1.5 The limits of memory
In unwritten music, a factor which places a constraint on the number of fixed elements – the degree of detail specified by any model – is memory. Whatever is fixed must be memorised; as a matter of necessity, therefore, performers in these traditions have evolved strategies which limit the load placed on their memories. Here is Nettl again:
Dividing music into elements, I hypothesise the need for some of these to remain simple, repetitive, stable, so that others may vary. There is probably some point beyond which it is impossible for any sizeable population of musicians to remember material … Recurring events or sign posts such as motifs or rhythmic patterns, conciseness of form, brevity, or systematic variation may, as it were, hold an aurally transmitted piece intact.
(Nettl, 1983, p. 192)
We shouldn't underestimate the ability of musicians to memorise pieces of considerable length and complexity. In the West for instance, pianists often perform lengthy concerts entirely from memory. It should not be surprising that musicians who never use notation may tend to develop even greater powers of memorisation than those who do. Nevertheless, there is always a limit. Indian maestros who perform without notation, for three hours or more each night (and the music different each time), are not reproducing memorised performances. So how do they do it?
One way of answering this question is Nettl's: certain aspects of music remain stable or repetitive so that others can vary. Another, complementary, approach is to say that musicians have to learn and memorise two types of information: the models, and the ways of turning those models into music. The jazz soloist, for instance, learns a stock of songs and compositions, as well as ways of generating solos appropriate for those pieces and his/her instrument. Similarly, in any unwritten music tradition, performers learn models for music-making (scales, modes, melodies; rhythms, metres; harmonies; song texts; and so on), and they also learn how to turn those models into acceptable performances.