Music Derived from Other Sources

One imagines composed music as originating in the mind of an author. Indeed, I have for the last five decades been composing music in this fashion. However, for the last four decades I have also been repeatedly attracted to various methods of deriving music from sources both inside and outside the field. These sources have been linguistic, acoustic (field recordings), visual and mathematical (parametric formulae) as well as other works of music. For most of these operations I have resorted to strict algorithmic means and the use of computer programming. In this paper I will outline the sources mentioned above and go a little into detail in some of them. For instance the linguistic: I have used text orthography, spectral analyses of human speech, digital recordings of the human voice, and synthetic semantic structures. In the case of the visual, it was the conversion of pixellated photographs, city silhouettes, abstract films and geometric models as well as the visualization of algorithmically generated music that played a central role in the compositional plan. In other words, the paper will provide a general overview and examples. This is the first time I have covered more than one type of source in a single article.


PHONETIC COMPOSITION
While not strictly representative of composition in the usual musical sense, I cite three text examples of my work in this field. The first and last were put together as texts embedded in a larger context and the one in the middle as a stand-alone phonetic exercise.

Haiku (1968)
This work for baritone and clarinet follows the Japanese Haiku formula of 5+7+5 in its overall form -5 short pieces for baritone solo, 7 for clarinet solo, then again 5 for baritone solo. Each piece contains 5+7+5 notes of which the 5+7 form a twelve-tone row and the 7+5 the retrograde inversion of this row. The pieces for baritone are sung to ten meaningless palindromes written in the International Phonetic Alphabet consisting of 5+7+5 serial syllables based on English and German phonemes. They are listed here below: The Cologne Courses for New Music were held by Mauricio Kagel from 1969Kagel from to 1975. Each course was devoted to a different theme, the one in 1970 to "Hörspiel", which normally translates as "radio play". Kagel however meant something different, "neither a literary nor a musical genre, but merely an acoustic genre of indefinite content", as he wrote in the course brochure in German, English and French. I was inspired to 'pan' the three versions from the first through the second to the third. My result, written in the International Phonetic Alphabet as seen below, was printed in a programme for a public concert including my radio play CCU (1980), more about which follows in this paper. As examples of my wordpanning, see [jənuf] as a cross between German "genug" [gɛnuk] and its English equivalent "enough" [ənʌf], or [wjae᷉ l], a cross between "well" [wɛl] and its French equivalent "bien" [bjae᷉ ].

Ode to St. Cecilia (1988)
This text specifying misuses of music forms part of Fruitti d'Amore for cello and electronics. The piece was written for Frances-Marie Uitti, who recites the text at the beginning. The stanzas of verse were so written that every stressed syllable of one line phonetically resembles the corresponding syllable of another, a technique I call "omniproxonance" as opposed to "finaequisonance" for the final perfect rhyme; all spoken lines were also recorded, synchronised and used as sound material rendered electronically near the end of the entire piece. Fig.3 arranges the phonetically similar syllables vertically, e.g. the second syllable of each line: "view -through -su -mu -you -thu -few -mu -you -beau" ("qui" in the penultimate line is an exception, of which there are many).

SPEECH RECORDINGS
As examples of using speech recordings for composition, I will outline three techniques I developed, synthrumentation (1981), sound wave surfing (1987) and spectastics (1994).

Synthrumentation
I invented this term -from "synthesis through instrumentation" -to denote the following technique. A spectral analysis of a recording, typically of spoken language, adapted to a fundamental, which can move as compositionally desired, is rendered as a series of short chords comprising 'telescoped' Fourier time windows represented as a MIDI data file, which is then 'calmed': harmonic partials, the MIDI velocities of which fall near multiples of a set tolerance value, are rounded to those multiples and made continuous, i.e. the interruption between a partial in a chord and the same frequency in the next chord is removed. One layer of Im Januar am Nil is the synthrumentation of German sentences, made according to phonetic considerations relevant to this piece (plosives and other noise phonemes excluded). Fig.4 shows five stages in the synthrumentation of the spoken words "In Armenien" (= "in Armenia"), starting with the sound file. Its Fourier Analysis -time windows with harmonic partial numbers and amplitudes -is passed into synthrumentation, yielding a MIDI file mapped at centre right as pitch (y-axis) against time (x-axis). When the MIDI file is played, the sound is clearly recognisable as "In Armenien". This MIDI file was made into the score shown below; played by musicians -bass clarinet, four violins, two cellos and a double bass -its similarity to spoken language is still strongly evident.  Figure 5 shows MIDI pitch-time maps of three phrases "why me", "no money" and "my way" in Orchideae Ordinariae. Against one clarinet and seven strings in Im Januar, 50 string players were involved here. The words "no money" are especially well recognisable in performance.

Sound Wave Surfing
This technique, one I first used in 1987 in software I named "Samploop" (from "sample loop"), employs forwards and backwards motion, looped if necessary, along the samples of a recorded sound wave. It has three parameters: first and final sample number of a sound wave segment as well as number of iterations of the segment. For backward surfing the final sample number must be lower than that of the first sample. I first used Samploop applied to the spoken words "Herre Gott erbarme dich, Christe erbarme dich, Herre Gott erbarme dich" in Herre Gott from my piece documissa '87 (1987). Here is the input to Samploop, one segment to every line, for the middle section, in which the segment "dich" is repeatedly truncated (see also Fig.6): Fig.5 -Orchideae Ordinariae: MIDI maps of three synthrumentated phrases in English. The first sample of each segment moves rapidly from #51660 to #53100, staying there until the end of the section. The final sample moves gradually from #56652 to #53163. The segment length is initially 4992 samples (or 113.2 ms at a 44.1 KHz sampling rate) and shortens to a mere 63 samples (1.43 ms). Repeated 144 times, this last segment gives a constant tone of frequency 700 Hz and duration 206 ms with the timbre of the samples in the segment. Figure 6 is a "surf chart" for Herre Gott (1987) -diagonal lines represent sound segments played forwards 'normally'. Horizontal lines asseen mainly on the right are constant tones caused by short sample segments looped several times: the vertical width of a horizontal loop line (not obvious in the chart) sets the frequency of the perceived tone, theiteration count (horizontal length) its duration, and the inherent timbre is maintained: these tones contain only the input sound wave vowels [ɛ e ɔ ɛ ɑ ə i…] from "Herre Gott erbarme dich" etc.
My tape piece fLvXv$ (1990) using a further Samploop development uses words spoken by John Cage, taken from a 1988 interview Zygmunt Krauze made with me for Polish television. By gradually shortening the sample segments, the piece organically moves from a form of concrete poetry through rap-like music to electronic clicks and bleeps.

Spectastics
30 years ago I invented this term (derived from "Spectral Stochastics") to denote a technique for converting spectral speech analysis into music. Whereas "synthrumentation" renders the analysis as a 'calmed' chord-succession (see §1.3.1 above), "spectastics" generates a rapid succession of single notes: first the spectrum's amplitudes, calculated for every partial, are interpolated between the partials to give an amplitude value for every chromatic scale degree upwards from the analysis fundamental frequency. These chromatic amplitudes are then used as probability values for the random generation of a rapid melody, ideally 20-200 notes per second: the 'louder' a note is during a certain stretch of time in the chromaticized spectrum, the more frequently it will appear in the melody during that stretch. A spectasized melody, as rendered on a synthesizer or a player piano, is remarkably like the original sound recording. My principal use of spectastics was in my composition Farting Quietly in Church (1992) for baritone and six computers, performed three times by the late William Pearson . In one part of the piece, his words were spectasised live and the results sent in real time to an accompanying computer-driven player piano. Figure 7 shows the words "mais oui, Joanna" in three similar-looking forms: a sonagramme, and MIDI-pitch maps of a synthrumentation and spectastisation of the words. See also references (Barlow, 1996(Barlow, , 2008(Barlow, , 2009b(Barlow, , 2014a. (2003), a piece written -instead of by spectral treatment -in a 'language' crafted for several percussion instruments, termed "Progéthal Percussian": here sentences of a given text are grammatically parsed into parts of speech -nouns, verbs etc. with their attributes -and converted into a music score by a fixed set of rules, using a vocabulary developed with the help of the 1852 Thesaurus by Peter Mark Roget (1779-1869) -hence the name of the language. The piece purports to be a course in the language.

I developed this technique for Progéthal Percussian for the Advanced Beginners
To each of Roget's thousand categories of meaning I allotted a new five-digit code reflecting -in addition to Roget's six classes and the sections and divisions therein -shades of meaning such as negation, etc. A text was then taken as source material, parsed into parts of speech and attendant properties e.g. plural, genitive, etc. for nouns; transitive/intransitive, etc. for verbs, etc. After an elaborate process including a comprehensive set of grammatical rules in part inspired by various existing, in part newly devised languages, I wrote a computer program to convert the parsed thesaurally encoded text into a music score. Texts thus translated were Hamlet's Soliloquy, United Nations resolutions on states in the Middle East, as well as texts on commercial posters with instructions to children as to how and how not to behave properly.
These are Roget's first 24 categories in Class I, Sections I-II with my five-digit code at left: Hamlet's lines "To be, or not to be: that is the question: whether 'tis nobler in the mind to suffer the slings and arrows of outrageous fortune, or to take arms against a sea of troubles, and by opposing end them?" are here expressed in a meta-language, parsed and algorithmically converted into a meta-score. My five-digit code is seen near the middle of each line.
Of the above, the code at left e.g. in the first line "1 6 a <M1---002…--->" (for the first word of the text and the first bar of the music) means: Compare this necessarily compact description with the 24 bars shown in Fig.8. Also note that the stressed low note (lower staff, for the bass drum) on the fourth pulse of bar 3 means "not": compare this to bar 1 ("To be") -there is no bass drum note there. Shakespeare's text is shown interspersed between the original and the meta version above.

ACOUSTIC SOURCES
Here I cite pieces made out of field recordings done with technical equipment in various places. My octaphonic electronic work CCU (1980) layers seven of 20 hours of stereo recordings made during three months in the winter of 1977-78 at 60 venues in Calcutta, both in-and outdoors, to form a 48-minute piece reflecting the course of a hypothetical 24hour day in that city. Here, the main diffusion parameters are dynamics (fading the sounds in and out) and spatialization: given that the eight loudspeakers are symbolically placed on the city᾿s rim (miniaturized to the bounds of the concert hall), each recording is panned between two of the loudspeakers joined by a line, the midpoint of which corresponds in direction and distance to the place of recording viewed from the city centre. The loudspeaker panning matches the location of the venues, north in front, south behind etc.; the distance of the venue from the city centre determines the separation of the two speakers diffusing the stereo recording -the further the venue from the centre, the closer together the loudspeakers. Figure 9 shows at left a map of the city in its geographical size of the 1970s. It has an elliptical shape, its major axis tilted clockwise 8 arc minutes from due north-south. The river Hooghly, a Ganges distributary,

31
curves its way downwards in the upper left of the map. The 60 venues are barely visible as starlike icons ("☼") placed at their locations. Fig.9 shows at right a series of eight concentric ellipses with loudspeaker positions marked "1" to "8" along the rim of the outermost one. Superimposed slightly tilted on the map at left, these ellipses show a total of 8x8 = 64 zones. These adjoin the midpoints of the 28 lines joining the eight loudspeaker positions. In the right half of Fig.9, each midpoint, five seen as large black dots, is adjoined by groups of three conjunct zones. A stereo recording made in one of the three zones would be panned between the loudspeakers at the ends of the line with that midpoint. This type of panning was a very difficult task with pre-digital technology in 1980, given the large number of simultaneously running recordings and was achieved by a team of several persons acting together to achieve the desired result.
On the other hand, the octaphonic 41-minute Zero Crossing (2001) contains 83 recordings made daily at the same stellar time in 31 places on a three-month trip around the world in the winter of 1999-2000. Each recording was carefully whittled down to an 'edition' lasting exactly two minutes, the central 30 seconds of which (the 'core') is played at the original volume with the preceding and following 45 seconds constituting a fade-in and -out. Fig.10 shows a map of the world with my airports of call, all within 30 degrees north of a great circle poled in the North-West Pacific and the South Atlantic. The system of the layering of the 83 editions is shown in Fig.11; each edition is faded in up to the core and faded out thereafter. Incomplete editions are numbered in square brackets.

TEXTMUSIC VERSION 8 (1971/1973)
This version of Textmusic (see §1.1 above) is based on a poem in Cologne dialect (a Moselle-Franconian language akin to Dutch) praising that city's virtues. Fig.12 shows at left abstracted postcards of I. the Opera House, II. the Cathedral and nearby bridge, III. the mediaeval city West Gate, IV. Rhine panorama with Cathedral and the Church of St. Martin Major, and V. St. Gereon's Church, which were applied as parametric curves in a computer program. The unique result is inextricably linked to these curves; e.g. postcard II sets the key-colours initially at 90% mixed (chromatic) and 10% white (diatonic) and finally at 100% black (pentatonic), due here to the bridge's proximity at the picture's right end -see the two scored examples at right: the upper one shows the mostly chromatic beginning, the lower one the piece's pentatonic end See also references (Barlow, 2009a(Barlow, , 2012a(Barlow, , 2018.  In 1995 I was commissioned by the "100 Years of Film" festival of the Art and Exhibition Hall of the Federal Republic of Germany ("Bundeskunsthalle") for music to be synchronised to the 1930 film Study No.6 by Oskar Fischinger . The original soundtrack by Jacinto Guerrero (1895Guerrero ( -1951, to which Fischinger had synchronised his images, was to be replaced for the festival by a contemporary work. One of the composition's layers is a metamorphosis of Fischinger's dancing objects; here I effected a phonetic linkage between the width and height of the objects and the sound colour of a resulting tone-cloud: horizontally stretched shapes yielded (as does the human mouth) an [i] ('ee') sound, squat shapes an [u] ('oo'), with sonic interpolations for in-between shapes. Estudio Siete (Spanish for "Study" or "Studio" No.7) was composed in Cuenca near Madrid and realised in the studio that I used to head in the Royal Conservatory The Hague (Studio 7). It is best film-synchronised on a player piano. Fig.13 shows schematically at left the first 12 frames of the Fischinger film with their spectastically (see §1.3.3 above) extracted notes, starting at frame 3. At centre right is a hypothetical frame with five objects, further right are the "X Y" coordinates of these objects' lower left corners and their respective widths and heights ("W H") followed at far right by phonetic formants as MIDI numbers and amplitude curves. Spectastically interpreted formant amplitudes are seen as overall probability curves and as their sum in the "Σ" box (top right corner).

KURI SUTI BEKAR (1998)
Written for the pianist Kristi Becker, whose name, pronounced in Japanese, inspired the title, Kuri Suti Bekar (1998) consists of a Prelude and a Chaconne. The short Prelude (see the left box in Fig.14) is taken from the pianist's name in Japanese Katakana script (top) -the right hand plays [ku-ri-su-ti] and the left [bɛk-ka-ro] simultaneously (MIDI pitch-time map below). For the Chaconne (see the lower half of the box at right), Ms Becker's photograph, scanned into an 88x88-pixel matrix (box top centre left), was filtered and mapped ten times to form a ten-page score with eight bars per mapped page: in each map, 88 horizontal pixels mark 88 pulses in time (11 per bar), and 88 vertical pixels the 88 piano keys. The first two maps are shown enlarged at top left of this box. For each map, pixels were selected by constraints, the three most significant being 1) the transection of the pixellated portrait by two sets of 120 radial lines three degrees apart, 20 of them unused -see comment below, 2) a harmonic matching of my piece ...until... #5 (1972/74) in 11/16 metre, frequently played by Ms Becker (each of the ten mappings scans eight bars or 88 pulses of this piece, reflected in the eight bars of each page of Kuri Suti Bekar), and 3) a phonetic spectral analysis, serving as a sonic filter, of three Bengali words also forming the title: "kuri" (twenty), "suti" (cotton threads) and "bekar" (jobless), referring to the 20 idle radial lines. Superimposed, the ten pitch maps yield the graph at top centre right, clearly resembling the pixellated photograph. Bar one of the score is at top right. See also references (Barlow, 2009a(Barlow, , 2010a(Barlow, , 2012(Barlow, , 2018.

LES CISEAUX DE TOM JOHNSON (1998)
The French title refers to the "O"s in composer Tom Johnson's name, for whose 60 th birthday this piece was written, and the six circles indicated below -"Six O" is pronounced [sizo] in French, as is "Ciseaux" or scissors, a pun. See also references (Barlow, 2009a(Barlow, , 2010a(Barlow, , 2017(Barlow, , 2018.

LE LOUP EN PIERRE ("==") FOR TWO ORGANS
This piece (=The Stone Wolf, "loup" = wolf, "pierre" = stone = Peter) was composed for the two organs in St. Peter's Church in Leiden, the mean-tone Van Hagerbeer organ at A=419 Hz and the equal-tempered Thomas Hill organ at A=440 Hz. This mean-tone-tuning has a so-called "wolf fifth" between G# and Eb, noted in the title. The first section of the piece involves a sonification of a pixellated image of the church building, calibrated to the Bark scale of subjective pitch (see Fig.16): the pixels in the image are allocated to keys on either or both organs, the Bark or 12tone tempered pitches of which are the closest.
The building, graphically rescaled to the MIDI-pitch-scale (see Fig.17), now provides the first 97 chords, initially filtered by a horn shaped centre on pitch No.68=Ab, the only note common in pitch to both organs. Starting with this pitch, the range gradually widens to large microtonal clusters engendered by the central Gothic window. After the first section, the piece continues with other considerations related to harmony (see §4.2. below) and synthrumentation (see §1.3.1 above). See also references (Barlow, 2009a(Barlow, , 2010a(Barlow, , 2018.

MATHEMATICAL SOURCES
Two sources are demonstrated here, one -ISIS -in the realm of Digital Sound Processing and the other -the quantification of harmony and metre -in the realms of pitch and pulse, melody and rhythm.

ISIS -FOR "INTRA-SAMPLAR INTERPOLATION OF SINUSOIDS"
This technique resulted from my having doubted, already fifty years ago, that white noise is the sum of all frequencies. I argued that if this were the case, each frequency, one of an infinite number, would have to have zero amplitude, in order not to add up in loudness to infinity. The model I preferred for white noise was that of a single sine tone moving randomly to all possible frequencies at high speed. After some experiments with MIDI in the 1980s that seemed to bear out this supposition, I finally got down in 2001 to the mathematics of determining the variable sine-tone frequency of any sound wave in general. Presuming that it was sufficient to set the speed of frequency change at the sampling rate (currently commonly 44.1 or 48 KHz), I interpolated hypothetical sine-wave segments between the samples of a sound wave, such that each segment would pass through limiting values ±1 once each and the final phase of one sine segment would equal the initial phase of the next. Figure 18: Formula for ISIS-analysis of a sound wave with a graphic example Figure 18 shows a set of seven samples (dots) connected by a spline, but more significantly by six sine segments, the frequencies and intersegment phase of which are given near the top. The formula yielding these frequencies is at top right, where "ƒ" is the frequency, "R" the sampling rate and "s1" and "s2" the values (within ±1) of two successive samples. It is easy to see why the frequencies centre around the sampling rate: if two successive samples are equal in value, the sine segment between them will form exactly one period of the sampling rate frequency.
This extraction of sine frequencies from a sound wave really turns the latter into a 'melody', albeit ultrafast and ultrasonic. But it is possible to transpose this melody down into the audible range and to decelerate it to a perceptible speed.
The reverse is equally possible, to transpose any melody in pitch and speed up to the sampling rate, thus converting it into a sound wave, as in Fig.19, which shows J. S. Bach's Jesu Joy of Man's Desiring applied to the reverse formula for ISIS-synthesis -and the resulting sound wave and DFT spectrum. The input is in the column at lower left: labelled "SIS" for "sustained interval sequence", it expresses in cents the Bach melody centred on B4 with the numbers in the right column "xn" signifying the extent of holding or sustaining the notes to their left.
ISIS thus combines the domains of melody and rhythm with that of timbre to form a continuum. I have used this technique in my Four ISIS Studies (2003-2008 and in the second of my Five Dodecaphonic Pieces (2016Pieces ( -2017 and plan to continue its use. See also references (Barlow, 2005b(Barlow, , 2010b(Barlow, , 2012(Barlow, , 2017(Barlow, , 2017.

THE QUANTIFICATION OF HARMONY AND METRE
In the late 1970s, in search of an effective way to numerically estimate how harmonic a pitch interval is (including so-called micro-intervals), I studied work by Pythagoras, Hindemith, Partch and other music experts as well as relevant work in the field by the mathematician Euler. None of the above helped me make an exact and usable measurement, leading me to research on my own. This resulted in a formula for "harmonicity", based on one for a coefficient I called the "indigestibility" of natural  numbers based on both their size and divisibility. Both formulae and corresponding tables are shown in Fig.20. Take the numbers 7, 8 and 9: the indigestibility of the prime 7 (=10.3) is high in comparison to that of 8 (=3.0), a number divisible down to the smallest prime 2. The indigestibility of 9 (5.3), lies between the values for 7 and 8. I define the harmonicity formula as the reciprocal of the sum of the indigestibilities of the numerator and denominator of the ratio or fraction for the interval, with an added feature: if the numerator is less indigestible than the denominator (as in 4/3), the upper note becomes the the interval's root and the harmonicity is made negative as determined by the plus-minus "sign" function "sgn(ξ(Q)−ξ(P))" where "P" is the denominator, "Q" the numerator and "ξ" indigestibility. The harmonicity e.g. of 4/3 or 3:4 is −0.214, as seen in the two right-most columns of Fig.20.
Parallel to these harmonicity formulae, I came up with another set of formulae for "metric indispensability", the relative metric significance of a pulse within a metre. Shown with these formulae in Fig.21 at bottom right for the metres 3/4 and 6/8 on an eighth-note level, these indispensability values, unique to each pulse, are for the former 5 0 3 1 4 2, and for the latter 5 0 2 4 1 3, reflecting the respective metric significance of the six pulses therein. A process called "rhythmic dilution" is also shown at bottom right, removing attacks in order of increasing indispensability, further demonstrating this system, intrinsic to Autobusk.

1981 (1981)
In my piano trio 1981, piano trios by the three composers Muzio Clementi (La Chasse in C, 1788), Robert Schumann (Trio No.2 in F, 1847) andMaurice Ravel (Trio in A minor, 1914), one movement each, formed the starting point. In Fig.22, each instrument, violin, cello and piano, describes an individual spiral starting at the centre of a triangle with the composers Clementi, Schumann and Ravel at apices C, S and R. The numbers beside the images of the three instruments are score markers of multiples of 18 bars. The distance of the instruments' position from an apex at any given time is mathematically related to the amount of the music of the composer at the apex in the general mix. Thus, at the start of the piece, at the triangle centre, 33% of the notes in the score come from each composer. But at bar 410, just to the left of marker 23, all three instruments are together at the Clementi apex: the score contains 100% of his music at this point, together with a small percentage of the music of the other two composers, above zero since they are not infinitely distant.  See also reference (Barlow, 2012).

WÜRFELGANG (1995)
As modular notes and instructions written by Wolfgang Amadeus Mozart, his Musical Dice Game published in 1793 made it possible through combinatorics to 'compose' a large number of new waltzes. For each of the 16 bars to be made he provided 11 alternatives, one of which is randomly picked by dice. This creates about 46 x 10 15 different pieces. So far, many computer programmers have made real-time realizations of this piece. New in this project was that a statistical Markov-analysis, named after its inventor Andrey Markov , was applied to Mozart's material causing as expected different degrees of 'note-fidelity' -at Markov level 4 it is not easy to tell the result from the original, at level 2 the music sounds interestingly distinct from the original, but remains stylistically relatively faithful. Level 0 is only able to produce clumsy but comical imitations of Mozart's music. I realised this Markov-treated Mozart work as the computer program Würfelgang ("Würfel" = "dice", "Gang" = "course"), a pun on "Wolfgang", which sends MIDI notes to a computerdriven player piano or synthesizer -see Fig.23.

)ERTUR( (2015)
When requested by my Los Angeles-based art-collector friend Raj Dhawan to write music for an exhibition of his collection of Alphonse Mucha (1860Mucha ( -1939 paintings, I at once thought of basing the music on that of a Mucha contemporary and compatriot. My choice fell on the composer Leoš Janáček , born, like Mucha, in Moravia (then in the Austrian Empire, today in Czechia). 37 selected Mucha paintings are matched by 37 Janáček pieces, many of them movements of larger works. The bigger the paintings, the longer the music selections are: the square root of the areas in inches of the Mucha works equals the durations of the Janáček selection in seconds. Each of the 37 music selections is accompanied by a video showing the corresponding Mucha painting, expanding in accuracy from its centre as described below. A sharp crack preceding each section stems from a clapperboard as used in movie recordings.
At first the music is constrained to the range of a minor 7 th , all notes outside this range being discarded. The notes are also redistributed among five spatially separated instruments -flute, clarinet, violin, cello and piano -and the pitch range gradually increases to just over four octaves, each instrument being allotted exactly ten different notes by then. Also every instrument except the piano plays microtonally -see Fig.24 for an example of a couple of bars of Janáček's original Barn Owl compared with its treatment in )ertur( in section 5 -the grey-shaded notes in the Janáček example are filtered out in )ertur (. Analogue to this, each Mucha painting is first shown only with its most widespread colour, the rest being in grey. During the run of each Janáček piece, the colours of the Mucha works are gradually restored in range, starting at the middle of the image, to finally include all original colours in the entire picture. Fig.25 gives an example of this in section 5 of the piece, albeit grey shaded here. This audiovisual composition bears the title )ertur(, which -resembling the widening ranges of pitch, area and colour -could be expanded to include words like aperture (English, French), apertura (Czech, Italian, Polish), abertura/cobertura (Spanish, Portuguese) or copertura (Italian), all implying opening or covering. The reverse parentheses in the title are also relevant, the left one representing a fragment of a preceding "p"or "b", the right one that of a following "e" or "a". Irrelevant: "ertur" happens to mean "peas" in Icelandic.

CONCLUSION
This is a comprehensive account of my use of a range of pre-existing sources to compose music by generative means other than the detailed imagination of sound. The processes outlined above were shaped by my taste, that subjective and descriptionally elusive phenomenon, towards results aesthetically welcome to me and potentially to others. The processes were convergent on the conceptual and/or executional level: on reviewing each developmental stage in making the pieces described above, my artistic judgement would usually lead me to make a change in the generation process, typically algorithmic, often followed by further alterations, until I achieved a gratifying result I could declare final. This is analogous to preparing a meal, gradually improving its taste and texture until the result satisfies its cook, who thereby 'overwrites' each stage of the meal's development, keeping the final result for consumption.