processing: the encoding and recognition of melodies. Melodic patterns, in turn, may be
investigated at many different levels of analysis, from the basic interval relationships between
tones to the semantic networks that are important for long-term representations of familiar
tunes, and the output mechanisms necessary for singing or playing them. Here, we shall con-
centrate on the way in which auditory cortical systems are engaged in perception of melodies.
One critical consideration in this respect is that all melodic perception (indeed, all auditory
perception) involves working memory mechanisms. Since sounds unfold over time, in order
to compute the relationships between successive events, an on-line retention system is
evidently necessary. In the case of even a brief melody, this would mean that relationships
between tones have to be computed and maintained over periods ranging from seconds to
minutes in order to achieve a relatively stable and coherent internal representation.
Many studies of melodic perception with brain-lesioned individuals have demonstrated
that damage to the right STG often leads to behavioural deficits in discrimination
tasks,7,14–^16 although most of these studies have also noted that damage to the equivalent
structures on the left may also lead to some degree of impairment. Indeed, Peretz et al.^17
(see also Chapter 13, this volume) have noted that patients with bilateral lesions that
involve portions of the STG and frontal cortical structures demonstrate much more severe
disturbances in melodic tasks than most patients with unilateral damage. In fact, in each of
the cases that Peretz and her colleagues have studied, the patients did not appear to show
the specific musical deficit until after a second lesion had occurred, implying that unilat-
eral damage had had at most only a mild effect not noted by the patients or their families.
This point is important in that it suggests that the putative hemispheric specializations that
are one theme of this chapter are likely to be relative rather than absolute.
The relative importance of right-hemisphere STG mechanisms has been demonstrated
specifically for tonal working memory in two studies from our laboratory. In one,^18 we
adapted a paradigm first developed by Deutsch^19 in which a target and a comparison tone
are to be compared for pitch; the tones are either separated by a brief silent interval or by
a series of interfering random tones. Deutsch had shown that memory for tones was rela-
tively specific because it was not disrupted by other sounds, but only by other tones
(see also Ref. 20). When this tonal working-memory task was administered to temporal-
lobectomy patients, performance was significantly worse following damage to the right
than to the left temporal cortex for the interference task. On the other hand, unlike the
studies of basic pitch processing discussed above, there was no difference in degree of
behavioural deficit as a function of encroachment onto HG. Thus, areas of auditory cortex
outside of the primary zone are most likely to be involved in this function. Finally, it was
also noted that right frontal cortical excisions produced an impairment similar to that seen
in the right temporal patients, a point that is taken up below.
To follow up on the question of tonal working memory function, Zatorre et al.^21 used
PET to test 12 volunteer nonmusicians with unfamiliar melodies and noise bursts that had
been constructed so as to approximate the acoustic characteristics of the melodies. Four
conditions were run; subjects were asked to (1) listen to the noise bursts; (2) listen to the
melodies without explicit instruction; (3) listen to the same melodies and compare the
pitch of the first and second notes (two-note condition); and (4) listen to the melodies and
compare the pitch of the first and last notes (first/last condition).
237