Report

Functional Specialization in Rhesus Monkey Auditory Cortex

See allHide authors and affiliations

Science  13 Apr 2001:
Vol. 292, Issue 5515, pp. 290-293
DOI: 10.1126/science.1058911

Abstract

Neurons in the lateral belt areas of rhesus monkey auditory cortex prefer complex sounds to pure tones, but functional specializations of these multiple maps in the superior temporal region have not been determined. We tested the specificity of neurons in the lateral belt with species-specific communication calls presented at different azimuth positions. We found that neurons in the anterior belt are more selective for the type of call, whereas neurons in the caudal belt consistently show the greatest spatial selectivity. These results suggest that cortical processing of auditory spatial and pattern information is performed in specialized streams rather than one homogeneously distributed system.

Hearing plays a dual role in the identification of sounds and in their localization. Although it is undisputed that auditory cortex participates in the analysis of spectro-temporal patterns for the identification of complex sound objects, including speech and music, the neural basis of auditory spatial perception remains a matter of controversy. Brainstemstructures play a significant role in the processing of binaural cues, which contain important information for sound localization (1). However, lesions of auditory cortex also impair auditory spatial analysis (2, 3). With the recent discovery of multiple cochleotopic maps in nonprimary auditory cortex of the rhesus monkey (4, 5), the question arises whether neurons in some of these areas show greater specificity for sound source location than in others. This could indicate the existence of a specialized cortical stream for the processing of auditory space, similar to what has been postulated for the visual cortical system (6). Alternatively, auditory spatial in formation may be encoded in a completely homogeneous, distributed manner with each cortical area contributing equally to auditory spatial perception.

One of the cortical systems in the macaque identified beyond the primary-like core areas of the auditory cortex is the lateral belt (7, 8). It contains at least three areas: an anterolateral (AL), a middle lateral (ML), and a caudolateral (CL) area (4). CL and AL receive largely independent and parallel input from different areas of the auditory core (A1 and R, respectively) and are only weakly interconnected (9,10). CL in addition has strong connections with the caudomedial area (CM) and projects to a different target region of the prefrontal cortex than AL (10, 11). ML, situated between AL and CL and connected with both, shows attributes of an intermediate stage between core and belt (10, 12). Neurons in all three lateral belt areas respond better to complex sounds than to pure tones (4, 13). Stimulus preferences include band-passed noise (BPN), frequency-modulated sweeps, and species-specific vocalizations (“monkey calls,” or MCs). The caudal part of the superior temporal gyrus (STG) contains neurons that are spatially tuned to the location of a sound stimulus presented in free field (14, 15). We therefore compared the spatial selectivity of single neurons in the caudal belt region of four rhesus monkeys with that in more anterior areas. At the same time, by using MCs for stimulation, we determined the extent to which neurons in the various belt regions differ with regard to their specificity for different kinds of sound.

First, BPN bursts of variable bandwidth centered at different frequencies were used to map the lateral belt areas on the STG to determine the borders of AL, ML, and CL (4,16). Then, a standard battery of seven selected MCs (17) was used to test the spatial tuning of the same neurons in azimuth and evaluate their specificity for MCs (Fig. 1). To determine a neuron's selectivity for the seven spatial positions with the seven MCs, we assembled a “response profile” on the basis of 490 stimulus presentations (18). In comparing response profiles of cells in AL (Fig. 2A) with those in CL (Fig. 2B), it became evident that CL responses were often highly specific for spatial position, whereas AL neurons usually responded equally to sounds from all locations. Some of the spatially selective CL neurons were also highly specific for the type of MC presented (Fig. 2B, right), whereas others showed broader tuning in that domain. AL neurons, by contrast, seemed overall to be more specific for the kind of MC regardless of where in space it was presented. Because only a limited azimuth range (120°) was tested, the overall number of spatially tuned neurons was almost certainly underestimated, but this was equally true for all three areas. Despite the different spectral bandwidth of the calls, there was no overall tendency for any of the call types to be associated with spatial-tuning width, nor was there any dependency of spatial-tuning width on the best center frequency (BFc) of the neurons (r 2 = 0.002, P = 0.4456). Selectivity for MCs was also not simply a function of BFc, because neurons selective for different types of calls were found in all sections of the cochleotopic maps within the lateral belt [MC preference index (MCPI) versus BFc:r 2 = 0.013; P = 0.0761]. Thus, MC selectivity is generated by complex integrations in both frequency and time (4, 13).

Figure 1

Testing of spatial tuning in single neurons of macaque auditory belt cortex. (A) Loudspeakers set up in near–free field at the height of the monkey's ears (zero elevation) in seven different azimuth positions. The calibrated speakers were mounted in 20° steps on a stationary horizontal hoop with a radius of 1.14 m, thus spanning a range of ±60° from the center of the animal's head. (B) Spectrograms of the seven standard monkey calls that were used for stimulation.

Figure 2

Response profiles of single neurons in areas AL (A) and CL (B) of macaque monkey auditory cortex. Two representative examples from each area are shown. The selectivity of these neurons for the seven communication calls is displayed along the vertical axes, and selectivity across the seven spatial positions along the horizontal axes. The 49 individual responses were plotted according to the color scale shown at the top. For display purposes, linear interpolation was performed between data points.

Mapping of spatial selectivity along the cortical surface in individual monkeys again demonstrates that neurons with great spatial selectivity are concentrated in the caudal belt (Fig. 3). Great care was taken to collect an equal amount of data from either end of the belt region (19). The interindividual variability of both spatial and MC selectivity was small in each area (P > 0.05, Kruskal-Wallis, df = 3), so individual data sets could be combined for further analysis.

Figure 3

Maps of spatial selectivity in the lateral auditory belt from two monkeys. Spatial half-width is displayed on a color scale from yellow (most selective) to blue (least selective), as indicated at the bottom of the right panel. Borders between cortical areas AL, ML, and CL (dashed lines) were determined by mapping best center frequencies with band-passed noise bursts along the lateral sulcus (ls). Electrode penetrations are denoted by the letter “p” followed by a number and are projected onto a digital picture of the cortical surface in each animal showing vascularization pattern and major sulci (sts: superior temporal sulcus).

The main results of the study are summarized in Fig. 4. Comparison of spatial tuning (Fig. 4A) shows that CL neurons were by far the most sharply tuned and AL neurons were most broadly tuned. ML, commensurate with its anatomical status, fell in the middle of the tuning range with a hint of a bimodal distribution. The difference between the three areas in spatial half-width was highly significant when compared with a nonparametric analysis of variance (P < 0.0001, Kruskal-Wallis, df = 2). The same result was obtained when CL was compared individually with AL and ML (P < 0.0001 andP = 0.002, respectively; Mann-Whitney Utest) (20). Comparison of MC selectivity (Fig. 4B) showed AL to be more selective than both ML and CL (P = 0.0006 and P = 0.0287, respectively; Mann-Whitney Utest). This difference was also highly significant when all three areas were compared together (P = 0.0026, Kruskal-Wallis, df = 2). MC selectivity in CL, when present, was often associated with the prevailing high selectivity in the spatial domain.

Figure 4

Distribution of spatial half-width (A) and monkey call preference index (B) in areas AL, ML, and CL. Summary data from all four monkeys are shown in histogram form. The number of units recorded in each area is given on the right. Neurons in CL show significantly greater spatial selectivity than neurons in AL or ML. By contrast, neurons in AL are more selective for monkey calls than neurons in either of the other areas.

Thus, we find a clear dissociation of auditory spatial tuning between anterior and caudal belt in rhesus monkey auditory cortex. Spatial selectivity is greatest in CL and lowest in AL. Our findings represent physiological evidence for a functional specialization within the auditory cortex of primates and support the idea of processing streams at higher levels of the auditory system (11, 21). Auditory spatial information is known to be processed in posterior parietal (22, 23) as well as dorsolateral prefrontal cortex (PFC) (11, 22). The caudal belt provides direct input to both of these regions (11,13, 24) and could thus be regarded as the origin of a dorsally directed “where” stream for auditory processing. The finding of an increased concentration of spatially tuned neurons in the caudal belt is consistent with other studies (14, 15, 25), which demonstrate that this organizational feature is indeed associated with sound localization behavior. Other aspects of auditory spatial perception, in which the dorsal cortical pathway may be involved, include distance perception (26) and analysis of auditory motion (23).

By contrast, AL seems to be part of an auditory-processing stream that is rostrally directed and continues into ventral and orbital PFC (11, 13, 27). The latter route is supported by recent human imaging data that demonstrate an antero-ventral “what” stream for the analysis of human speech (28–30). However, our results suggest that signals used for auditory communication are also relayed to the caudal STG, where they are combined, at the single-unit level, with information about the location of sounds in space. This type of neuron could play an important role in sound segregation (and possibly identification of speakers) on the basis of spatial cues (31, 32).

The usefulness of natural complex sounds as stimuli in higher auditory areas has been emphasized previously (4,13). The selectivity for specific types of MCs, as found in auditory belt neurons, is higher than expected. However, even in AL, neurons rarely responded to a single call [although they sometimes responded to calls within the same phonetic category (33)]. This suggests that AL is still far from the end-stage in processing auditory objects, and recordings from awake animals in even more anterior and lateral areas of the STG may be promising. On the other hand, lack of extreme selectivity may also indicate that complex auditory patterns, such as vocalizations, are coded by networks of neurons rather than a single cell.

  • * To whom correspondence should be addressed. E-mail: rauschej{at}georgetown.edu

REFERENCES AND NOTES

View Abstract

Navigate This Article