Selectivity for 3D Shape That Reveals Distinct Areas Within Macaque Inferior Temporal Cortex

See allHide authors and affiliations

Science  16 Jun 2000:
Vol. 288, Issue 5473, pp. 2054-2056
DOI: 10.1126/science.288.5473.2054


The anterior part of the macaque inferior temporal cortex, area TE, occupies a large portion of the temporal lobe and is critical for object recognition. Thus far, no relation between anatomical subdivisions of TE and neuronal selectivity has been described. Here, we present evidence that neurons selective for three-dimensional (3D) shape are concentrated in the lower bank of the superior temporal sulcus, whereas neurons in lateral TE are generally unselective for 3D shape, though equally selective for 2D shape. These findings reveal that TE consists of at least two distinct areas, one of which processes a specific object property.

The inferior temporal cortex (IT) is part of the ventral visual stream, which is known to be critical for object recognition (1, 2). Neurons in the anterior part of IT, area TE (Fig. 1), respond selectively to object attributes (3). For over three decades, researchers have been puzzled by the organization of this large cortical region [up to 380 mm2 (4)]. Although TE can be divided into a number of subregions on the basis of anatomical criteria (5), no clear link between the anatomy and neuronal selectivity has yet emerged. Here, we present evidence, based on 3D shape–selectivity, that at least two areas can be distinguished within TE: the lower bank of the superior temporal sulcus (STS) and lateral TE.

Figure 1

Anatomical reconstruction of the recording sites. (A) Superposition of the MR (blue) and the CT coronal section (black). The vertical bar is the guiding tube, the red line the electrode track. (B) Lateral view of the macaque brain showing the anterior-posterior extent of the recordings (vertical lines). The shaded area corresponds to area TE. (C) Estimated recording zones in STS (blue) and lateral TE (green). The red lines indicate the medio-lateral range used in (D), the right one corresponds to the penetration shown in (B). Arrows indicate fundus (a) and lip (b) of the STS, and lateral border of the anterior middle temporal sulcus (AMTS) (c). (D) Average depth (in micrometers) of the silent zones (dotted lines) and the skull (full line), for three medio-lateral bins spaced 2 mm apart. Same scale as in (C). (E) Flattened map of the estimated recording area (average of the two monkeys). (a), (b), and (c) are as in (C). The pale blue area indicates that posteriorly, STS neurons were weakly responsive. Most responsive neurons were recorded in rostral STS. (F) Shown are the 2D shapes used to derive the 3D stimuli.

Thirty-two pairs of disparity-defined (6) curved 3D shapes served as stimuli. The 3D shape–selectivity was assessed by comparing the responses of single TE neurons to the members of pairs of 3D shapes differing only in the sign of their binocular disparity (concave versus convex) (7). Reconstruction of the position of each recorded neuron (Fig. 1) (8) showed that 142 neurons were recorded in the lower bank of the STS and 82 in lateral TE (9). Figure 2A illustrates the responses of an STS neuron that responded strongly to the concave, but not to the convex shape (10). Monocular presentations to left or right eye separately evoked only minimal responses (11). This 3D shape–selective response contrasted markedly with the responses of lateral TE neurons. Although the lateral TE neuron of Fig. 2B responded to the concave shape, the response to the convex shape was equally strong. Moreover, the monocular presentations elicited responses that were similar to those in the stereo conditions. This response pattern, combined with the 2D shape–selectivity revealed by the search test (7), indicates that this neuron was not sensitive to 3D structure, but simply responded to the 2D shape.

Figure 2

Responses of STS (A) and lateral TE neurons (B). The left panel illustrates the images presented to left and right eye for the preferred (top row) and the nonpreferred (bottom row) 3D shape. The middle panel illustrates the perceived 3D structure (semi-cylindrically concave and convex). This illustration does not display the actual borders of the stimulus. The right panel shows the neuronal responses to binocular (“stereo”) and monocular (“left eye” and “right eye”) presentation of preferred and nonpreferred 3D shapes. Bin width is 20 ms.

Overall, 56% of the tested STS neurons were selective for 3D shape, compared to only 12% in lateral TE (χ2 = 40.9,P < 0.001) (12, 13). Moreover, the degree of 3D selectivity was much lower in lateral TE (14), though the mean net response to the preferred 3D shape did not differ significantly between the two subdivisions (20.2 and 19.1 spikes per second for the STS and lateral TE, respectively. Kolmogorov-Smirnov test, nonsignificant). The difference in the population response to preferred and nonpreferred 3D shape is larger for STS than for lateral TE neurons (Fig. 3A) (15). Differences in population responses contain an analysis-induced component because the stimulus evoking the largest response is always selected as the preferred stimulus, even for nonsignificant response differences. The population of lateral TE neurons was no more sensitive to 3D structure than a population of statistically nonselective neurons (n = 135, Fig. 3A), derived from STS (47%) and lateral TE (53%).

Figure 3

Population responses. (A) Population peristimulus-time histograms (PSTHs) for preferred (red) and nonpreferred (blue) 3D shape, for STS (left panel), lateral TE (middle panel), and all statistically nonselective neurons (right panel). (B) Population PSTHs for the response to the preferred 3D shape in the stereo and in the monocular conditions, for STS (black) and lateral TE neurons (green).

We tested whether 3D shape–selective neurons respond to particular disparity values within the stimulus (far versus near), or, alternatively, to spatial variations of disparity (concave versus convex) by presenting the preferred and nonpreferred 3D shapes at five different positions in depth. The large majority of STS neurons tested (37/42) responded to spatial variations of disparity, not to local disparity values. None of four lateral TE neurons tested could be classified as responsive to disparity variations.

The difference in 3D shape–selectivity between STS and lateral TE might reflect a mere generalized difference in stimulus selectivity for the stimuli we used. However, lateral TE neurons were as selective for 2D shape as the neurons in the STS (16), which indicates that the difference between STS and lateral TE was indeed due to 3D shape–selectivity. Although lateral TE neurons are clearly selective for 2D shape, they are generally not selective for 3D shape, even for large disparity values (17).

The response of the STS neurons to the preferred 3D shape was generally larger than the sum of the monocular responses (Figs. 2A and 3B). Typically, the opposite was true for lateral TE neurons (Figs. 2B and3B). We defined a binocular summation index as (response to preferred 3D shape) − (sum of the monocular responses)/(sum of the monocular responses). The median binocular summation index for the STS neurons (+0.18; n = 142) was significantly larger than that for the lateral TE neurons (−0.41; n = 82) (18). The binocular summation observed in STS neurons reflects their selectivity for 3D shape, because binocular presentation of the nonpreferred 3D shape, which contained the same monocular images as the preferred 3D shape (Fig. 2A), evoked no binocular summation (median summation index = −0.65).

The rostral part of the lower bank of the STS corresponds to cytoarchitectonic areas TEa and TEm (5). The connectivity of this region is distinct from that of lateral TE (9). The cortex in the intraparietal sulcus (IPS), the terminus of the dorsal visual pathway, connects with the lower bank of the STS, but less so with lateral TE (19, 20). In addition, the lower STS is more closely connected to TEav than to TEad (21). Our results suggest a specific role for this pivotal brain region: the lower bank of the STS is the part of TE which processes disparity-defined 3D shape ( 22, 23). This finding has two major implications. Firstly, our results provide a functional role for the parietal connections of the STS, because the caudal bank of the IPS contains neurons selective for disparity-defined orientation in depth (24). Thus, these two interconnected areas, belonging to the dorsal and to the ventral streams, are both involved in the processing of 3D-structure (25). Secondly, in view of the distinct cyto- and myeloarchitecture, the specific pattern of connections and the functional specialization of the lower bank of the STS, this region should be regarded as a distinct area within TE.

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


View Abstract

Stay Connected to Science

Navigate This Article