I am peripherally involved (as a mentor) with a proposal to create a newsgroup for music theory (namely rec.music.theory), and so want to open this column with a rather curious remark which came up during the debate over the formation of that discussion group. You can find the debate on Dejanews [old link discarded 04/04/10], if you're inclined to read such things, but the remark went as follows: "rec.music.theory is superfluous, because it already exists, as rec.music.compose." I don't expect that point of view to have a real effect on the discussion (which is mostly revolving around tedious technical issues), but I thought it was a peculiar notion and worth a little discussion. The contention is apparently that theory & composition are synonymous, or rather that theory is an aspect of composition. This idea is clearly incorrect, but also seems like a good starting point for a more general discussion.
It is a trivial fact that composers & theorists are almost always different people. Of course, one studies the other, but how much interaction is there? Theorists could not exist without composers — or improvisatory performers, to take the more general case. On the other hand, composers must take the notation itself as a first theoretical construct. Composers need to understand notation, both to be able to write it and to realize how their work will be apprehended. In the modernist & postmodernist schools there is a distinct trend toward innovating notation on a per-composer basis, and often investing substantial energy into describing the nuances of the notation to the finest level of detail, a practice which meets its antithesis (yet within the context of innovative notation) in aleatoric techniques. There is a question raised by all of this, namely whether the conception of the notation actually predates that of the music. Generally speaking, even in the most theoretically-oriented composers, there must be a give & take... the composer thinks of the "sort" of music he'd like to write, and how it does not fit very well in the existing notation, conceives the notation, and then conceives the music in more detail. I think the latter ordering is significant, and do believe it is a general feature of contemporary notation related innovation, at least of the most prominent sort. While this may raise questions about whether the "sonic surface" or the theoretical underpinning is the primary act of creation (a basic issue in musical expression discussed previously), I do not believe one can make strong causal statements in that direction. In other words, the most influential works usually have a strong degree of simultaneity.
What we do see is a far larger amount of innovation in notation than in the past. While ideally one can conceive a piece of music, and then write it out, there is always the danger that some level of detail will be lost or readily misinterpreted. There is a natural evolution to notation in that sense. The most sudden and influential changes in Western notation still date from the late medieval era (the adoption of mensural notation in the early 13th century and the extra rhythmic possibilities of Ars Nova notation adopted in the early 14th century), with more gradual developments from there. A tangent which is perhaps relevant to this topic is the implicit belief among some HIP performers that composers' musical ideas are constrained by their available notation. The belief is expressed in the "hard" articulation of notated rhythms in medieval music, leaving aside any possibility for a fluidity which is impossible to notate. This is highly significant, as it makes some rather firm meta-decisions regarding performance, and to my ear always leaves the music sounding distinctly odd & forced. In other words, I believe the notation should not be interpreted to imply things whose negation cannot possibly be specified by available notation. I am likewise disinclined to advocate notational innovation which over-specifies. It is quite a quandary I suppose, especially for hypothetical performances 500 years in the future, but I firmly believe the old aphorism about squeezing something too tightly forcing it out of one's grasp.
Beyond notation, which is integral to composition, there is a much larger body of descriptive theory. For the theoretician, this provides the framework in which to "analyze" a piece of music, but more basically it provides us with simple language with which to describe or discuss a work. The basic language of tonality is theoretical in construction, as indeed there were works to which a key signature could be applied before there were key signatures. Likewise, there were composers moving to the subdominant before there was such a term, etc. In that sense, theory is always developed a posteriori and is not necessary for composition. One need not know that one is modulating to the subdominant to do it, or even to do it well, but the language can increase one's perception of this technique elsewhere and consequently one's sophistication. These theoretical concepts also gain power over time, as the invention of a term will naturally stiffen the underlying practice. So today we have composers thinking explicitly in these terms, and feeling bound by them in ways the composers whose music originated them did not. But let us not go too far... this is not a tragedy, any more than following a tradition is a tragedy, and the words are simply one embodiment of the tradition. This leaves theory & practice always at odds with each other, to the extent that theory attempts to catch up to composition and even dictate to it, making the original remark even more curious. It can only be interpreted as a prescription for composition, and in fact one which the vast majority of great composers did not follow.
A more interesting question concerns the degree to which musical innovation has been motivated by theoretical discourse as opposed to purely "sonic" ideas. In the major technical shifts from medieval to Renaissance to Baroque, the impulses were clearly sonic, but buoyed almost simultaneously by theory. While sufficient documentation does not exist to be certain, the use of thirds at cadences was apparently carried out strictly for "musical" reasons, and only subsequently validated by the designation of the third as a new consonance. The development of basso continuo is better documented, and indeed it was Monteverdi himself who provided much of the rhetorical basis for its practice, something which helped establish him as the leading composer of the era. But make no mistake, continuo was motivated sonically, by a basic desire to set a virtuosic soprano voice in sharper relief, and used in principle by composers before Monteverdi. After both of these shifts, the practice was slowly tamed by theory, respectively in the severe dissonance treatment of the Palestrina school and the key signatures of the later Baroque. This basic dynamic finds its reflection in Schoenberg's dodecaphony, as a theoretical approach which was developed a priori to restrict a free atonality which had arisen spontaneously. This signals something of a shift in the relationship between theory & practice, which I believe is motivated partly by the proliferation of academic certification and the relative ease of evaluating theoretical theses. Nonetheless, these relationships remain more or less typical, mediated primarily by the fact that music demands (or at least has demanded) two acts of creation. I will stop here, but do plan to take up this topic with respect to other traditions at a later date, with the hope of illuminating how dual creation can affect the relationship of theory to practice.