The corresponding modal composites represent chord functions. Tonal compo-
sites over longer durations represent keys, and modal composites represent
modes. If metrical bias is added, say in the form of pulsing activation (Gjer-
dingen, 1989a), then a tonal or modal composite would encode an interaction
between tonal/modal and metrical information.
If persistence is large and activation is phasic, a tonal composite roughly
represents the probability distribution of pitch classes in a segment of music.
Krumhansl (1990) has shown that distributions of pitch classes are strongly
correlated with empirically determined key profiles of Krumhansl and Kessler
(1982). A tonal composite with large persistence is thus a representation of
the hierarchy of prominence or stability of pitch classes as determined by their
frequencyofoccurrenceinasegmentofmusic.Withbothtonicandphasic
activation, the tonal composite would represent something between the distri-
bution of occurrences of pitch classes and the distribution of durations of pitch
classes, both of which are highly correlated with key profiles (Krumhansl,
1990). What Parncutt and Huron (1993) have attributed to echoic memory is a
tonal composite of pitch class; their demonstration that some of Krumhansl’s
probe-tone results can be modeled by such a memory is support for the exis-
tence of tonal composites as representations.
Although the distribution of pitch classes in a piece of music has a substantial
influence on our perception of the relative stability of pitch classes, long-term
representations of structural regularities (sometimes referred to asschemas)
also exert an influence. In a cross-cultural study by Castellano, Bharucha, and
Krumhansl (1984), Western and Indian subjects heard a rendition of a North
Indianra ̄gand then judged how well a probe tone fit with the preceding seg-
ment. Probe-tone ratings were obtained for all 12 pitch classes following each
of 10ra ̄gs. For both Western and Indian listeners, the probe-tone ratings were
highly correlated with the distribution of total durations of pitch classes in the
segment, consistent with a temporal composite representation. However, the
Indian subjects showed an influence of prior exposure to the underlying scale
ortha ̄t, whereas the Western subjects did not. In a multiple regression analysis,
a significant contribution to the regression was made by the distribution of
durations for both groups of subjects, but only for the Indian subjects was a
significant contribution made by the membership of pitch classes in the under-
lyingtha ̄t. This latter variable—tha ̄tmembership—was assessed by using a 12-
element vector of binary elements representing the presence or absence of each
pitch class. The contribution of this vector to the multiple regression for the
Indian subjects suggests that whereas the responses of the Western subjects
were based entirely on the distribution of pitch classes in the most recently
heard segment, the responses of the Indian subjects were based also on prior
knowledge of which pitch classes are typically present (i.e., the hierarchy of
stability of tones was internalized).
This prior knowledge is implicit, schematic, and acts like a cultural filter.
Implicit knowledge can be studied by priming. In a priming task, a target
stimulus is presented following a context (prime stimulus), and subjects are
instructed to make a designated true/false decision about the target. If the
speed and accuracy with which the decision is made are greater following
contextC 1 than following contextC 2 ,thenC 1 primesthetargetmorethanC 2
464 Jamshed J. Bharucha