|
[Page numbers reference text version of document, located at above link]
p.2 How do people improvise? How is improvisational skill learned and taught? These questions are the subject of this paper. They are difficult questions
p.2 In a previous article (Pressing 1984a) I summarized a number of general properties of the improvisation process... In this article a much more explicit cognitive formulation is presented, the first proper (though by no means necessarily correct) theory of improvised behaviour in music.
p.4 The starting point for nearly all the existing theories is the 3-stage information-processing model mentioned earlier, based on sensory input, cognitive processing, and motor output. To this must be added the notion of feedback (auditory, visual, tactile, or proprioceptive).
p.5 As discussed by Shaffer (1980), a plan is an abstract homomorphism representing the essential structure of the performance and allowing finer details to be generated or located as they are needed during execution.
Other related theories include Allport's proposal of a system of condition-action units which are links between sensory calling patterns and categories of action (Allport 1980). Also related are adjustable control or description structures from artificial intelligence such as frames and scripts
p.5 But as of this writing there seems only one alternative in the area of motor behaviour. This is the organisational invariant approach of Turvey, Kugler, Kelso and others (Turvey 1977, Kugler, Kelso and Turvey 1980, cf Kelso 1982 for further references). This approach draws on two sources: the ecological perspective of Gibson (1966, 1979) and the dissipative structure model of non-equilibrium thermodynamics (Prigogine 1967, Prigogine and Nicolis 1971, Prigogine, Nicolis, Herman, and Lam 1975). Essentially the theory de-emphasises notions of cognitive process and control, replacing them with, in so far as is possible, 'organisation invariants'. These organisational invariants are characteristic constraint structures that allow the emergence of specific spatial relationships and dynamic processes in the behaviour of non-linear systems when the parameters controlling these systems fall in certain critical ranges... Organisational invariant theory seems also likely to apply primarily to the dynamics of motor program execution
p.6 Feedback can also be considered to operate over different time scales. Thus short term feedback guides ongoing movements, while longer term feedback is used in decision-making and response selection. Still longer term feedback exists in the form of knowledge of results (KR) for skills where external evaluation is present or result perception is not sufficiently precise or immediate.
p.7 The idea of preparation is very important for improvisation, where real-time cognitive processing is often pushed up near its attentional limits... For improvised performance that aims at artistic presentation, where discrepancies between intention and result must be kept within strict bounds, practice must attempt to explore the full range of possible motor actions and... effects, to enable both finer control and the internal modelling of discrepancies and correction procedures, including feedforward.
p.9 skilled learning... By the time advanced or expert stages have been reached, the performer has become highly attuned to subtle perceptual information and has available a vast array of finely timed and tunable motor programs. This results in the qualities of efficiency, fluency, flexibility, and expressiveness. All motor organisation functions can be handled automatically (without conscious attention) and the performer attends almost exclusively to a higher level of emergent expressive control parameters
p.16-17 A second and wide-ranging [psychological] approach [to intuition] is found in the recent work by Bastick (1982), which includes a search of over 2.5 million sources for common properties underlying intuition. After the identification and detailed analysis of some 20 of these properties, Bastick ends up describing intuition as a combinatorial process operating over preexisting connections among elements of different 'emotional sets'. These emotional sets apparently contain encodings, often redundant, of many different life events (intellectual activities, movement, emotion, etc.). By giving strong emphasis to the role of dynamics, bodily experience, and the maximizing of redundancy in encoding, and by a series of suggestive diagrams of intuitive processing, Bastick seems to be on an important track parallel to emerging ideas of improvisation.
p.18 Search techniques come in several variants, including depth-first, breadth-first and best-first. All use a generate-and- test procedure to find solutions to a problem. Clearly there are possible connections with improvisation. Generate-and-test could be applied to learning to improvise
p.18 The need of the improviser is for a good solution, not the best, for there is probably no single 'best' solution, and even if there were, it would take too long to find it.
p.20 Any theory of improvisation must explain three things: how people improvise; how people learn improvisational skill; and the origin of novel behaviour. It must also be consistent with the numerous recurring themes reviewed above. The model given here seems to satisfy these conditions.
p.20 The first part of this model describes the process of improvisation. It begins with the observation that any improvisation may be partitioned up into a sequence of non-overlapping sections... Let each of these sections contain a number of musical events (e.g. a group of notes, one or more gestures or phrases, etc.) and be called an event cluster Ei. Then the improvisation I is simply an ordered union of all these event clusters. Formally,
I = E1, E2,....EN (1)
p.20 Our first major assumption is that every improvisation is actually generated by triggers at specific time points t1, t2, ....tn that instigate the movement patterns appropriate to effect intended musical actions. Each time point is thus the point at which decided action begins to be executed. Note that it is schemas for action that are triggered, not precise movement details, and subsequent motor fine-tuning based on feedback processes goes on after each time point. Most often time points will have clear musical correlates, with adjacent event clusters being set off from each other by local musical boundary criteria: pauses, phrase junctures, cadences, grouping by sequence, etc; but this need not always be the case.
p.20 With this interpretation, equation 1 is a unique specification of the timing of central decision-making made by the improviser. The improvisation may then be viewed as a series of 'situations', where the (i + 1) th situation is confined primarily to the time interval (ti, ti+1) and entails the generation of the cluster Ei+l on the basis of the previous events E1, E2,....Ei = E i, the referent R (if one exists), a set of current goals, and long-term memory M. The referent R is an underlying piece- specific guide or scheme used by the musician to facilitate the generation of improvised behaviour (Pressing 1984a). The process of event cluster generation may then be written
p.21-23 1. Ei is triggered and executed... 2. Each aspect of Ei may be decomposed into three types of analytical representation: objects, features, and processes... Let this decomposition... be represented by three variable-dimension arrays O, F, and P, and assume that they represent all information about Ei needed by the improviser in decision-making... 3. The structures of the three types of arrays are as follows... 4. Production of Ei+1 occurs primarily on the basis of long term factors... stylistic norms and on-going musical and movement processes.. and by evaluation of effects 5. Associative generation is based on either similarity or contrast... 6. Interrupt generation is based on the resetting of all or a significant number of the strong array components without regard to their values in the current event cluster Ei... 7. The choice between association and interrupt generation may be formally modelled by a time-dependent tolerance level for repetition, L(t)... 8. Once Oi+1, Fi+1, and Pi+1 are selected for all relevant aspects, tunable cognitive and motor subprograms are set in motion that generate, on the basis of these higher constraints and current motor positions, a specific action design... These, then, are the salient features of the model in outline.
p.27 The specific cognitive changes that allow these properties to develop in improvised musical behaviour are considered to be:
- an increase in the memory store of objects, features and processes - in musical, acoustic, motor (and other) aspects
- an increase in accessibility of this memory store due to the build-up of redundant relationships between its constituents and the aggregation of these constituents into larger cognitive assemblies
- an increasingly refined attunement to subtle and contextually relevant perceptual information.
p.28 The refinement of improvisational skill must depend partly on increasing the efficiency of perceptual processing to allow the inclusion of more and better-selected information in the improviser's decision-making procedures. The need for this efficiency is imposed by every performer's more or less limited individual capacity, per unit time, to process novel sensory input.
p.28-29 This paper has attempted to illuminate the process of musical improvisation by first examining the modelling tools available from a number of different disciplines. Based on this examination, a cognitive model has then been presented for the process itself, followed by a brief discussion of its relation to improvisational skill acquisition... The model seems to be specific enough to allow its use as a basis for the design of 'improvising' computer programs.
|
