the generative procedures are in some respects handled more freely, there is less need of extra-schematic material. Also, as the text, which ultimately generates pitch-content, comprises a greater number of (generally shorter) lines per stanza, the overall texture is fuller, much repetition of pitches notwithstanding. The penultimate section, however, does to a certain extent summarise, and thus draw together, the structure, by alluding to passages from the voice at closer proximity, in effect telescoping the vocal line, and indeed the text itself (see Ex. 9). In formal terms, then, this provides a codetta, before the final section, representing the textual envoi, to which we must now return.

The final instrumental section of this movement is drawn from a previous work of mine, the rhapsody for violin and orchestra, …alone, surrounded by darkness and silence… (2000-2001). In that context, the material furnishes a brass-quartet chorale. Three questions must therefore be addressed: first, why I chose to use this pre-existing material, and in so doing depart radically from my approach in the rest of this work; second, how this translocation has also necessitated translation of the original material; third, what ramifications this translation has for the text-music amalgam. I abandoned my generative schemes because the tone and style of the envoi struck me very differently from the rest of the text. The sudden change to a much more rhetorical, even Romantic language seemed to demand a different approach. The richer harmonic language, and the obviously freer control of pitch-content strikes the ear as a sudden change of landscape strikes the eye, thus achieving a similar musical effect to that of the envoi itself. Departure from the schemes is not total, however, for the envoi still governs the duration of each phrase. The original order of the chorale phrases has been altered, as have their durations, although the phrases were coincidentally of appropriate proportions. Obviously, the altered instrumentation heightens definition of the dynamic contours and adds variety. This section is wordless, so the proportion of vocal to non-vocal music here is not as different from LAutomne as first seemed possible. In terms of meaning, there is less need to reflect the envois content on a representational level. The structural influence remains, but musically, the coda seems to reflect upon the text already declaimed as much as to represent the semantic level of the envoi. The music, then, fulfils a similar function in the setting to that of the envoi in the poem and thus, in some senses, supersedes the text itself. For any listener independently aware of the poetic content, this function is inherently dialectical, for the omission of the envoi significantly alters the semantic balance of the text: the first section describes the situation in a manner clearly indicative of the poets response thereto; the envoi reflects upon and generalises this particularity.

I began by presenting this technical approach to text-setting as a way into writing music, a means of starting, and a more interesting approach than merely pursuing an imitative or representational agenda. I hope that this article at least partially answers the question I am most frequently asked by non-composers, musicians and non-musicians alike – how do you even know where to start?. I hope also to have demonstrated that this technique offers much more than an easy way out of that most difficult aspect of composition, but rather provides one response to the very pressing aesthetic problem of embedding a pre-existing autonomous literary work into a new musical one. Of course, one need neither to incorporate the text into the work literally, nor indeed to use a text at all: any number-sequence serves as a starting-point. Perhaps, then, this also answers another thorny question: how to move beyond serialism but still use pitch as a means of formal articulation, without reverting to functional tonality.

Many aspects of this generative process are open to subjective variation. Why omit the pitch-classes selected by the number-sequence? Why start at, and work upwards from, an arbitrary given? Prohibiting repeated pitch-classes and repetitions of a note, or using the number-sequence to control transposition of the row, the octave of each pitch-class, or local durations, for instance, are further ways in which this technique might be adapted. There are, of course, several textual structures – visual, grammatical, metric, rhythmic, lexical – and therefore, more than one possible response to these structures. One might envisage a polyphonic or striated structure, with one level of textual structure organising pitch, another rhythm, and so on. Textual proportional ratios could be represented intervallically and rhythmically; rhythms could be transposed, using the number-sequence as a basis for multiplication-/division- or even addition-/subtraction-sequences. It is in making these strategic decisions that ones interpretation of the semantic level of the text comes into its own.

Structuring vocal music in this manner has implications for musical understanding, in both a semantic and an autonomous sense, which provoke questions of structural audibility and comprehensibility that I have only touched on very briefly here. I intend to explore these issues more fully at a later stage, when my techniques and methods of applying them have been tested further, and when I am at liberty to draw on a wider variety of examples.

The marriage of textual and musical structures requires that one find a common ground where transactions between textual and musical structures may occur freely, without concession on either side. However, when a text is not conceived simultaneously with the music, it follows that in such transactions it will always have the upper hand. In this case, though, the music is truly – to use Boulezs phrase – crystallised around the text. But the text does not govern its setting: in many ways, indeed, these movements are less bound to their texts than many traditional settings, whose texts are merely structurally-irrelevant appendages. I believe there is room for both approaches, but personally I have come to find this one the more stimulating and creatively liberating. With this approach, I can freely project my response to my chosen texts, through music that, fundamentally, they have not only inspired, but also helped to organise.7


1 Pierre Boulez, trans. Martin Cooper, Poetry – Centre and Absence – Music, in Orientations (London: Faber and Faber, 1986), p. 186. Back

2 This work was first performed on 2 March 2001 in the Picture Gallery, Royal Holloway University of London, Egham, Surrey. The performers on this occasion were the Royal Holloway New Music Ensemble, conducted by Francisco Sassetti-Corrêa, with Peter Garner-Winship (baritone). The title of the work refers to the title of the collection from which my texts – LAutomne and Place du Carrousel – are drawn: Jacques Prévert, Paroles (Paris: Éditions Gallimard, 1949). The texts are to be found on pp. 203 and 232-3 respectively. Back

3 I refer especially to the chapters Constructing an Improvisation, Pli selon pli, Sound, Word, Synthesis, and Poetry – Centre and Absence – Music in Boulez, trans. Cooper, Orientations, pp. 155-73, 174-6, 177-82, and 183-98 respectively. Back

4 Boulez, trans. Cooper, Poetry – Centre and Absence – Music, pp. 195-6, italics in original. Back

5 Boulez, trans. Cooper, Pli selon pli, p. 175, italics in original. Back

6 Boulez, trans. Cooper, Poetry – Centre and absence – music, pp. 186-7. Back

7 An early version of this article was delivered as a paper at the Postgraduate Study Day in Music, Royal Holloway, University of London, 27 June 2001. My thanks to all those present on that occasion for some useful feedback and a fruitful discussion. Back

© Matthew Mills, 2001-2Back to contents page