Old Fashioned Composing from the Inside Out:

On Sounding Un-Digital on the Compositional Level

by Laurie Spiegel


This is a personal subjective survey of traditional pencil and paper composing and editing techniques from one traditionally educated composer, to try to better communicate the freedoms and abilities fostered by these techniques to authors of music composition software who may have had less experience with music composition per se. It attempts to describe what a notational composer is accustomed to being able to do and how and why. This is only one composer's view, and other composers' experiences should be sought as well because no single artist's experience can ever be comprehensive.

This description hopes to give authors of compositional software food for thought in designing extensions and alternatives to both the currently-dominant sequencer model (which simulates multi-track tape recorders), and the process-oriented methods of logical algorithmic generation which have finally begun to be more widely accepted and in which I have specialized in my own software.


Despite my 20 years experience composing primarily in electronic and computer media, I often still compose on paper the old way because of the freedoms and depths and kinds of expression the paper and pencil give me which are still unavailable to me in any computer composition system I know, not excepting those I've created myself. I usually write music on paper at a keyboard or other instrument, but not always, since for me the imagination is still the ultimate unparalleled creative tool, and the processes of envisionment, focusing, clarification of detail, and external realization of whatever one holds and elaborates in the imaginative memory is, in my view, an irreplaceable creative experience of the highest level, providing freedom and expression well beyond any other technique I've tried.

The "improvise and edit" method of composing is now becoming the dominant compositional technique of our culture, in large part due to the tremendous proliferation of recording technology, both tape and computer based. Though I use both recording and algorithmic generation often in making my music, I would not want to be limited to them. What follows, therefore, is an attempt to describe from subjective experience some composing and editing methods which have worked for centuries so well that we can't afford to neglect them in our new software tools. Some of these observations describe simple direct operations, and others just try to give a feel for the compositional process. Much that they refer to has been somehow left out or misunderstood in the design of most music composition software currently available.


I know of no commercial software with this function.

Appending to the end is the commonest technique, usually to all voices (what would be called all "tracks" these days) together. You add small enough amounts that you can safely hold them in your head without losing them until you can write them down.

The major skill practised in the formal study of composition in my opinion is the ability to expand as much as possible the amount of musical material you can hold clearly in your memory. Music theory, harmony, etc. are in part mnemonic methods for increasing the amount one can remember. They enhance this central mental skill. The amount of music you can keep clearly in your head at a time grows with years of practise.

Nonetheless, you still keep the chunks you append small because they are concentrated areas of intense decision making activity, multi-dimensional, and very context dependent.

Non-Sequential Addition of Material

You usually add at the end of the piece or section, but not always. Sometimes you skip beats or measures and go back to fill them in later when you've gotten a better idea of the whole context, what something is leading toward. Sometimes you go back before the beginning of the entire piece and write introductory material that works up to it.

Scanning for Context when Appending

When adding material at the end, you generally back up in the piece and take a running start reading or playing through what you've already written and then continue beyond it creating new material. You stop playing beyond the written ending point when you've done as much new stuff as you can keep in your memory, or have accumulated enough indecisiveness in the transition and the material to require some testing and deciding. Then you write down or merge in the new stuff, back up and repeat the process again.

If you can't decide on something, you may back up further and do a longer run through to provide a greater sense of context.

Where to Start Scanning

The point at which you start playing the written stuff that leads up to the place where you'll start appending is not a single hard and fast place specifiable by rule. You might go back a phrase or a couple of measures, or sometimes more. You might start at a simple 8-bar phrase boundary, but not usually. You may have to chase a lot further back in some voices than in others to set up the context for where your mind will take the piece next, for example, getting long-sustained bass notes sounding, setting up the rhythmic feel, or the experience of a tempo change such as an accelerando. You might run through the material further back in harmonic reduction (playing just the chords on the strong beats of the measures), and when you get closer to what you're working on now, start playing literally everything that's written before coming up with new stuff to write down at the end.

Not Just a Plain Append Operation

This is not really like having your sequencer automatically punch into record mode at the current end of a multi-track piece for several reasons.

One is that you are reviewing what you've already written as you run though it, and you refine it as you go, adding to or changing it, because you've clarified your ideas in the meantime and also you now have some idea what it's leading up to. You look at the score, pick a starting point for the run through and "record" (remember or notate) any differences between what's already written and what you are playing (or hearing in your head).

The second reason is that the stuff you come up with to add at the end is less likely to be added like a splice than via some fairly complex merge, modify, and filter operation. When you stopped appending the last time, you didn't know where the piece would go next, and couldn't write in preparation for it.

You hear changes and buffer them mentally, trying variants, before you decide exactly how and what to commit to paper. It's as though you've recorded them alongside the score in a separate synchronized sequencer in your head. After you merge and append the new stuff, the result probably contains the previous material modified slightly at its end, the new material with it's beginning modified to make a good transition, and possibly some altogether new bridge material, created during the actual written appending process to make a good transition.

A third reason is that you usually try several takes of this "running start" operation, as exploration, clarification, and experimentation, and as part of your decision making and refinement processes, and you may wait a while to decide and want to keep all of them in the meantime.

If you write down several alternatives, you find yourself with a branching piece to deal with (usually branching to several sheets of paper with different versions). So you take a break, or else start another scan from further back to try to decide, or else skip some blank space in the score and go on, figuring you'll fill in the missing transitional part later.

You may also just decide to use whichever one of your branches turns out to be easiest to extend. Some material just seems to take off by itself and other stuff just dies and won't go any further no matter what you try.

Sketching: Shorthands, Symbols, and Speed

Sometimes you jot down just enough to capture the new stuff or changes in very rough sketch, in order to keep going and not lose the music's flow and momentum. Then later you'll go back and fill in the details. This means writing a first pass leaving out notes, voicing, dynamics, phrasing, orchestration, lacking exact rhythm, maybe marking just single pitches here and there, or chord symbols for the downbeats, or drawing visual curves for dynamic contours or for pitch patterns. You may have personal symbols and private notational shorthand you've invented for new or non-standard things you often do.

Parts may drop out and back in again as you sketch. Some areas will have more detail filled in that others. You may leave some sections completely blank, to be filled in later, in order to capture as much of the total concept of the piece as fast as you can.

On Interdependencies and Top Down Design

Composing can be thought of as a bit like doing a crossword puzzle. There are hard words you leave blank the first pass through because you know that after you've filled in context, the words that cross them running the other way, they'll be a lot easier to get. This rough-it-in-and-refine-later composing method is also a bit like doing computer pseudo-code first then conforming it to proper syntax later. It's also to some degree like "top down" structured programming, except that, once you've started actually writing, doing an overview just about always gets triggered by specifics on the most detailed micro-event level of a what's already written down, instead of the other way around.

Sometimes you design the overall structure of a piece top down and then fill it in, especially in scoring accompaniments for dramatic material such as film. But in composing for concerts or records, pre-structuring often ends up sounding forced or artificial. More commonly, the overall structure of your work only becomes clear afterwards, in retrospect of the process, as an accumulation of moments built up intuitively during the your immersion in moment-to-moment detail.

Multiple Passes and Harmony versus Line

You can't think about all aspects of the section you're working on at once. Usually harmony, line, shape, pitch, or rhythm, or some combination of these are first. Voicing (the distribution of the notes among the parts), dynamics, orchestration, and the specification of articulation and phrasing tend to be done in later passes, but notes for them may be made in places in the rough draft at any time.

The technique of doing all aspects of each voice (each instrument) as a separate pass, as in tape overdubbing or MIDI sequencing, hasn't been used in European composition since the cantus firmus technique of the medieval Notre Dame School died out. It's death can be strongly linked to the evolution of tonal harmony, which gives the relationship between all pitches sounding at one time (chords) a higher priority than it gives the integrity of each melodic line which runs through the chord.

Free Movement Between Data Types

The idea that you are making multiple passes through the score working on different dimensions of the information in it is an oversimplification. You may be working mainly on pitch-time stuff, or dynamics, or whatever, but your access to other types of information is unrestricted.

You freely make pencil notes about any other aspect, or add information of any other type, at any time, without any experience of interruption such as is created by having to switch modes in a computer based editor. All information of all types is available random access at all times for all relevant editing operations. It is also open to you for the attachment of verbal comments, asterisks to mark inter-connections, or whatever else you find useful.

Non-Sequential Groupings

The groupings you work with are not commonly sequential in time or restricted to one part (one "track"). One moment across all the parts for some dimension is probably most frequent (harmony, loudness, instrumentation, ...) You might put accents on all the downbeats of the measures in a section. You might write chords or put rests on the strong beats for all the voices, and then make melodies and rhythms within the measures later. You might specify dynamics for the whole orchestra or make all the strings go pizzicato at once.

You often group the material for one of your editing operations non-sequentially by musical dimension, timbral similarity, or temporal placement. Sequential grouping for editing is probably just a carry-over from word processors that was never well thought through for musical use.

You don't always compose the notes in the order in which they'll be played. If you rough in just the chords on the downbeats first, you may not think at all about which specific notes will fall in which instrument. You have no idea how many events you'll break those chords into later when you write them out and fill in, working out melody, counterpoint, and rhythm from them.

Filling in Abbreviated Sketches

Aside from voicing between chords, there are other ways of placing events then interpolating between them later. For example, you might choose a few pitches that will be high points in a series of successive upward-running melodic lines, then go back and write the lines that run up to each of these successive high pitches later. In this case, the last note of each measure may get composed first for several measures in a row, then the earlier notes in each bar that lead up to them are composed after. Or you may decide on a series of musical arrival points, then figure out the best routes by which to approach them later.

Not Getting Carried Away with Improvisation

Improvisation can be limiting in that it prevents you from working easily with materials too complex to play, or perhaps wrongly spaced for 2 hands on any single instrument, or which require several instruments you don't play at all. Your muscles and nervous system have formed reflexes and habits which will lead you down paths you've already tried, which is counter-productive. Improvisation may be an indispensable technique for composing, but it is only one techniques among others.

Improvisation may also put you at risk of getting carried away, of developing too large an amount of new material in a singe take to append or integrate well. For one thing, you may not be able to remember all of it when you go to write it down. Just as importantly, you can get off on tangents, wander, get lost, jeopardize the continuity or distend the proportions of your piece.

You can lose the mood or momentum of where you began when improvising extended takes, and may have to spend a lot of time re-establishing the feeling of the starting place in your mind before you can decide on what you did or can try alternatives. You may need to go back to the beginning of the entire work and read through it all, possibly several times, to put the new stuff into perspective within the piece, to ensure its continuity, appropriateness, and proportionality.

Often a fair number of changes must be made while appending new material on both ends of the transition. With music of any complexity, working with too much new material at once can be disorienting and inefficient. Each composer knows their own limits on this, and they change considerably with the type of material being written too. Some things are written one note at a time. On the other hand, with very simple material, one can sometimes "append" an entire piece start to finish to a completely blank page in one pass.

Forest or Trees? Multiple Perspectives

Other reasons to compose bit by bit on paper rather than taking down long improvisations are to maintain selectivity about material, and to optimize for qualities that can only be gotten by constantly changing your perspective on the piece. Qualities such as controlled complexity, consistency of material, or integrity of form can be kept in adjustment because you can step out of real time. You can see the overall piece from a vantage point outside of time. You can jump around freely within time as it is represented on paper.

Composing is the process of designing an experience for others. It requires a designer's overview perspective as well as a builder's preoccupation with detail. You're always outside of the work, objectively viewing it, pretending you're hearing it for the first time, at the same time as being lost and immersed in the full detail of all the parts and dimensions each moment contains. You simultaneously maintain and ignore both levels of view.

Outtakes and Separate Starts

Often the new stuff you compose doesn't work where you intended, but it might belong in the piece somewhere else, so you put it aside and go on writing other stuff. You keep a little heap of such outtakes - chord sequences, melodies, rhythmic figurations, counter-melodies, variations, sonorities, etc. Some get used later and some don't. Some may end up in other pieces or turn into pieces themselves eventually.

Sometimes you take such an axe to a piece that you virtually start all over and the whole first draft goes on the scrap (resource) heap as an outtake. Or there may be several generations of sketching which successively approximate or refine a vision. Parts of all of these may get reworked into the final piece, so access to multiple piece (or fragment) files from within a music editing program would be essential. Sometimes pieces which you start as totally separate turn out to be part of the same unified work (the subconscious mind again) but you were completely unaware of their relationship when you wrote each of them, and are surprised to find that they fit together perfectly (with a bit of transitional or other editing).

You may need to start over with a clean piece of paper and build the final piece from scratch while referring to all the other bits and pieces that somehow are part of it. This can be done in a very premeditated structured fashion or it can be done spontaneously, intuitively, so that you never know moment to moment what you're going to try putting where next as you brainstorm on what to try. There is a lot of comparative looking back and forth among 2 or more pieces or parts of the same piece in such processes. (Multiple window displays are optimal if a screen is big enough, but hardcopy will do, as it has for centuries in the past.)

You may want to try re-using a line, a rhythmic, an articulation pattern, a sonority, etc. from an outtake scrap or another piece you did last year, or maybe a line, or a rhythmic ostinato or other material from a different movement of this same piece or from your stash of out-takes and related ideas. You want to be able to get at anything you've ever written fast, with minimal interruption to what you're working on now, and review it or copy whatever aspects of it you want to reuse, while changing what you need to about it (key, tempo, meter, timescale, rhythm, ...). It's almost as though each aspect (pitch, rhythm, articulation, doublings, ornamentation, etc.) were independently recyclable.

Debugging a Piece

After you've added the latest extension, you realize that the pacing is wrong. A climax was reached too fast, too early, and you go back and rework one or two measures into a half dozen to make the climax happen later. Or a section feels like it's moving too slowly after what was before it, so you try breaking up the longer notes into shorter ones, or adding rhythmic density by repeating notes, or find other ways to increase the density of activity without altering the rate at which the basic pitch content (the "harmonic rhythm") moves. Or you might delay the problem section by inserting a repetition of earlier material, changing it to prevent it's getting boring.

You may try various solutions to a problem on separate pieces of paper instead of erasing and rewriting. If none of them work, you may get rid of the offending part altogether and stash it someplace just in case.

Unity and How it Gets There

As you work more and more with material, you sometimes realize that fragments of the music which you thought were unrelated are actually all variants on some simple or central theme that you just had not been aware of before. (The subconscious does a lot in composing.) Then you might go all the way through the piece looking for other such similarities and do some editing to clarify them. You might create a section at the beginning or end which makes the camouflaged commonness more explicit, which plays whatever has been recurring in a clear and distilled form, making it easier for the listener to experience unity in the work.

When they analyze scores, musicologists often think that composers have intentionally made variations on a single theme in places that a composer would suspect they may have just found or exploited similarities they wrote without planning to.

Computer and electronic compositional tools have strong biases toward musical repetition (looping, copying, etc.). Just about all music editing programs, both scoring and sequencing, support the technique of copying and changing to create variations. But they universally lack even the most simple support for finding and grouping similarities (such as pattern-oriented search or search-and-operate functions), let alone highlighting or increasing similarity.

Unplanned similarities among different parts of a piece also happen consciously or unconsciously within single dimensions, such as a second melody having either the pitch line or the rhythmic pattern of an earlier one, but not both. It's hard to pinpoint the similarity and yet there is a vague familiarity of feeling in the second one when it's played which somehow enhances the emotional experience of the piece. You find you've re-used something as abstract as a melodic shape, maybe even upside down, or in major instead of minor the second time, or with different harmony or rhythm, or in another key or sound. I sometimes find things like this which I was unaware of in my pieces even years after I finish writing them.

Sometimes you realize that what you just wrote leads back to something you wrote earlier, so you try appending a variant of it. This may be anything from a literal repetition of the earlier stuff to something which only vaguely resembles it. Some kind of transformation is usually needed, because the key, rhythmic structure, position in the bar, density of surrounding material, meter, or other factors are likely to be different. Even if it fit perfectly, it would not have the same impact in its new context.

Sometimes something done earlier seems like it might be good in counterpoint with (simultaneous to) what you've just done, and you copy it there, and edit pitch and rhythm in both to make them fit each other.

You want to be able to get rid of such attempts at least as easily as you can try them. You try many many more things than you end up using. You want to be able to make multiple versions of many of the changes you make to your score, and quickly and easily try them one after another while deciding. Paper is not much use for this but is of some help. A well developed musical memory and visualization technique are vital and there is no technology outside of the body that I know of which has its properties. This deserves much more study.

Polyphonic Writing

If you're writing polyphonically (in distinct parallel parts each of which has real coherence) instead of homophonicly (which is what is too often meant by synthesizer people by "polyphonicly" - "homophonic" means several sounds at once but not arranged into a fixed number of distinct parallel voices), sometimes you write one or more voices way ahead of the others, and then you go back later to fill the others in to catch up. You need to be able to leave gaps and spaces, and actually just hang bits of music "in thin air" on the page. You might know that at the end of 8 bars you're going to be in a certain octave or chord and write a melody starting on it there, leaving space on the paper to fill in later whatever will work toward it well.

Sketching in Time

It's sometimes easier to compose a stretch of music if you've left yourself verbal notes, or hints, or a map of where you're headed further on. Blank music paper maps the time which lies ahead waiting to be filled with information. Just as the invention of "zero" as a placeholder for numbers increased our arithmetic ability, blank time-space on paper facilitates our composing. The opposite of written music is neither blankness nor silence. It is unfilled metric (or temporal) structure, and that's what blank music paper shows.

Barred music paper can be written on, vaguely or precisely, at any timepoint, no matter how discontinuous from anything already written. I can draw large-scale dynamic curves pages ahead of where I'm composing in orchestral writing. I can draw this curvature of musical intensity in real time, counting through the bars as my pencil draws the curve, using no actual musical notation symbols at all till I fill that curve much later with specifics. I can write partial notations, symbols, curves, chord names, and patterns across a blank metric page and have much of the actual composing done before I write a single note.

Polyphony and Imitation in Open Score

If you're writing polyphonically in "open score" (doing the voicing and orchestration at the same time as composing the notes) you tend to write one voice till it's longer then the others, then while reading it, you go back to a shorter one and write it until it's longest then while reading those, you go back to another shorter one and write it forward until it's longest, etc. You may copy material from one voice to another, so it's heard at a time delay and usually at a transposition too, with or without inversion or some other transformation,. (That's how you build things like fugues.) But more often you just write freely.

When you're extending one of the shorter voices, it has to fit in with what's already going on in all the others. You need to be able to see its immediate past, its present, and also, importantly, its future context - the parts which have been written ahead of it. If you can't see where you're going (you can't in tape overdubbing), you have work from the limitations of memory and can't read ahead. (Notation is an adjunct to memory as much as a communications medium.)

The part written farthest out ahead of the others, is free to move wherever it wants. The others will conform to it when they catch up. But each note you write limits your choices for the notes after it, and must drag the other voices along in the directions it goes, so you don't want to write any voice too far ahead of the others. Though you're writing lines, you still want to stay fairly close to working beat by beat across all the voices in the piece at once.

Homophonic Writing

If you're writing an accompanied melody instead of polyphony, you sometimes make some kind of filler material, such as chords or repeating patterns, to give your melody harmonic and rhythmic context. Then you go back after you've written your line against it, and replace or change the filler, perhaps breaking the pattern into several melodic lines, or otherwise making it seem to respond to the melody you wrote on top of it. Then you may make further changes in the main melody. You may work on them alternately or both at the same time, to make them seem to respond to and rely on each other. You use and edit all their musical parameters: orchestration, pitch, rhythm, dynamics, phrasing, articulation.

Integrating Algorithmic Material (An Aside)

This doesn't pertain to normal pencil and piano composing, but sometimes, when I've composed at a keyboard that links to a computer, instead of standard musical patterns (arpeggiations, etc.) as filler. I've used computer algorithms to provide a texture as background to something I'm working on. (Algorithms can be viewed as a new kind of musical descriptive shorthand.)

I've tried to work rules into these logical processes that let them respond to what I do live, as a simulation and shortcut of the process described above. I also use them to generate material which will be used strictly for background, for example, to a voice-over for a commercial announcement. I sometimes record or write down a length of such a texture, then go in and change individual notes, and add lines that stand out and have real character against them. I may orchestrate from that. Or I may use them as spring boards for my auditory imagination, fantasizing material from them which I then write out.

Symbols, Shorthands, and Standard Terms

Standard symbols (chord symbols, old fashioned figured bass, ornaments) are often used as "shorthand" for groups of notes. Substitutions of many fast notes for single long ones (trills, tremolos) are done by writing a symbol for what is to be substituted by the performer when he plays it. You don't ever have to write out all those little fast notes. Chord names or other symbols for groups of notes speed up writing, making it closer to real time or to the mind's creative speed.

I sometimes just write the stems and beams for the rhythmic aspect of a passage without any pitches at all, and then come back, often much later, after the rest of the piece has been worked out and the context for the section is clear, to fill in the actual notes.

You write verbal reminders to yourself on the score about things you're not bothering to work out now - instruments you might want to use at certain points, interpretive ideas you might forget later, names of pieces or composers you're reminded of, qualities of sound you're after. You also often notate on the fly, while concentrating on something else, articulations, fingering, tempo changes, dynamics, phrase marks, possible alternative notes or lines you just can't make up your mind about right now, cross references to other parts of the piece or to stuff in your heap of outtakes, references to film action cues, or lyrics. All these kinds of things can easily be marked anywhere on a paper score with a pencil at any time. It's as though there were "comment fields" or linking mechanisms available for every aspect of your piece.

Free Movement Between Operations

The particular process you're using while composing changes about as easily and quickly as the direction of your eyes can change. You find a note by looking at it and make it an edit point by putting your pencil there. You can write with one hand while playing notes with the other, or you can hear all the notes in your head from reading what you wrote even if you can't play them all or at speed, and as you hear them in your head you try edits, changes, or additions in your mind without writing. (What you hear in your head can seem quite real.)

With a pencil, you are always in play, record, insert, delete, merge, and separate modes all at the same time. Not only can you change between these modes instantly, but it's as though there's always another sequence being recorded in your mind, from which you decide what to filter out and how to integrate new material or changes. What's stored in human short term memory is always remembered in sync with the piece as written, with whatever triggered the generation of its contents in the mind. (We night try to simulate it with a default sequence which is always recording whenever we play back a MIDI piece in progress)

Movement Within a Piece

You can move with equal ease from one stave to another or along one part, or in both dimensions at once. Jumping around is as easy as moving continuously. It helps that full page shows a lot of material.

You refer back and forth throughout the score as you work. You might re-read and improvise on something early in the piece to make something new which is related to it, or you might simplify it down. You check on exactly how you did a similar thing before, such as how you spelled some accidentals, how 2 voices fit together, how long it's been since a theme recurred, or to refresh your memory about earlier sections of the piece.


There are by now tens of thousands of people composing, often for the first time, with MIDI sequencer software which assumes that notes are members of various "channels" each of which has the same sound for all it's notes. I am not aware of any sequencers which model the "channel" as a property of the note though it would be easy to do so. The implication of notes belonging to channels is that orchestration is both first, before notes are composed, and fixed for the notes once entered. It's prohibitively difficult to write material and then freely distribute the notes among various instruments afterwards as would be easy if a channel were viewed as a property of a note.

This is an even bigger problem for professional orchestrators and arrangers I've talked with about MIDI than for composers. Sometimes pitch and timbre are composed simultaneously instead of the notes being done first and then orchestrated. This may ultimately be more efficient in some ways because you conceive the pitch-time stuff already properly fitted to the instruments which will play it.

But it has always been common practise to write a piece unorchestrated (in "reduction") and then to orchestrate it in a later compositional pass. (Much Hollywood film scoring is done this way; the composer credit is accompanied by an orchestrator's credit.)

If the association of sounds with notes is done after the pitch-time content is determined, you can concentrate completely on on melody, harmony, and rhythm, writing your "piano reduction" freely and quickly. You can think about instrumental ranges, voicing, tessitura, articulation and orchestral special effects later, when you're clearer on what it is you want to bring out and how you want to present it.

When you orchestrate, one of the commonest things to do is to move material up or down an octave, since accoustic instruments have different timbres in different octaves. Also, melodic lines may be simplified or elaborated during orchestration, since different instruments have different kinds of agility and awkwardness and tend toward different kinds of characteristic figuration. For example, it's hard to trill across register breaks on wind instruments, so you may need to rewrite a line when you decide to use clarinet on it. Orchestration as a phase often involves considerable editing and rewriting, and the filling in of sections still left sketchy.

Doubling one instrument with one or more others is one of the commonest things to do. The simplest way is to write the names of the additional instruments into your first-sketch score right near the notes. Later, when all the creative thinking is done, you'll have time to copy the doublings to the appropriate staves of the full score (usually at unison, 8va, 3rds, or 10ths). Right now you're just getting the ideas down while they're rushing through your mind.

You may put extra instruments on a whole line, or on just a few selected notes of it for emphasis, or break up a line altogether and have it go back and forth between several instruments. You sometimes move notes or melodic fragments back and forth between the voices, trying different orchestrations, switching the staves, octaves, and instruments of notes or passages repeatedly until you're satisfied.

Effectively, you are assigning, distributing, individual notes and musical gestures throughout a group of available sounds.

In general, larger media such as orchestra are written for more simply than smaller media such as string quartets or solo instruments. There may be only 3 real melodic-harmonic voices in an orchestra piece, but built up for the 100 or so players through doubling, highlighting of specific notes in some instrument with other sounds (such as chimes, plucked strings, or percussion), gradual thickening of orchestration, extensive use of dynamics ("mixing"), and the recurrence of the same material in different registers or orchestrations.

Converting Orchestration to MIDI

In addition to each instrument having different timbres in different pitch ranges, each instrument also sounds different at different loudnesses. This makes the explicit composition of dynamics (loudness) absolutely essential when you orchestrate. Unless we can compose loudness exactly and unless our synthesizers are programmed to use these dynamics to modify their timbres appropriately, it is easier and truer just to imagine the orchestration in your head than to try to compose while hearing your delicate blends of sound color completely distorted in timbrally flat electronic simulation.

It's worth noting that a lot of the graphical score symbols for articulation (different kinds of accent marks, etc) are actually symbols for sonic qualities that a computer could use as envelope modifiers, mixing instructions, or in sound selection. The use of such specifiers as "pizzicato", "col legno", or the standard graphical symbols for brass mutes and percussion mallets in a computer-based score should imply their translation to their sonic MIDI equivalents when the computer plays the file. Not to be able to use existing standard musical symbols and language for timbral variation misses the boat in terms of trying to provide compositional fluidity via computers.

This is no less true for the description of abstract electronic sounds. Paper based composition makes it easy to invent new symbols and notations for new sounds and effects. In a medium as conducive to inventiveness as computers, the absence of provision for users' notations, macros, etc. is strange.

Musical Debugging: Tools and Techniques

There is a lot of problem solving in composing. Often you can ponder over why a passage isn't effective and try a million changes to it only to discover much later that something you did several measures back was the culprit and set up the listener's ear with certain unintended expectations, or that the material was fine all along but just happened a little too early or too late or too briefly.

It's not unlike debugging a computer program, and you want the techniques you use while doing it to be as transparent as the most ideal operating system and editor would be while programming - simple, direct, customizable, and with some facility for the grouping of operations into frequently used combinations. Editing operations should be learnable to the point where they become reflexes. You need to be able to concentrate on the musical problem you're solving.

The editing function group should be extensive enough to do everything you need (and probably a lot you don't need) even if it takes a considerable period of learning and study to master it. There should always be more techniques available than any one person can master or use, more left to explore, and enough that different composers can specialize in the subsets of available technique which fit each of their musics best and don't all end up having to work the same way.

Having spent at minimum 10 to 15 years becoming fluent in notation, musical structure, and instrumental technique, composers as a group are more than willing to invest time learning in order to increase their ability and power to create. The idea that no program should take longer than about 20 minutes to learn seems meant to make computers and software sell quicker by making them seem easier and "friendlier". As a design philosophy, it may work against making computers as productive as possible for a specific task once their ownership is assumed.


Working on a piece for a long time (a symphony can take years), you keep going back to the piece, and finding that it's not as you remember it. Your sensativity to and vision of its potential have grown, and your memory of it has changed with them since you left it. You keep refining and changing it to bring what is recorded (notated) up to the standard of (into conformity with) what is now in your memory/imagination as your mind has refined it.

What you've created in the end is a written hypothesis. Your score is a set of instructions to be read by others who will translate your description into sound. Hopefully those sounds will be the same as what you think you wrote, what you imagined.

Copyright ©1988 by Laurie Spiegel. All rights reserved.
back to Laurie Spiegel's writings page | go to Laurie Spiegel's home page