Chapter 1: CONNECTING LANGUAGE
AND MUSIC
Linguistic
processes in musical exploration
Aspects of
linguistics in musicology and ethnomusicology
Aspects of
linguistics as tools in musical analysis
Aspects of language
as tools in computer music composition
Connections between
music and poetry
Chapter 1
CONNECTING
LANGUAGE AND MUSIC
In the 1980's, uses of language or linguistic
techniques in the analysis or composition of music were explored in some depth,
especially in computer based composition of music. These explorations outline
some methods for using aspects of language in musical composition. Focussing
less on the possible relationships between the surface levels of music and
language, these methods mostly clustered around the larger scale structural
aspects of language and how to use them in composing music. Because they are
based on what is actually heard, the surface relationships of music and
language are most fertile for music composition.
In the 1990's linguistic and language-based
approaches have been less used and explored, or at least less documented, by
computer/electronic music composers. This does not mean that language based
approaches are less fertile or valid, just that the interests of these
composers have shifted. Areas that now seem to have the interest of
computer/electronic composers are based in science and mathematics, especially
in the computer modelling of natural events, such as the flow of water,
cellular automata, or the behaviour of flocks of birds.
Linguistic processes in musical
exploration
In 1932 musicologist Gustav Becking drew the
relevance of phonology to music. He presented the paper Der musikalische Bau
des Montenegrischen Volksepos where he pointed out a similar basic problem
in phonology and musicology, that is: that musical construction and phonemic
construction are culturally dependent[1]. This was
probably the first contribution the relatively new science of linguistics made
to musicology.
Linguist Roman Jakobson
took up
Becking's point stressing that the particular property of music, as of poetry,
is that its conventions are wholly phonological in operation and do not concern
etymology or vocabulary. He urged music analysts to study the model of
phonology[2].
Phonology, being the study of the sound systems of
language, is an important area when discussing the similarities between music
and language. If a musical composition is viewed as the organisation of a sound
system then tools used in exploring the sound system of a language become
appropriate. This use of linguistic tools is particularly appropriate for
analysing highly stylised compositions such as those of the mid-century serial
composers.
Work relating linguistics and
music fits into two broad categories: theoretical and practical. Below I give a
brief background of some discussions relating linguistic theory and music. I
then discuss in some of the writings more relevant to my explorations in
greater detail, under the headings: Aspects of language in musicology and
ethnomusicology; Aspects of language as tool in musical analysis; and Aspects
of language as tools in computer music composition.
Theoretical approaches
Since Chomsky's Syntactic Structures[3] and Aspects
of the Theory of Language[4] there have
been many attempts to use his systems for divining the root structures of
language in the analysis and composition of music. Examples of these approaches
are seen in the works of musicians, theoreticians and composers such as Leonard
Bernstein, who suggests that "there are similar cognate functions
operating in both music and
language which are
discoverable by linguistic method"[5]; Lerdahl
and Jackendoff, whose "theory of musical syntax provides hypotheses about musical universals"[6].
The studies of morphology and semiotics supply
linguistic tools for exploring larger areas of language than phonology. Here the
rules for ordering phones into syllables; syllables into words; and words into
sentences and so on are explored. These tools can also be used to explore the
unfolding of larger areas of music, such as notes into motifs; motifs into
phrases; and phrases into sections, as explored by Roads[7] and Cope[8].
The efforts of the first three theorists and
composers are to create or suggest a general analytic process that can be used
effectively for analysing tonal or non-tonal music. Cope, on the other hand, is
more interested in using the same or similar linguistic concepts in the
composition of music.
Semiotics is another area where the linguistic
process of investigation has been appropriated for musical research. Nattiez[9], Lidov[10], Smoliar[11] and
Monelle[12] use this
approach to divine and examine the unique laws that link musical events within
a specific piece. Ethnomusicologists often look to the phonemic and semiotic
structures of a culture's language when examining the evolution of its music
and language. This can be seen in the work of Levman[13], Chenoweth[14], Bright[15] and Powers[16].
Psychologists have used music to examine learning,
human perception, recognition and parsing regarding memory and the perception
of time. Music offers insights into these human mental processes that are not
available when text is the only medium for testing. The works of Deutsch[17], Trainor
and Trehub[18], Clynes[19] and
Tsunoda[20] are
examples of this.
As the perception, recognition and parsing of
information is important in artificial intelligence there is interest in
codifying the perception of music and language. Programs based on linguistic
premises, such as that of Terry Winograd[21], have been
used to analyse tonal music. Marvin Minsky has suggested that the sonata form
is an excellent process for introducing and teaching new information[22].
Practical
approaches
Aspects of speech are used to create both background
and foreground structures in music. Steve Reich's Different Trains[23] and Paul
Lansky's Smalltalk[24] are
examples of speech being used to create musical surfaces. In Different
Trains Reich uses the rhythm and melodic inflections of the recorded speech
fragments to provide the rhythmic and melodic source for a string quartet,
these fragments are heard throughout the composition. In Smalltalk Lansky filters a recording of a conversation
in such a way that the rhythm and melodic inflections of the conversation are
enhanced while the denotative nature of the language is obscured. In each of
these two cases the composer's approach is based not on linguistic theories but
on a direct appropriation of speech to create music. This approach has a
precedent in the approach of Guido of Arezzo.
During the eleventh century Guido
mapped vowel sounds to pitches to create liturgical melodies. His process was to
match the distribution of vowel sounds to the distribution of pitches when
composing plainsong. In this century the Young Farmers Claim Future, a group
centred in Belgium, translate textual information into musical events using
Opcode'sª MAX program. Their process is to "match distribution of text
characters to distribution of pitches in classical tonal systems"[25].
While there is a substantial amount
of musical composition based on linguistic concepts textual composition based
on or influenced by musical concepts is comparatively rare. One possible
approach is that taken by John Cage in composing his Sixty-two Mesostics re
Merce Cunningham[26]. By using
a letter row, rather than a tone row, and applying an extremely rigorous
approach based in those of serial composition, Cage's mesostics show a direct
use of musical techniques to construct text. The result is words and word
fragments that produce a language that flips in and out of English. When seen
on the page the result is akin to the work of concrŽte poets in its visual
representation. The computer program MESOLIST[27], used when
composing the sixty-two mesostics, changes the fonts and font sizes of each
letter based on I Ching principles. These changes can be used to represent
possible pitch and amplitude changes in performance. By using these methods
Cage accentuates the acoustic, not semantic, properties of the text and the
result resembles a hybrid between sound poetry and vocal music.
Another approach linking the sound
of language and the sound of music is that offered by Bettina Knapp[28], who
claims that phonemes are emotionally loaded. She uses this theory when
discussing James Joyce's use of alliteration in his short story Eveline[29]. Knapp
suggests that Joyce constructs his phrases musically, ensuring that the
strategic rhythmic placement of particular phonemes reinforce the emotions he
is trying to convey.
The explorations listed above show
some of the varied relationships that have been drawn between music and text.
When used cooperatively the two can shed light on the culture and psychology of
their practitioners, particularly offering insights into their subset, music
theory. Practical cooperation between music and text can lead to expanded
artistic horizons, as seen in some of the textual works of Cage.
Aspects of linguistics in
musicology and ethnomusicology
Language holds a fascination for musicologists and
ethno-musicologists. Observing the similarities in a culture's music and its
language and then contrasting them with those of other cultures is one method
used to define the uniqueness of that culture.
An example of this is given by William Bright when
describing aspects of southern Indian instrumental music as:
largely
based on vocal music [, therefore] we find patterns of speech reflected in [southern
Indian] music of the flute or the violin
[and that] Indian poetry is normally recited in traditional melodic
patterns, rather than simply spoken[30].
Bright also sees that poetic metres are correlated with
linguistic patterns in European cultures. He suggests that stress in English
poetry is more important than the length of syllables, and that metres are
based on the stress patterns of syllables. French poetry, on the other hand,
has little contrast in syllable stress or length but has verse based on
syllable count[31]. The
different approaches to sound and rhythm in English and French poetry may be
seen to have a correlation to the differences in musical style of the two
cultures. This difference is more obvious in music prior to current times,
where the ease and speed of communication has blurred cultural boundaries.
Harold Powers gives another example of the
similarities between music and language when comparing written western musical
discourse to Indian classical music and Middle Eastern and South East Asian
improvisations[32]. According
to Powers written western music has very little scope for the improvisation
that is a constant and integral part of verbal discourse. However, it can be
said that verbal discourse follows strict rules, therefore we are not really
inventing novel sentences when we speak but are making improvisations within
set structures or making variations on themes. This process of improvisation is
seen in classical Indian music and in the improvisations of the Middle East and
South East Asia[33].
In these cultures improvisations are guided by models
and therefore have more parallels to language than improvisations which
elaborate or are based on fixed items, such as the more Western styles of
improvisation using figured bass notation, a chord structure, or the cadenza.
Bryan Levman draws the most obvious similarities
between music and speech. In his discussion of the origins of music and
language, he states that the differences between the performance of music and
language are of degree, not kind. Pitch, dynamics, duration and speed of
articulation, timbre and articulation are all used in speech and music. In
music the gamut of each of these areas is much wider than in speech[34].
Levman sees three trains of thought in the field of
music and language development:
1. that music and language developed over separate
paths and are of two distinct faculties, this is taken by glosso-geneticists,
who specialise in language origin and evolution;
2. that music evolved out of language, ie. song is
exaggerated speech; and
3. that music and language developed from the same
inherent 'proto-faculty': as language became more expressive of ideas than
feelings, accent decreased but consonantal articulation increased[35].
Music and
singing are also used in rituals, and as ways of controlling the environment[36]. In these
cases the lexical significance of the words is of less importance than the
music. An example of this is the use of melisma in liturgical text, where the
actual word may be obscured but its meaning heightened.
Aspects of linguistics as tools in
musical analysis
Musical analysis can fit into two broad categories:
semantic analysis, dealing with the meaning of the composition; and structural
analysis, dealing with how the composition is put together. These two areas are
distinct but not exclusive, borrowing from and informing each other constantly.
Four approaches are given below: Robert Hatten's[37] semiotic
approach, Raymond Monelle's structural semantic approach[38], John
Sloboda's[39] reductive
semantic/grammatic approach and Lerdahl and Jackendoff's syntactic approach[40].
Hatten argues that
competencies
in semiosis embrace meanings besides the syntactic or formal[41]
and that
a semiotic
perspective can take us beyond the perception of aural gestalts or processes to
the cognition of a stylistic signification. [Therefore] music cognition can
then be defined as the study of the stylistic competency as it guides not only
identification but also interpretation of [the] significance of musical events.[42]
Therefore a semiotic
understanding of a piece will include a syntactic or formal understanding.
Monelle puts forward a number of linguistic terms
taken from A. J. Greimas's SŽmantique structural in an analysis of
Bach's Fugue in B flat minor, from Book I of Das Wohltemperierte Clavier.
As the piece unfolds, each note, phrase, harmony and so on is placed in a
context and therefore given a role, or semantic meaning. This results in, for
example, the F A C triad having the role of the dominant chord and therefore
creating a need for resolution in the listener, rather than a sense of repose.
This context based understanding of the F A C triad is like the use of the word
'court' in Kafka's The Trial, where we understand it in a legal
sense, not to mean 'royal entourage'. Monelle concludes that:
instrumental music lends itself to
description in terms of structural semantics. The workings of semic nuclei,
classemes, sememes, and isotopies are discerned. The distinction of sense,
established by redundancy, and signification, by variancy, illuminates the
musical features like variation, development and reprise[43].
Monelle also suggests that for the piece to have
meaning broad and detailed levels of understanding are required by the
listener, ranging from an historic and stylistic knowledge to knowledge of the
surface interactions within it. He suggests that at the surface level music
relies on thematic and motivic repetition. This sort of repetition also occurs
at the semantic level of language, where a writer may repeat an idea using
synonyms, thus reducing the sense of repetition. A composer may use repetition
to reinforce an idea and as a structural device[44].
At a deeper level John Sloboda contrasts Monelle's
view of the semantic similarities between music and language, putting forward
that while Chomsky's deep structure and Schenker's Ursatz share similar
ideas there are three significant contrasts between their views:
1. Schenker's Ursatz does produce a melody but Chomsky's deep structure does not
produce a sentence;
2. Schenker's Ursatz can be applied to long
and complex pieces but Chomsky's deep structure can be applied only to
sentences, and often results in simply a condensation of sentences; and
3. The meaning of music is dependent on its surface
structure, in fact there are very few different types of Ursatz, but the
meaning of a sentence lies almost completely in its deep structure, of which
there is an almost infinite amount[45].
Sloboda goes on to say:
One suggestion is that the mental
substrate of music is something like that which underlies certain types of
story. In these stories a starting position of equilibrium or rest is
specified. Then some disturbance is introduced to the situation, producing
various problems and tensions which must be resolved. The story ends with a
return to the equilibrium. The underlying representation of music could be seen
as a highly abstracted blueprint of such stories, retaining only the features
they all have in common[46].
That is, the story or
musical composition begins at a period of repose, moves to a period of
agitation, and ends in a period of deeper repose; Schenker's I V I harmonic
structure in tonal music.
In
A Generative Theory of Tonal Music composer Fred Lerdahl and linguist
Ray Jackendoff put forward a theory whose "central purpose was to
elucidate the organisation which the listener imposes mentally on the physical
signals of tonal music"[47]. The
theory of musical grammar they propose provides a hypothesis about musical
universals which are appropriate regardless of style and of which certain
aspects "constitute universals of musical perception and can be taken to
represent innate aspects of musical cognition"[48].
According
to Harold Powers,
Preliminary investigations have
indicated that [Lerdahl and Jackendoff's] theory can be modified to produce
structural descriptions of pieces in styles as diverse as Macedonian folk
music, North Indian music, and 14-century French music [ though] a limited amount
of the world's music, notably a good deal of 20th-century art music, would not
appear to meet all four conditions[49].
These conditions, or as Lerdahl and Chabot put it,
hierarchical components, are:
1. grouping structure;
2. metrical structure;
3. time span reduction; and
4. prolongation reduction[50].
Lerdahl
and Jackendoff's concern with process rather than product, their use of
well-formedness rules, and their use of reduction diagrams show the Chomsky
influence in their approach.
Aspects of language as tools in
computer music composition
David Cope devised Experiments in Music Intelligence
(EMI) to assist and provoke him in musical composition that fits his personal
style and goals. This program assumes "that musical intervals can be
defined linguistically and that style emanates from the manner in which
composers catagorize (sic) these definitions"[51]. To ensure
that the EMI parsing mechanism would be linguistically based EMI was
initially
conceived and written to construct lexically correct haiku. This
followed the principles that they:
(1) Have a
correct number of syllables (1 2 2; 1 [2 or 3] [3 or 2] 1; [2 or 3] [3 or 2]
for a total of 17 with lines 5/7/5 syllables)
(2) Be
linguistically valid
(3) Be
contextually meaningful
The
following poem provides an example:
A seagull flying
A nighttime
talisman for
Caressing startrails.
These concepts were then translated, through
surprisingly few adaptive alterations, to the musical parsing functions [used
in Cradle Falling][52].
Cope used a hierarchical system of identifiers,
giving four types of musical events: Antecedent, Consequent, Ornament and
Statement (ACOS). These areas each exist in three levels: foreground,
middle-ground and background, the constraints on possible actions within the three
areas become more rigorous as the levels get deeper[53].The
composer sets the rules, weighting values and the style dictionary by which
these areas interact.
EMI follows
the general principles of sentence parsing [using augmented transition networks
(ATN)] In the case of natural language,
prepositions require subsequent noun phrases and verbs are preceded by nouns or
noun phrases and followed by nouns or noun phrases. Flexibility, however,
abounds in this linguistic approach. Noun phrases can be staccato two- or
three-word structures or extended sections with furthest sub phrases. The same
is functionally true in EMI with the design of large scale works of music.
Phrases vary in length content and meaning, but style dictates that rule apply
and that certain events inevitably require certain reactions[54].
Cradle Falling, for soprano and chamber
ensemble, was composed by Cope using EMI, which provided about one-third of the
piece. Prior to the composition Cope created an extensive dictionary of his
style, thereby ensuring that the computer would produce something that he would
see as musically valid and that would fit his body of works. He has also
created dictionaries which follow the rules and styles of other composers,
including Bach, Beethoven and Bartok[55]. Figure
1.1 is an example which
demonstrates
the line-producing characteristics of the EMI system with an attached
Cope-style dictionary. The letters and
numbers indicate the entrances of other instruments with the same material at
different tempi, creating a heterophonic texture[56].
Figure 1.1.
Heterophonic texture created with Cope's EMI system.
When composing Stelling VI. Stephen Dydo
developed the program 'LIMID', which creates music, and text related to that
music, from a set of user input words and morphemes. He sees the relations
between music and language as co-incidental rather than confluent, for example,
seeing the intonation curves of speech as creating a 'speech melody'. To
illustrate this point, Dydo refers to the
classical song of the Sung
dynasty, where the tonal features of the language were used as compositional
determinants. [That is] the melody was created from the pitch contours of the
words[57].
Thus the classical song of the Sung dynasty is a historical
precedent to text generating melody and to Dydo's approach.
Dydo took a set of pitch curves typical of American speech,
each containing three pitch levels and of five lengths, from short exclamations
to long and complex phrases. From this data he formed a
library of
timbre definitions vocal in quality
[and] imitative of some typical American speech types. However, the extent to
which they resembled actual speech was sharply curtailed by a number of
factors:
[1] only two formants could be
specified;
[2] the projection of the
pitches was regarded as being of overriding importance, so that a kind of
singing, rather than speaking was imitated;
[3] most of the sounds had to
function both in isolation and in conjunction with other sounds, so the natural
permutations caused by sounds being adjacent to other sounds was not taken into
account; and
[4] the sounds had to function
over relatively huge registeral and durational domains, necessitating
compromises with the concept of truly human sounds[58].
When linking the pitch curves he found that the
phrases "formed a sentence which would typically be inflected with that pitch
curve when uttered by an American."[59] Figure
1.2a shows a melody generated by the LIMID program and function tables used
when assigning the sounds, that is, the tables of subjects, verbs, objects and
so on. Figure 1.2b shows the resulting words used.
Figure
1.2a. A melody generated by LIMID assigning sounds to Subjects, Verbs and
Objects.
Figure 1.2b
Words attached to the melody generated by LIMID
Fred Lerdahl and Xavier Chabot developed a theory
that music and poetry have fundamental similarities and went on to produce a
very simple melody based on Shakespeare's Sonnet 29. Their summary of the
componential mappings between music and poetry is shown in Table 1.1 below[60]:
Table 1.1
Lerdahl and Chabot's summary of the componential mappings between music and
poetry
|
Music |
Poetry |
|
grouping = |
prosodic hierarchy |
|
meter = |
meter |
|
time-span reduction = |
phonological stress hierarchy, |
|
timbrel prolongation = |
phonological prolongational structure |
|
pitch prolongation = |
semantic prolongational structure |
From this mapping they develop a hierarchy of
semantic, timbrel and metrical attributes, much like the structural tree
diagrams and reduction maps of Lerdahl and Jackendoff's A Generative Theory
of Tonal Music. To give an example of the technique samples are taken from
a reading of Shakespeare's Sonnet 29 and used as a base from which the theory
can be tested. The first line of the sonnet is shown in Figure 1.3[61].
Figure
1.3 Lerdahl and Chabot's melody based on Shakespeare's Sonnet 29.
These three approaches show the forging of concrete
relationships between text and music. Cope takes a structural approach, making
the connection obvious through his parsing of sentence structures; Dydo
produces a melody and an accompanying text, each of which are based on both
intonation and sentence structures; and Lerdahl and Chabot use the actual sonic
structure of a line of poetry as a starting point to a musical composition. In
each case the composers have been at pains to ensure that their work has a
concrete and well defined relationship to a natural language, English in this
case.
Connections between music and poetry
Unlike
a painting or a sculpture, where it is possible for one to view the entire
artwork in one moment and to then review specific areas at leisure, music and
poetry can only be experienced over a period of time, the length of which is
dictated by the performer or reader. Within the context of time music and
poetry share three similar areas: the sonic, the structural and the cognitive.
In
the sonic area, music and poetry exist as purely physical events, the motion of
air particles. These events can be measured using amplitude against time
graphs; decoded using spectrographs; and described or notated in ways which do
not consider the perceived meaning of the information, such as musical scores
or written transcriptions. These methods of transcription have no audible
quality as they are not in the sonic domain, therefore they are divorced from
the sonic events they are representing. By notating and describing the sonic
area of a poem or musical composition it becomes possible to gain a deeper
understanding of the rules, or grammar, that give structure and cohesion to the
sonic surface, thus illuminating the structural area.
Examining
the structural area through a visual representation of the sonic string will
often show a deep, well understood, and well used knowledge of the various
grammars of music and poetry. These grammars operate on microscopic and
macroscopic levels, concerning, for example, the note to note relationships in
traditional harmony or the phoneme to phoneme relationships in a given
language, or on the macroscopic level, the rules concerning phrase or sentence
construction or the structure of a poem or musical composition.
Knowledge
of any given grammar can be acquired at a very early age and there is much evidence
to show that humans are predisposed to learning the various grammars of
different languages[62],
both musical and linguistic. Because this learning happens at an early age it
can appear to be intuitive, and we become so accustomed to this knowledge that
we often take it for granted, incorporating it seamlessly and essentially into
daily life. An example of knowledge of musical grammars and styles can be seen
in a listener being able to tell if a performance of a twelve bar blues fits
into the "jazz" idiom or the "heavy metal" idiom. This
distinction is made quickly and accurately even though the tempo, harmonic
structure and harmonic rhythm may be very similar.
The
cognitive area concerns how the sonic input is interpreted by the listener.
This level can be divided into three realms of exploration: psychological,
considering how the sonic information is being understood; physical,
considering how the sonic information is being computed and what physical
changes are occurring in the brain and body; and semiotic/semantic, considering
how the listener relates the sonic input to extra-musical events or items. In
each of these cases the skill and education of the receiver governs their
ability to decode the transmission and the methods used to decode it.