A Cortical Area Selective for Visual Processing of the Human Body

See allHide authors and affiliations

Science  28 Sep 2001:
Vol. 293, Issue 5539, pp. 2470-2473
DOI: 10.1126/science.1063414


Despite extensive evidence for regions of human visual cortex that respond selectively to faces, few studies have considered the cortical representation of the appearance of the rest of the human body. We present a series of functional magnetic resonance imaging (fMRI) studies revealing substantial evidence for a distinct cortical region in humans that responds selectively to images of the human body, as compared with a wide range of control stimuli. This region was found in the lateral occipitotemporal cortex in all subjects tested and apparently reflects a specialized neural system for the visual perception of the human body.

One of the most fundamental questions about visual object recognition in humans is whether all kinds of objects are processed by the same neural mechanisms, or whether instead some object classes are handled by distinct processing “modules.” The strongest evidence to date for a modular recognition system comes from the case of faces [(1), but see (2)]. In contrast, relatively few studies have considered the mechanisms involved in perceiving the rest of the human body. Neuropsychological reports suggest that semantic knowledge of human body parts may be distinct from knowledge of other object categories (3). In addition, functional neuroimaging studies have implicated regions of the superior temporal sulcus (STS) in the perception of biological motion (4–6) and have associated regions of left parietal and prefrontal cortices with knowledge about body parts (7). Finally, single-unit recording studies in monkeys have identified neurons in the STS that respond selectively to the appearance of the body, including the face (8–9). None of these findings, however, provides conclusive evidence for a region in human visual cortex selectively involved in processing the appearance of human bodies. Here, we report a series of fMRI studies that provide the first evidence for such a region.

Subjects in these experiments were scanned while viewing images of objects from several different categories. In 19 out of 19 subjects scanned, we found a region in the right lateral occipitotemporal cortex (Fig. 1) that produced a significantly stronger response when subjects viewed still photographs of human bodies and body parts than when they viewed various inanimate objects and object parts (10). We have provisionally named this candidate body-selective region the “extrastriate body area” or EBA (11). After identifying the EBA in each subject with these “localizer” scans, we then ran a new set of experimental scans in the same session, in order to measure the response of the EBA to a large number of other stimulus categories (Figs. 2 and 3). This procedure enabled us to characterize the response profile of this region to a variety of different kinds of visual stimuli (12) in order to test a number of alternatives to our hypothesis that the EBA is selectively involved in visual processing of the human body.

Figure 1

EBA activations in three individual subjects. Each row shows coronal anatomical slices from a single subject, arranged from posterior (left) to anterior (right), overlaid with a statistical map showing voxels that were significantly more active for human bodies and body parts than for objects and object parts. The EBA is visible in the right occipitotemporal cortex of each subject (arrows); in some subjects an activation was also observed in the corresponding location of the left hemisphere. Scale indicatesP value of activations in colored regions.

In experiment 1, the response [percent increase in the MR signal (13) in the EBA] to human body parts (1.3%) and face parts (1.0%) was significantly greater than that to object parts (0.5%;P < 0.001 in each condition). It is noteworthy that the response to whole faces (0.6%) was significantly lower than to face parts (P < 0.01) and body parts (P < 0.001) and no greater than that to whole objects (0.5%; P > 0.20) (14). Experiment 2 revealed that the response to hands (1.4%) did not differ from that to assorted body parts (1.4%). In experiment 3, the response to human bodies (1.9%) was greater than that to human body parts [1.2%;P < 0.025 (15)], which in turn was greater than that to object parts [0.4%; P < 0.005 (16)]. Thus, relative to the control conditions, the EBA responded strongly and selectively to a variety of pictures of human bodies and body parts, with the single exception of faces.

In order to minimize differences between body and nonbody stimuli on low-level image properties such as texture, shading, and spatial frequency composition, we measured the response of the EBA to line drawings. In experiment 2, the EBA response to body parts was greater than to object parts, whether represented as line drawings (P < 0.005) or as photographs (P < 0.001). Similarly, in experiment 4 we found a significantly greater response to entire bodies (1.6%) than to cars (0.7%;P < 0.01), whether they were presented as line drawings or photographs (17). To further rule out low-level visual confounds, in experiment 6 we tested the EBA response to stick figure representations and silhouettes of human bodies, compared with scrambled versions of the same stimuli. The response to stick figures (1.7%) was significantly higher than that to the control items (1.0%; P < 0.01). Likewise, the response to human silhouettes (1.8%) was greater than that to scrambled versions (1.0%; P < 0.05). Finally, experiment 5 showed that the EBA does not respond generally to any object that, like the human body, is composed of rigid subparts connected at flexible joints. The EBA response to common articulated inanimate objects such as scissors (0.7%) was significantly lower than to bodies (1.5%;P < 0.005) and body parts (1.3%; P < 0.005) and did not differ from that to object parts (0.7%). We conclude that the selectivity of the EBA for body stimuli is not due to differences in the surface or structural properties of the stimuli (18).

Previous reports [e.g. (19)] have shown greater responses to animals than tools in a region near the EBA (in addition to an activation in the lateral fusiform gyrus), raising the question of whether the EBA is more generally responsive to all animals, rather than specifically to humans. In experiment 3, the response to nonhuman mammals (1.0%) was significantly lower than that to humans (1.9%;P < 0.01) and human body parts (1.2%;P < 0.05), but greater than that to object parts (0.4%; P < 0.001). In experiment 5, we tested several animal categories, revealing that the responses to nonhuman mammals (1.0%), birds (0.9%), and fish (0.8%) were all significantly below that to human bodies (1.5%; P < 0.01,P < 0.005, P < 0.005, respectively). Further, the response to mammals was only marginally higher than that to articulated objects (0.7%; P < 0.08), and this difference was not significant for birds or fish. Thus the EBA responds selectively to humans as opposed to other animals, with a possible modest preference for nonhuman mammals as opposed to other animal categories. However, this region is clearly not selective for the category of animals in general.

To further assess the relationship between the EBA and other visually selective cortical regions (Fig. 4), we tested for anatomical overlap (20) between the EBA and the following: retinotopic cortex (21), the fusiform face area (FFA) (22), the parahippocampal place area (PPA) (23), the lateral occipital complex (LOC) (24), and the visual motion area MT/V5 (25). We found no anatomical overlap between the EBA and retinotopic cortex, the FFA, or the PPA, as functionally defined in each subject individually. In some subjects, the EBA overlapped partially with either area MT or LOC (26). However, in 11 out of 11 subjects, a set of voxels was uniquely activated in the EBA localizer and not in either the MT or LOC localizers (mean 1.5 cm3, SEM 0.33 cm3). Thus, the EBA is clearly distinct from these other previously described visual regions (27).

Figure 2

Stimulus examples. The EBA response was high to human body parts (A) and whole human bodies (B) whether presented as photographs, line drawings (C), stick figures (D), or silhouettes (E), and was not attenuated to images that depict little implied motion (F). The low response to whole faces (G) was the single exception found to the preference for human bodies. In contrast, the EBA response was significantly lower to object parts (H) and whole articulated objects (I), whether represented as photographs or line drawings (J), as well as to scrambled control versions of stick figures (K) and silhouettes (L). The responses to face parts (M) and to mammals (N) were intermediate.

Figure 3

Grand mean activation time courses from experiments 1 to 6 [(A) to (F), respectively]. Each block is indicated with a color band; repetitions of conditions within a scan are indicated by shared color. Data are in terms of percent signal change from fixation baseline (shown in gray). All data were extracted from regions of interest defined in independent scans, in each subject, within a session. Conditions in all experiments are labeled as follows (some conditions superceded by other analyses are not discussed here): OP, object parts; BP, human body parts; FACE, faces; HAND, hands. (A) FP, face parts; SCENE, outdoor scenes; WO, whole objects; SO, scrambled objects. (B) L-OP, line drawings of object parts; L-BP, line drawings of body parts; MOV, oscillating low-contrast rings; STAT, static low-contrast rings. (C) HUM-F, human bodies, without faces; MAM-F, mammal bodies, without faces. (D) P-CAR, photographs of cars; L-CAR, line drawings of cars; PA-HUM, photographs of alert humans; LA-HUM, line drawings of alert humans; PI-HUM, photo of inactive humans; LI-HUM, line drawings of inactive humans. (E) HUM-F, human bodies, without faces; MAM, mammals; BIRD, birds; FISH, fish; TREE, trees; AO, articulated objects. (F) F-CL, filled clothes; E-CL, empty clothes; SIL, human silhouettes; STICK, human stick figures; S-SIL, scrambled silhouettes; S-STICK, scrambled stick figures.

Figure 4

Coronal slices from a single subject, arranged from posterior (top left) to anterior (bottom right), showing the EBA, FFA, MT, LOC, PPA, and face-selective region of STS, all identified within a single scanning session. Colored voxels are those which reached significance (P < 10−7) in a standard localizer scan for each region; regions of overlap are not indicated.

In conclusion, our results reveal a region in human lateral occipitotemporal cortex that responds selectively to visual images of human bodies and body parts, with the exception of faces. These findings suggest that the EBA is a specialized system for processing the visual appearance of the human body. At present, we can only speculate on the precise functional role of the EBA. It may be involved in the identification of individuals, perhaps under conditions in which face recognition is not possible (e.g., when the face is not visible because of viewing direction, distance, occlusion, poor lighting, and so on). Alternatively, the EBA may be critical for perceiving the position and/or configuration of another person's body, perhaps as part of a broader system for inferring the actions and intentions of others. Finally, it may be involved in perceiving the configuration of one's own body, for example in the guidance of actions.

In its strong selectivity for a specific object category, the EBA resembles other previously identified regions of human extrastriate cortex such as the FFA and the PPA. Although all three respond somewhat to nonpreferred stimuli, each shows a general and strongly selective response to stimuli from its preferred category. The existence of these category-selective regions in human extrastriate cortex supports the hypothesis that high-level vision is not accomplished by a single functionally undifferentiated system. Rather, visual perception and cognition appear to be served by distinct mechanisms for at least a select few categories, including faces, places, and bodies.

How many category-specific regions like the EBA exist in human extrastriate cortex? In ongoing studies, we have tested a wide range of other object categories and have so far found no compelling evidence for other category-selective regions in occipitotemporal cortex. This result suggests that objects from many categories may be represented by a “general-purpose” recognition system, a role proposed for the lateral occipital complex (28, 29). Perhaps the most fundamental unanswered question about category-selective regions in human extrastriate cortex concerns the origins of these structures. Are the FFA, PPA, and EBA largely specified in the genome, or do these regions primarily derive from the extensive lifetime experience an individual has with faces, places, and bodies? Methodological advances now being developed may enable us to answer even these most challenging questions about the organization and origins of object representations in the human brain.

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


View Abstract

Navigate This Article