As indicated in Formal Conception: Its Elements, the idea of traversing a parallel landscape of material from several perspectives began on the basis of very elementary sketches (cf. Exs. 1, 2, 12, and 13) and evolved over a period of years. The first complete, formalized plan can be seen in Example 41, and is dated 20 February 1998. It provides detailed specification for component durations and also defines the parallelisms between the S and D versions. The last version of the overall plan was done on 29 January 2002, and is significantly different in regard not only to the level of detail (which is yet more abundant), but in the degree to which the inevitable generality of the original is revealed. The first version of the plan, for example, proposed that a repeat of Theme 2 would occur midway through Other and that the COMB2/4 would continue during Theme 3. Its S version included a complete COMB3/5 and failed to anticipate the participation of the Ensemble in a supporting role during the piano version of TR2 to 4. Throughout the process of envisioning and then actually composing the two parts of The Angel of Death, I used the various versions of the overall plan as guides. And not only proportions, coexistences and alternations, but also dramatic succession, sectional as well as thematic character, and — when the D part was to be extrapolated from the existing S one — the specific assignment of source identity for each of the new part's subsections.
I would certainly not expect that the early plans for a complex work could remain fixed over time, unchallenged by the developing reality of the music that arises out of them. On the other hand, if, in a sense, "anything goes", and the durations, proportionalities, and identities are not adhered to after their initial positing, then there would be no reason to make such preliminary specifications, to plan for architectonic consistency. Evidently, there needs to be a reasonable line between disregarding the normatives established in the preliminary stages of a work's evolution, on the one hand, and, on the other, an inflexible commitment to fulfill their every letter without regard to emerging realizations as composition proceeds. This interplay between the proposed and the realized is more complex the more elaborate the work (and its purposes). The duration of the compositional process is also a factor in reconsiderations. As it happened, Ircam was closed for renovation during the 1999-2000 season, and this caused a delay of a year for the Angel Project. Thus, the plans themselves were, for me, both more essential, and yet less stable than might normally be the case.
As has been indicated earlier, the original conception of the ensemble as being comprised of four familial layers, could not be sustained. In considering why this was so, I note three sources of difficulty. Firstly, the generality of the pianistic medium — where range, breath and bow length are not constraining factors — invites music likely to exceed the boundaries of the ensemble instruments when considered as individual resources. What one instrument cannot do must be redistributed to several, creating ensemble coordination issues. Secondly, the piano has considerable registral flexibility — there is an absence of the inertia that, for example, resists an upward leap in the wind instruments. Thus, the necessarily concentrated and highly characterized thematic material conceived in reference to a pianistic medium provokes the creation of elements that require at one moment agility in the low register, at the next great force in the high register, and so on.
The capacities needed by the ensemble instruments assigned to carry the pianistic figurations change continuously with the intricate local variation of musical character in the themes. Different instruments (from different families) must continually be assigned to meet the changing needs. Thirdly, it became clear that, with the exception of Theme 3, whose character is explicitly and unavoidably restricted by its nature, a high degree of heterogeneity is almost certain to arise if one is seeking to define idiosyncratic miniatures. Perceptual distinctiveness is the sine qua non in this situation. But the third theme, while constrained in range, density, and force, requires at the same time a flexibility and agility that, for example, the brass family could not hope to manage.
At an early stage in our collaboration, my psychologist colleagues asked whether I could compose a section of material that was "wrong". In other words, could I work within the boundaries I had determined in such a way that the music produced was unacceptable. This was a startling request, and my response was that I would not know how to go about doing such a thing honestly (by other than a self-conscious contrivance). But, while I was working on the textural designs for the various themes, I realized, one day that the design I had created for Theme 2 was unacceptable. If one examines the differences in the design shown in Ex. 48 with that in Ex. 49, it is obvious that they are significant. In fact, they constitute an "error" in the first version that was corrected in the second.
Although there is the same number of basic chordal/harmonic units in the two versions (11), the first seems to have been generated under the assumption that each of the seven subsections could be thought of as essentially self-contained: in other words, that the relation of the elements of this theme across its subsections was not crucial. One can see that the appearances of chord "d" (which appears once in subsection 2, once at the end of subsection 3 and three times in subsection 5) are distributed such that it brackets the central core. Still, too little attention has been given to how the harmonic resources are handled: with regard to concentration or breadth of distribution, timing periodicity, and bracketing (as a guarantor of unity across the entire theme). The scalar passagework resource in the second version is also treated with more intricacy and flexibility. Thus, while this may not meet the needs of their original request, my colleagues can find here, as well as in the continuing evolution of the governing overall plan (it had at least 4 versions), instances in which the first version of an aspect of the work turned out, after consideration to be unacceptable, requiring a new manifestation.
The Subsequent Elements (Treatment and Adjustments) section above discusses changes with respect both to COMB3/5 (partially mitigated, thereafter by the content of S11 which was derived from the same two source themes) and also COMB1/2/3, which was completely replaced. This need not be reiterated here. The addition of the piano Epilog is another factor involving not omission or alteration, but the recognition of an incompleteness in the original overall concept for the piece: two parts stemming from differing views of the same material terrain with the addition of a subsidiary, but formally propulsive and culminating influence from the computer. As matters evolved, the computer emerged more prominently (even dominantly) than was originally foreseen, both between the work's two primary parts and after the close of the second.
I was uncomfortable with the prospect of the computer trumping the final instrumental section, Theme 5. This would be all the more true if D was performed second and the final theme was thus represented primarily by the solo pianist. It was unacceptable to have the computer close the piece while the instrumentalists sit uninvolved, and this dissatisfaction intensifies if the computer's ending were particularly forceful. What was needed was a lyrical instrumental section, a powerful computer summation, and then a return to the live dimension — something in a reflective mood but brief and familiar enough so as not to give the impression of embarking on new terrain.
It was surprising to find that one of the seemingly most attractive and promising ideas for a computer image, "Staggered Convergences", proved very difficult to realize effectively, and, in addition, seemed destined to fatally mar the effect of Theme 3. I decided, after considerable effort, that this image could not be effectively realized in the available time, so it was omitted. The other 10 images, as a result of dynamic balancing and the effects of spatialization, would interact well, I thought, with the already completed S materials. There were three sorts of relationships between the computer images and the live instrumental components: exclusivity/dominance (D10, D1, S7), intensification (D5, S4, S8), and transparency (D6, D2, S11, D9). The pitch component of each portion of the S part was noted, and possible conflicts or unwanted dissonance were mitigated by adjustments.
Finally, thought was given to the possibility that some elements of the computer's role could be done interactively, with real-time processing. However, tests underscored the fact that, while some approaches were reliable and of high audio quality (e.g., recording and playing back over various time intervals, from fractions of a second to very large values), real-time instantiations of the more powerful transformations such as filtering, transposition, and analysis and resynthesis, were not of sufficiently high quality to be viable along side the pre-computed playback of such transformations.