I thought I'd write a song or two, so I tried, and tried again. It seemed like a perfectly natural thing for a young composer to do. Everyone else was doing it, so why couldn't I? But nothing worked, it felt wrong, it sounded bad, awkward, self-conscious, pretentious, even ugly. Was I just incompetent or was it that my musical frame of mind was antithetical to the task. Since I couldn't accept the former conclusion, and it was too late to change my major, I trudged on assuming the latter. On the other hand I fundamentally believed, and still do, that voice, speech and music are inextricably intertwined, and, as a result, have spent much time and effort during the past twenty years experimenting with ways to use voices in musical contexts. What follows are some observations on this trek. With 20/20 hindsight, it appears that my musical world-view with respect to speech and poetry entails the following beliefs:

1) Speech itself is inherently rich in musical properties. It is not useful to make a binary distinction between speech and song, but rather is better to imagine a spectrum with each at either end and an infinite variety of mixtures and inflections in between. Furthermore, we often impose speech-like perceptions on music and musical perceptions on speech, sometimes out of necessity and frequently with great invention and creativity. In fact, many of such perceptions are necessary to carve out meaning and intention.

2) It's tempting to view poetry as intensely musical speech. Many of music's most powerful aspects: timing, rhythm, phrase, contour, form, repetition, cadence, closure, memory, counterpoint, harmony, tone color, etc., are equally central to poetry. Perhaps music is merely intensely poetic sound. Or perhaps it is poetry without the burdens of the real world. There is an implicit assumption in this view (admittedly not universal) that poetry comes to life through performance. That is, the text of a poem is a score, and the poem is brought into being through a reading.

3) Text and poetry can be utterly transformed by music. Words often become inextricably linked and fused with their musical context. Everyone has had the experience of remembering a snatch of text only because of its musical context, and being unable to recall that text without its full-blown musical setting. Another way to look at this is to speak a text which was learned in a musical setting and notice that it usually sounds naked and absurd. The implications of this for setting poetry are important--it's easy to understand why some poets resist, and some composers find it hard. Music can destroy, recreate, magnify and alter the inner music of poetry. There is, however, no implicit virtue or fault in any of these potentials.

4) The presence of a voice palpably alters the contours of a musical context. It is insufficient to assert that vocal sounds are merely another timbre in an arsenal of instrumental color. A binary switch is activated by the presence of any voice, and its presence inalterably changes the nature of the musical landscape.

What follows is a discussion of three musical projects which deal with text, voice, and poetry in different ways. The first involves using the poetry and writings of others; the second, with creating and manipulating the illusion of speech; and the third, with composing text as part of the compositional process. All of these projects use computer processing, synthesis, and accompaniment.[1]

Six Fantasies on a Poem by Thomas Campion[2], is a set of six computer-made settings of Thomas Campion's famous poem "Rose-cheekt Lawra", from his Observations on the Art of English Poesie (1602). Campion had a didactic purpose in writing this poem, namely to demonstrate that it was possible to write quantitative verse in English, as it is in ancient Latin and Greek, i.e. verse in which meter is determined by vowel quantity rather than by accentual rhythm, as is most often the case in English verse. In the poem, Lawra, Petrarcha's love, is lauded for her ideal beauty, which our earthly music can only crudely approximate [3]. What first attracted me to the poem was the extent to which it was a play on vowel sounds.

             Rose-cheekt Lawra, come,
          Sing thou smoothly with thy beawties
          Silent musick, either other
                      Sweetely gracing.
             Lovely formes do flowe
          From concent devinely framed;
          Heav'n is musick, and thy beawties 
                      Birth is heavenly.
             These dull notes we sing
          Discords neede for helps to grace them;
          Only beawty purely loving
                      Knowes no discord;
              But still mooves delight,
          Like cleare springs renu'd by flowing,
          Ever perfect, ever in them-
                      selves eternall.

It is almost as if Campion is strumming our vowel box like a guitar. My interest, however, was in setting a reading of the poem. That is, rather than choose my own pitches, contours and rhythms for the text, it was much more interesting for me to observe what a skilled reader does with the poem, (in this case, Hannah MacKay) and respond to that musically. I am effectively asserting here that the poem is a composition, and a reading of it is a performance. I therefore wanted to have access to the `music' which the poet intended to create through a performance of the poem. My music, then, would dance around, in, and about the music of the reading. ( Consequently, the success of the piece owes a great deal to the quality of that reading.) Just as Campion had didactic purpose in writing the poem, I wanted to teach the listener to hear the implicit music in speech. The parallel with Campion's comparison between heavenly and earthly music, and Lawra's beauty, and my attempt to draw distinctions between the implicit and explicit music of the performance of the poem, is an added bonus.

Each of the six movements engages an aspect of spoken English, holding the reading of the poem up to the light as if it were a crystal, letting the reflections and refractions reveal and highlight its facets. The first movement, her voice, studies the contours of the reading by exaggerating them slightly, and by having three parallel transpositions sound simultaneously. This makes the fact of spoken language more abstract by building an intervallic dimension to create a chord-like context. It is almost as if the spoken poem is being sung by a trio of virtual Andrews sisters, and this throws a short circuit across the switch we normally engage to distinguish between spoken and sung language[4]. They seem to be talking, but the chordal projections confuse our perception of it as speech. The second fantasy, her presence, combines inverted forms and exaggerated statements of the phrases of the poem, landing on sustained vowels at the ends of lines. The coda of this movement is a paraphrase of a Campion song, Tune Thy Music to Thy Hart, and is made from the collected, sustained vowels. The resonance of the reading now resides in these vowel sounds, a key aspect of the poem itself. Next, her reflection, tunes the resonance of the voice with large banks of comb filters[5]. The intention was to create the effect of a wide variety of resonant chambers which physically respond to the vocal sounds, now spreading out like the ripples on a pond. Here it is the physical aura of the speech which emerges most palpably, rather than its vowels, or contours. The fourth movement, her song, emulates the precise and deliberate mouthing of words by good jazz singers. The choral-like setting slows down the speech and lingers on articulations, stressing the physical effort required to utter the text. Next, her ritual creates a percussion orchestra from consonants, plosives, and words, thus emphasizing the noisy articulations used in speech. The movement is rather noisy and violent and leads to the final movement, her self, in which the original, unaltered reading, is played against a background made of corresponding vowel sounds, tuned approximately to the actual pitches of the reading. Almost like a set of theme and variations in reverse, the resonance of the first five fantasies now resides in the sound of the simple, unaltered reading, making the reading itself seem like a rich and complicated piece of music. The implicit qualities of the reading, which were explicated by each of the movements, are now meant to resonate in the sound of the original [6].

Word Color [7], a setting of verse 17 of Walt Whitman's Song of Myself ("These are really the thoughts of all men, in all ages and lands...") is my only other attempt to set poetry. Accompanying a relatively straightforward reading of the poem are a number of arbitrary words which are fed through chords of comb filters with very long reverberation times, making these words into sonic objects which ring and hang in space. The general strategy is to create an environment for the reading which obliquely reflects the sense of the poem. The reading itself is relatively unaltered, accompanied only by chords of comb filters which it activates. Creating a musical context around spoken text runs the risk of sounding like melodrama, which I have always felt fails because of the friction between the sense of speaking and the implication of song. The context seems to imply that the voice should be singing but is forced to speak instead. In Word Color, this problem is addressed by having the voice actually activate musical sounds. This is intended to create a sense of purposefulness in the relation between the voice and the accompanying harmonies. The speaking voice naturally lacks a sharply pitch-quantized contour (notes) but it activates sounds which have a clear sense of pitch (chords) and in doing so creates a concrete relation between the speech and the music.

Parsing words in a musical context requires that we simultaneously engage both verbal and musical minds and probably intertwine the two in the process. Furthermore, the actual physical effort required to pry apart and understand the elements of a musical context that involves a voice induces a particular kind of attention. It might be assumed that these sorts of observations are normally ex post facto for a composer, but for one involved in electronic and computer music composition this is not so. Electronic music which lives in recording rather than live performance must seek new ways to engage listeners without the benefit of the tensions of live performance.[8] When observing a performance we use a range of cognitive skills: listening to the music; watching the performers; admiring their abilities; observing how they succeed or fail; noticing their skill at interpreting a score; being alert to the innate randomness of the process, and so on. On the other hand, music which lives in recording has none of these advantages in occupying the listener's attention and, I assert, must revert to other means in order to sustain concentration. To this end, the addition of a voice and a text offers many obvious advantages. Two suites of computer pieces I've composed address these issues by adding peculiar complexity to the task of decoding speech, cashing in on the presence of voices and the effort required to understand the words, and using unusual means to short-circuit normal assumptions about speech.

Idle Chatter, just_more_idle_chatter, and Notjustmoreidlechatter [9], all use chopped-up segments of words rhythmically scattered about with some degree of randomness to create the impression of a host of people busily chattering away, and in harmony. The fragmented words are barely intelligible, and when tossed together in a complicated mix, they are totally incomprehensible. But, when confronted with this morass, listeners seem to want to figure out what is being said, and lean forward to try and understand, perhaps attempting to depend on the well-known cocktail party effect. (In the years since I wrote Idle Chatter, many people have told me what they've heard, and nobody has ever heard the same thing, including listeners in many different languages.) The point of the pieces is not to bewilder, however, but rather to provide an engaging context which is amenable to repeated listenings. The randomness and complexity of the texture creates a listening environment which encourages the ear to wander and be curious. It takes some work to follow along, even though the background harmonies are tonal and relatively simple (a conscious choice made to allow the listener to concentrate on other aspects). To help hold the texture together, as well as provide listeners with a place to rest their weary ears, I've provided long sustained vowel sounds as a constant backdrop. Speech in these pieces is ultimately an illusion, not contributing meaning, but merely creating a lively rhythmic and harmonic texture. And, the crowds of chatterers and sustained background singers seem to create the illusion of not-quite-real singers in an absurd landscape.

Smalltalk, Late August [10], and Same Scene, Nine Years Later [11] also use incomprehensible speech but from a different angle. These three pieces are all made by processing informal, pleasant conversations between a man and woman. The conversations trigger sounds which track their pitch and rhythmic contours in a highly quantized way--the musical sounds are generally diatonic and rhythmic, and reflect their shapes back onto the conversations, effectively interpreting them from a musical perspective. Very softly, beneath these triggered sounds it is possible to make out the fact of the conversation, but, generally, not the words. The result of this process is that the pieces map the feeling, sense and spirit of the conversations into music, interpreting them from a musical point of view, but masking their contents. Again, as in the `chatter' pieces, the listener is encouraged to lean forward in a vain attempt to understand the words.

In all these pieces, the fact of speech is critical. Furthermore, in being unintelligible, the speech creates the illusion of a physical human presence as the motivator of a musical context, rather then as a contributor to a music whose goal is to project and color the contents of some text. Confronted with this the listener attempts to understand the speech, is frustrated, and, as in the Campion Fantasies, consequently learns to hear, through my interpretation, the implicit music in sounds which are assumed to be explicitly unmusical.

My sense that poetry and text are rich in implicit music; that music multiplies the resonance of text to create a result which, ideally, does not consist of the expression of one by the other but rather fuses the two into one, or, in other words, that effective sung text is inseparable from its real-music context; and that poetry is composed with a rich set of musical characteristics in the first place; have all led me to the conclusion that there is no compelling reason why I should not compose my own texts as a part of the compositional process [12]. The fact that the texts I write are likely to fail as poetry is immaterial since they will never exist as such, they are inseparable from their actual musical context, and my own motivations in composing them are cast entirely in terms of their eventual and explicit musical use. Furthermore, my own musical skills will lead to text composition which reflects and aids in the creation of the musical qualities I care about. Finally, my use of text has always involved processing a spoken reading of the text, so that the implicit musical qualities of spoken language naturally come into play[13]. Since the texts I write will generally not be able to stand on their own, it is my job as composer to write music which makes them work. In this respect, it is comforting to read a description by Sir Michael Tippett of a discussion he had with T.S. Eliot, in which he was trying to persuade Eliot to write a libretto for him. According to Tippett, Eliot stated `You don't want words which already have the magic which your music should provide. You'd better learn how to do it yourself.'[14] Tippett subsequently became his own librettist.

Two pieces for which I composed my own text are Now and Then [15] and Things She Carried [16]. The latter is an hour-long piece consisting of eight movements. In both cases the text was written in collaboration with the eventual reader, Hannah Mackay.

In these pieces there are a number of significant aspects concerning the relations between voice, text and music.

First, the particular quality of a female speaker is critical to the projection of the contents of the text. In Now and Then, which consists of connecting phrases from children's stories, most often concerning aspects of time (now and then, once upon a time, etc.), the sound of a woman's voice has a more intimate connection with our experiences as children hearing stories read. In Things She Carried, the text is concerned with the stuff of some woman's life, and a female voice creates a more sympathetic backdrop. A male voice would add unwanted implications, particularly since some aspects of this woman's life involve difficult relations with men.

Second, a central idea to both pieces is to provide short, suggestive statements which will encourage the listener to build mental representations of the subjects of the pieces. For example, in Now and Then, the music provides space for the listener to fill in the details left out between the connecting time-based phrases according to his or her own personal experiences in this genre. Similarly, in Things She Carried, we hear lists of statements about aspects of this woman's life, but no linear narrative describing anything specific (with the exception of a fractured mystery story which, again, leaves much to the imagination). As a result, a basic rhythm in these pieces is one in which short statements of text are interpolated in an ongoing musical context. (These phrases, moreover, usually activate musical sounds, thus creating a purposeful relation between vocal and instrumental sounds.) The sense and function of the music, therefore, is to create a context which leaves room for the listener to build mental images. By means of harmonic, textural and procedural changes (in synthesis and processing techniques, for example), the music helps to articulate the sense of the text, and create a large-scale rhythm which moves the text along. The composition of the text itself entailed the assumption that there were not determinate time spans between utterances, and that the text was not continuous. In as much as a list is a text with little implication for timing between items--a poem generally has highly determinate inter-linear timings--a particular function of the music here is to activate a meaningful sense of timing between items in these lists, or, in other words, to make the reading of a list seem compelling.

Third, the pitch and amplitude contours of individual phrases, as read and recorded, are treated musically by choosing carefully and matching them appropriately to context. In other words, larger phrases are constructed by selecting particular readings of texts which together have an effective relative sound of consequence. (In preparation, as many as ten or fifteen readings were recorded of many lines, and the order of recording of the lines has little or nothing to do with their eventual placement in the composition.) The vocal expression thus often has the sense of consequence and logic even though the contents of the text may not.

Fourth, since the voice performs a kind of narrative function, composing the text actually involved assessing the ways in which this particular reader was able to utter the words. Subsequently, many of the tasks confronted in the compositional process involved adapting the music to the expressive modes of her voice. I have no doubt that if I had used another reader the music, and text, would have turned out to be entirely different.

Finally, both pieces use what might be called non-linear narrative. That is, there is at best a very loose logical connection between one utterance and the next and rarely is there any sort of antecedent/consequent statement. Lines are often grouped together according to content, but little attempt is made to make one line seem like the consequence of another. The cumulative sense of the text requires, however, that the listener eventually be able to piece together a set of logical relations between statements. The ability to do this is aided by the fact that the text of each movement is basically a list of items of a similar sort (the contents of her purse, things she remembered, noticed, etc.), but much more importantly, it is the magical ability of music itself to create a complex set of relations between a variety of objects separated in time that makes it interesting. The music is thus a thread which weaves in and around time, mediating the flow of the text.

These, then, are my observations on words, poetry and music, extracted from my efforts to write music using voices and texts. Perhaps someday I might write a song, but right now it feels distant and difficult. What I think I have accomplished is the creation of some music which examines the complex relations between words and music, and if there is any measure of my success it will be in the extent to which it has made these relations seem even more complicated.

1 Almost all of these pieces are available on compact disc. This article will make more sense if the reader is familiar them.

2 Six Fantasies on a Poem by Thomas Campion, (1978-79), Composers Recordings Inc., Fantasies and Tableaux CD 683, 1994.

3 I was introduced to this poem in a seminar held at Princeton in the late 70's by the poet Larry Weider, in which we examined the relations between poetry, music, poets and composers.

4 It is interesting to note that rap music, which began to emerge at the time I was writing this, has the same effect, although its methods are quite different.

5 A comb filter resonates a signal at a specified frequency. When the frequency is very low the result will be an echo, otherwise it is much like a feedback loop.

6 If the original reading had come first, the didactic feel of the piece might have been unbearable.

7 Word Color (1992), More Than Idle Chatter, Bridge Records 9050, 1994.

8 I'm begging the question of listening to recordings of live performances. This presents added complexity beyond the scope of this essay.

9 More Than Idle Chatter, Bridge Records 9050, 1994.

10 Smalltalk, New Albion Records, NA030, 1991.

11 1997, not yet released.

12 This attitude is probably close to the frame of mind of writers of songs in popular music, jazz, rock, etc.

13 My views on and uses of spoken text owe a lot, conceptually, to Laurie Anderson.

14 New York Times, 9/5/93. Eliot's well-known resistance to allowing his poems to be set undoubtedly reflects his view that his poetry has its own music. See his essay "The Music of Poetry", On Poetry and Poets, pp.17-33, N.Y., 1957.

15 Homebrew, Bridge Records 9035, 1992.

16 Things She Carried, Bridge Records 9067, 1997.There are actually two songs among the movements, sung in male falsetto. The songs are, however, wordless, and highly processed.