In a second condition of the same experiment, Zatorre and Samson presented patients with the same task, but with the retention interval filled by interfering tones. The right temporal lobe patients were impaired in this version of the task, regardless of whether Heschl's gyrus was involved. A group of patients with right frontal lobectomies were similarly impaired. Although the laterality of the frontal effect could not be assessed, the right temporal lobe effect was clear-cut and indicated a critical role in auditory working memory for secondary auditory cortex in the right hemisphere.
David Perry carried out a study of ear-of-presenta-tion asymmetry (a small behavioral effect presumed to result from predominance of the contralateral ascending auditory pathway) for the subsequent recall of melodies by experienced pianists who replayed them on a computer-interfaced keyboard. The interval between the end of the melody and the beginning of attempts at keyboard reproduction was recorded. During this interval, most subjects reported mentally resinging and rehearsing their memory of the melody before beginning to play. A right ear asymmetry was observed in sequence recall accuracy. Furthermore, an asymmetry was observed in this delay interval that was positively correlated with that in accuracy, such that the delay between stimulus and response (lasting several seconds on average) was longer after presentation to the preferred ear.
This asymmetry in an interval associated with internal rehearsal of just heard information by ''inner singing'' suggested the possibility of an asymmetry in the neural substrate for auditory-tonal working memory, conceived of by analogy to Alan Baddeley's ''phonological loop.'' Just as inner speech can serve to refresh the contents of a specialized auditory-verbal store (as when remembering a just heard phone number), so too inner singing (based on vocal fundamental frequency control rather than articulatory control processes) can refresh the contents of a specialized tonal store through the functioning of the ''tonal loop'' (Fig. 3). It is possible that for instrumentalists ''inner playing'' might also be utilized as an internal rehearsal strategy, although whether mental
Figure 3 Tonal working memory. Tonal loop model of auditory-tonal working memory based on the phonological loop model of Alan Baddeley for auditory-verbal working memory. Music or nonmusical tonal information is first held in a time-limited tonal store, whose proposed neural substrate includes auditory cortex and mid-ventrolateral (VL) frontal cortex. The contents of this store can be refreshed and held on-line through inner rehearsal (e.g., ''inner singing'') based on vocal fundamental frequency control processes (rather than on articulatory control processes as proposed for inner speech by Baddeley). Inner or imagined singing is proposed to depend on most of the areas active during actual singing. The supplementary motor area (SMA) is particularly strongly associated with imagined singing. Finally, the contents of working memory may be consciously monitored, e.g., to update an ongoing record of events or actions, an executive working memory function for which the mid-dorsolateral frontal cortex is critical (adapted with permission from Marin and Perry, 1999).
motoric rehearsal and its associated auditory image can be fully dissociated from inner singing remains an empirical question.
Based on their studies of anatomical projections in the macaque brain between the frontal lobe and posterior sensory association cortex, Deepak Pandya and associates hypothesized that retention of auditory information might involve specific temporal-frontal projections, just as Patricia Goldman and colleagues had proposed for visuospatial retention and parietal-frontal projections.
Based on a long series of behavioral lesion analyses in nonhuman primates and further analyses of posterior association cortex-frontal projections, one of Pandya's associates, Michael Petrides, has articulated a hierarchical theory of frontal contributions to mnemonic processing. Petrides proposes that each major sensory modality represented in the posterior association cortices (auditory, visual-spatial, visual-object, and somatosensory) projects, with some degree of topographical specificity, to each of several frontal lobe cortical regions critical for distinct executive functions. As mentioned in the discussion of absolute pitch, the posterior dorsolateral frontal cortex is proposed to play a role in conditional associative learning, one higher order frontal lobe memory function.
First in the hierarchy of proposed frontal lobe contributions, ventrolateral frontal cortex is hypothesized to be critical for the repetition, selection, comparison, and judgment of stimuli held in working memory. Auditory cortex may be sufficient for the passive retention of tonal information, but ventrolat-eral frontal cortex may be required for any form of more active maintenance. Furthermore, as suggested by the patient study of Zatorre and Samson and the asymmetry in the aforementioned rehearsal intervals, there may be a complementary right hemispheric asymmetry in frontal as well as temporal cortical contributions to pitch processing.
In a PET experiment inspired in part by the lesion analysis of pitch retention described previously, Za-torre presented eight-note melodies and instructed subjects to simply listen or to compare the pitch of either the first and the second notes or the first and final notes. The melody perception condition was discussed previously. When the two-note condition was contrasted to melody perception, frontal activation was observed in right frontopolar cortex. For comparison of the first-last pitch vs perception, additional increases were observed bilaterally in both ventrolateral and mid-dorsolateral frontal cortex, and there was also an increase in the right middle temporal gyrus.
Thus, only in the first-last condition was activation seen in the frontal regions predicted by Petrides' model of frontal lobe contributions to working memory. The two-note comparison, despite the fact that it requires judgment of pitch direction, was designed to make minimal demands on active maintenance of pitch information, which may not have been sufficient, particularly in contrast to melody perception, to result in measurable ventrolateral frontal activation. However, the first-last condition requires maintenance of the first pitch across interfering tones and then comparison to the last pitch. Since all the tones in the melody are related, it may not be possible to hold one tone in mind and simply ignore the rest. If subjects hold the entire sequence briefly in mind and compare the first and last tones mentally, then they are monitoring the contents of their working memory store. In a sense, this task thus resembles the n-back tasks, in which subjects are asked to compare the current item to one n items back, which have been demonstrated to activate mid-dorsolateral frontal cortex just as the self-ordered tasks do. Therefore, activation of both stages of Petrides' model by the first-last eight-note melody task is consistent with the task demands.
In a PET study designed to isolate the ''monitoring'' component of auditory-tonal working memory, Perry, Petrides, and Zatorre presented subjects with two tasks requiring them to monitor the contents of working memory. In a self-ordered condition, subjects were asked to sing six-tone sequences composed of two pitches by choosing each at random until a sequence with an equal number of both pitches was completed. In an externally ordered condition, they listened to five-tone sequences and sang the final pitch such that the sequences again had an equivalent number of high and low pitches. Both conditions were contrasted to a sensorimotor control involving singing a single pitch at the same rate. Increases were observed in both mid-dorsolateral and ventrolateral frontal cortex (Fig. 4), consistent with Petrides' two-level hypothesis of frontal lobe contributions to working memory that emphasizes the role of projections between mid-dorsolateral and mid-dorsolateral frontal cortex in monitoring. The activations within mid-dorsolateral frontal cortex were both greater in the right hemisphere, whereas no consistent asymmetry was observed within ventrolateral frontal cortex.
The portion of ventrolateral frontal cortex activated was actually deep within the upper bank of the horizontal ramus of the Sylvian fissure, part of cortical area 45 according to recent cytoarchitectonic analyses of this region of human frontal cortex by Petrides and
Si self-ordered pitches vs. singing /^externally-ordered pitches vs. singing III first-last pitch direction vs. melody perception II pitch direction in heard songs vs. reading ifi pitch direction in imagined songs vs. reading
Was this article helpful?