Functional Mapping of the Primate Auditory System

See allHide authors and affiliations

Science  24 Jan 2003:
Vol. 299, Issue 5606, pp. 568-572
DOI: 10.1126/science.1078900


Cerebral auditory areas were delineated in the awake, passively listening, rhesus monkey by comparing the rates of glucose utilization in an intact hemisphere and in an acoustically isolated contralateral hemisphere of the same animal. The auditory system defined in this way occupied large portions of cerebral tissue, an extent probably second only to that of the visual system. Cortically, the activated areas included the entire superior temporal gyrus and large portions of the parietal, prefrontal, and limbic lobes. Several auditory areas overlapped with previously identified visual areas, suggesting that the auditory system, like the visual system, contains separate pathways for processing stimulus quality, location, and motion.

The full extent of the monkey's auditory system remains unknown. The principal cortical areas that have been identified by neuronal recording occupy approximately the caudal two-thirds of the superior temporal gyrus (STG) (1–3), although small sectors have also been found in other cortical and subcortical areas, including, for example, the inferior parietal (4,5) and the lateral prefrontal cortices (6–8), and the superior colliculus (9, 10). Yet this cannot be the full extent of the auditory system, inasmuch as the caudal STG sends dense projections to the rostral STG as well as to numerous other cortical regions outside the sectors identified thus far (1,11–13).

In an earlier study (14), we mapped the monkey's entire visual system by comparing glucose utilization in an intact (“seeing”) hemisphere and a visually deafferented (“blind”) hemisphere of an awake animal passively viewing visual stimuli (15, 16). Here, we used the same strategy in an attempt to map the entire auditory system of the monkey.

Three chair-trained, adult rhesus monkeys (Macaca mulatta) weighing 4.0 to 4.8 kg were prepared with unilateral ablations of the inferior colliculus and transection of the forebrain and tectal commissures to yield an intact (“hearing”) and an acoustically isolated (“deaf”) hemisphere in each animal. Five to 10 weeks later, the 2-[14C]deoxyglucose (2-DG) method (17, 18) was applied while the animals listened passively to a variety of acoustic stimuli (16) (fig. S1). The 2-DG autoradiographs generated by this method revealed hemispheric asymmetries in local cerebral glucose utilization (LCGU), thereby indicating which areas had participated in auditory processing during the experiment (Fig. 1) (16, 19, 20).

Figure 1

2-DG autoradiographs from a monkey prepared with a right inferior colliculus ablation (section –2) combined with forebrain and tectal commissurotomies, showing relative rates of glucose utilization in pseudocolor (see color bar). The left hemisphere is on the left. Numerals refer to the coronal levels of the sections in millimeters anterior (+) or posterior (–) to the interaural plane. Comparisons between the left, intact, hemisphere and the right, deafferented, hemisphere reveal the asymmetries indexing auditory regions, many of which are labeled. Scale bar, 5 mm. Cortical abbreviations are the cytoarchitectonic designations of Bonin and Bailey (24). Subcortical abbreviations are as follows: Alat, lateral nucleus of the amygdala; IC, inferior colliculus; MGBsg, medial geniculate body, including the suprageniculate nucleus; PUTv, ventral portion of the putamen; SCm, midportion of the superior colliculus.

Overall, the results demonstrated that cortical auditory tissue occupies the entire extent of the STG from the caudal tip of the lateral sulcus to the rostral tip of the temporal pole, and from the fundus of the lateral sulcus to the fundus of the superior temporal sulcus, as well as parts of the limbic cortex and large portions of both the inferior parietal lobule and the prefrontal cortex (Table 1 and Fig. 2). Subcortical structures shown to be involved in auditory processing include not only the tectum and medial geniculate body but also parts of the amygdala and neostriatum.

Figure 2

Schematic summary of cortical areas related to the processing of auditory, auditory plus visual, and visual stimuli (red, pink, and gray areas, respectively), based on a comparison of the present results with those obtained in the earlier 2-DG visual experiment (14). Standard coronal sections through the right hemisphere are at levels indicated by the vertical lines on the lateral surface view (upper left). Numerals refer to number of millimeters anterior (+) or posterior (–) to the interaural plane. The frontal pole (FDE), surrounded by a dashed line, was not assessed for LCGU uptake in the 2-DG visual experiment. Boundaries were determined by visual inspection followed by quantitative examination of the autoradiographs for the individual cases, and the boundary positions were averaged across the cases within each study. Cortical abbreviations are the cytoarchitectonic designations of Bonin and Bailey (24). Sulcal abbreviations are as follows: ai, inferior limb of arcuate; as, superior limb of arcuate; ca, calcarine; ce, central; ci, cingulate; ip, intraparietal; l, lunate; la, lateral; oi, inferior occipital; orm, medial orbital; ot, occipitotemporal; p, principal; rh, rhinal; tma, anterior middle temporal; tmp, posterior middle temporal; and ts, superior temporal.

Table 1

Cortical and subcortical auditory regions. Values are means and standard errors of LCGU in micromoles per 100 g of tissue per minute in the three animals. For all structures listed, paired t tests indicated significantly greater LCGU in the intact hemisphere than in the deafferented hemisphere (P < 0.05). Percent difference in the final column was calculated as (intact – deafferented)/intact × 100. Cortical divisions in bold are those of Bonin and Bailey (24). Subdivisions of the core, belt, and parabelt areas listed under TC, TB, and TAc are based on those of Kaas and Hackett (1); the subdivision listed under TAr is that of Pandya (12); and subdivisions listed under FD, FDΔ, and FD/FF are those of Petrides and Pandya (45). The highest levels of glucose metabolism are found in two subdivisions of the core (A1 and R), which have also been shown to have the densest immunoreactivity for parvalbumin (53). For divisions with abbreviations: see Fig. 1; CDt, tail of caudate nucleus; IA/IBd, dorsal one-third of insula; TAc, caudal two-thirds of TA; TAr, rostral one-third of TA; TAts, upper bank of superior temporal sulcus; TGd, dorsal one-half of TG. For locations of Bonin and Bailey divisions and of subcortical structures, see Figs. 1 and 2.

View this table:

Among the largest hemispheric asymmetries observed were those in the areas of the STG comprising the primary (core) and the secondary (belt and parabelt) auditory cortices (1) (Table 1). However, as indicated, metabolic asymmetries extended throughout the entire length of the STG (–4 to +20 in Fig. 2), including its rostral third (rSTG).

The rSTG contained a columnar pattern of activity (Fig. 3, A and B). This pattern, which was observed in all three animals throughout the lateral surface and dorsal bank of the rSTG in the intact hemisphere, was not seen elsewhere in this hemisphere (21, 22), nor was it identifiable in the rSTG of the deafferented hemisphere. In two intact control animals (23), the columns in the rSTG were apparent in both hemispheres but were wider than in the intact hemisphere of the experimental monkeys and were separated by correspondingly narrower, relatively inactive columns (Fig. 3, C and D). The findings suggest that, in intact monkeys, the pattern consists of adjacent thick and thin columns representing ipsilateral and contralateral auditory inputs, respectively, and of a third, thin, unactivated column.

Figure 3

2-DG autoradiographs showing examples of acoustically activated columns within the rostral third of the left STG (TAr/TGd) from (A and B) the intact hemisphere of an experimental monkey, and (C andD) a control monkey. [(A) and (C) are pseudocolor renditions of autoradiographic images (B) and (D)]. In the coronal plane, the metabolically active columns in the experimental monkeys were 300 to 600 μm wide, separated by inactive columns 400 to 600 μm wide. Within each active column, the 2-DG uptake usually extended from layers II to V, being densest in layer IV. Reconstruction of these columns across several coronal sections revealed a stripelike pattern in the horizontal plane, with the active stripes extending 200 to 700 μm in the anterior-posterior direction, separated by 100- to 500-μm gaps. In the control monkeys, the columns were apparent in both hemispheres but were wider (500 to 800 μm) than in the intact hemisphere of the experimental monkeys and were separated by correspondingly narrower (200- to 400-μm) inactive columns. Scale bar in (D), 1 mm.

Outside the STG, which includes cytoarchitectonic areas TC, TB, TA, and the dorsal TG (24), the only area in the temporal lobe that showed metabolic activation was the posterior parahippocampal region; here, the activated tissue was located in parts of area TF (+2 to +8, relative to the interaural plane, in Fig. 2) and TH (+2 in Fig. 2). Other activated limbic cortices included a small patch of area LC in the retrosplenial region (+2 in Fig. 2), and, widely separated from this, a long strip of area LA in the anterior cingulate gyrus just below the callosomarginal sulcus (+14 to +32 in Fig. 2).

Within the parietal lobe, the active area covered approximately the rostral two-thirds of the inferior parietal lobule, comprising portions of both areas PG and PF (–4 to +8 in Fig. 2). This activation extended into the dorsal part of the caudal insula, areas IA and IB (+8 to +2 inFig. 2). Finally, the active frontal lobe tissue included a substantial portion of areas FD and FF in the orbitofrontal (+26 to +32 in Fig. 2) and inferior frontal cortex (+32 in Fig. 2) and, separated from these, a strip within areas FDΔ and FD along the upper bank and lip of the principal sulcus (+32 in Fig. 2).

The subcortical structures activated, beyond the major auditory relay nuclei, were the mid-portion of the superior colliculus (probably corresponding to the intermediate layers), the lateral nucleus of the amygdala, and parts of the ventral putamen and tail of the caudate nucleus (Fig. 1).

The boundaries of the metabolically active regions in the intact hemisphere were sometimes sharply delineated. This was the case, for example, for the rostral border in the inferior parietal lobule (+8 in Fig. 2), the medial border of the superior temporal gyrus at the level of the insula (+8 to +14 in Fig. 2), and the ventral border in the temporal pole (+20 in Fig. 2). In other locations, however, LCGU uptake tapered off only gradually across a cortical area. These latter locations were often regions in which the auditory maps overlapped with maps previously obtained in visual areas (14,15), as described below.

Comparison of the visual and auditory areas, in terms of where they do and do not overlap, suggests that the auditory system is organized along lines similar to those of the better-studied visual system. Previous work has identified three major multisynaptic corticocortical pathways or streams diverging from the striate cortex, each specialized for processing a different type of visual information: (i) An occipitotemporal or ventral stream dedicated to processing stimulus quality; (ii) an occipitoparietal or dorsal stream, concerned mainly with processing stimulus location; and (iii) a third stream coursing through the upper bank of the superior temporal sulcus, contributing (among other things) to processing stimulus motion (25,26). The present results suggest that an auditory region resembling and paralleling the unimodal, ventral visual pathway extends through the entire length of the supratemporal plane together with the exposed surface of the superior temporal gyrus (–4 to +20 inFig. 2); like the ventral visual pathway, this auditory region appears to be modality-specific, suggesting that it is dedicated to analyzing acoustic stimulus quality for purposes of stimulus identification and recognition, just as the ventral visual pathway does for visual stimulus quality (27, 28). In contrast, other large auditory sectors overlap extensively with visual areas. One such sector occupies the caudal half of the inferior parietal lobule (–4 to +8 in Fig. 2), where it is coextensive with a substantial portion of the dorsal visual stream, whereas another extends through nearly the entire length of the upper bank of the superior temporal sulcus (–4 to +14 in Fig. 2) and is therefore coextensive with a major portion of the third visual stream. Multimodal processing in portions of these two sectors has been demonstrated physiologically (2,5). For example, a majority of a sample of neurons in the lower bank of the intraparietal sulcus (area LIP) (+2 in Fig. 2) have been found not only to be bimodal but also to show the same spatial tuning in their responses to the locations of auditory and visual stimuli (5). The present data are consistent with the notion that the two auditory sectors share the functions of the two visual pathways, which, therefore, are probably better viewed as contributing to an analysis of the location and motion, respectively, of both acoustic and visual stimuli.

Sharing of functions also applies to subcortical areas of auditory/visual overlap. Thus, the overlap in the middle layers of the superior colliculus is known to serve the elicitation of orienting to both visual and auditory stimuli (9, 10), and, in the lateral amygdaloid nucleus, to mediate evocation of fear by particular stimuli in the two modalities (29,30). By implication, the overlap in the ventrocaudal neostriatum indicates that auditory input may mediate stimulus-response learning or habit formation in this modality, just as the visual input does in that modality (31).

As in the case of the visual cortical pathways, the current data (+26 to +38 in Fig. 2) support the notion (10) that auditory cortical pathways continue into the prefrontal lobe, where only small sectors appear to be modality-specific. Auditory/visual overlap is also present at much earlier stages of cortical sensory processing, suggesting that cross-modal analysis and integration are likely to be serial/hierarchical processes just as it is in the case of unimodal analysis and integration. In some of the areas of sensory overlap, neuronal responses are enhanced by bimodal stimulation (32). Behaviorally, bimodal combinations of sensory input can markedly strengthen the detection and discrimination of stimuli and increase performance speed and accuracy (33, 34).

Neuroimaging studies in humans (35–41) have revealed a substrate for auditory functions that is proportionately far more extensive than had been revealed heretofore in nonhuman primates (42–45). Some of these studies have implicated a dorsal pathway in the processing of sound sources (37–40), and a ventral pathway in the processing of tonal differences (38), environmental sounds (37), and verbal stimuli (35,36, 41, 46). By providing both a comprehensive and a high-resolution map of the auditory territory in the monkey, including evidence consistent with the notion of functionally different auditory processing streams, the present results help close the gap between the human and monkey cerebral auditory systems, in terms of both their extent and their organization. Perhaps more important, by delineating the many auditory regions in the monkey located beyond the primary and secondary auditory areas, the results lay the groundwork for an investigation of the likely neural substrates of higher-order auditory functions.

Supporting Online Material

Materials and Methods

Fig. S1

  • * To whom correspondence should be addressed. E-mail: amy-poremba{at}


View Abstract

Stay Connected to Science

Navigate This Article