The brain may use a shared code for gender across faces, bodies, and objects

A neuroimaging study suggests that the brain processes cues about gender from faces, bodies, and objects in many overlapping regions at the back and lower sides of the brain. One region, in the right middle temporal gyrus, appeared to represent gender in a similar way whether it was signaled by a face, a body, or an object. The paper, a reviewed preprint, was published in eLife.

Humans are remarkably good at inferring gender from visual information. They make such judgments quickly and with little conscious effort. Faces provide especially powerful cues, but gender can also be inferred from body shape, posture, clothing, and even objects that people associate with men or women.

For faces, this ability remains surprisingly intact even when an image is degraded, partly hidden, or reduced to a silhouette. Such flexibility suggests that the brain may represent gender at a level that goes beyond any single type of visual cue. In other words, the same general concept may be extracted from very different kinds of images.

Scientists have already identified several brain regions involved in judging gender from faces, particularly areas associated with visual and social processing. Much less is known about how the brain handles gender information conveyed by bodies, and gender cues from objects had not been studied in this way before. An important question is whether these different sources ultimately converge on a common neural representation.

Study author Wenjie Liu and colleagues at the Institute of Psychology of the Chinese Academy of Sciences used functional magnetic resonance imaging (fMRI), which tracks changes in blood flow as an indicator of brain activity, to record participants’ brain activity while they viewed images of faces, bodies, and objects associated with men or women. They aimed to establish whether any brain region encodes gender regardless of what type of image a person is viewing. Identifying such a region could help researchers understand how the brain turns diverse visual signals into abstract social categories, such as gender, that guide everyday perception and interaction.

Study participants were 22 healthy adults between 19 and 28 years of age, with an average age of 22. Eleven of them were men.

In the main experiment, participants viewed six types of images while their brain activity was recorded: male and female faces, male and female bodies, and objects associated with men or with women. Each type included 30 different images. Face images showed Chinese male and female movie actors, and body images showed a body in underwear from below the head to above the knees. The object images, which were strongly associated with one gender or the other, came from a previous study.

Images were shown in 12-second blocks, each containing images from a single condition, such as female faces. Each of the six conditions appeared three times per run in random order, and each run lasted 5 minutes and 30 seconds. Participants completed six runs, except for the first two participants, who completed seven. To make sure they paid attention, participants pressed a button whenever the same image appeared twice in a row, which they did with about 97% accuracy; they were not asked to judge gender.

To analyze the scans, the researchers used a machine-learning approach that tests whether patterns of activity in a given brain area can be used to tell male images from female images. The regions carrying this gender information differed somewhat depending on the type of image. For faces, gender information was found in several visual areas on both sides of the brain (the cuneus, calcarine cortex, lingual gyrus, superior and middle occipital gyri, and middle temporal gyrus), as well as in the right superior parietal lobule and the left fusiform gyrus.

For bodies, gender information was found in a large portion of the occipitotemporal cortex, along with clusters in the superior parietal lobule, inferior parietal lobule, and postcentral gyrus on both sides of the brain. For objects, it was found in the cuneus, calcarine cortex, lingual gyrus, fusiform gyrus, superior, middle, and inferior occipital gyri, and inferior temporal gyrus on both sides of the brain, as well as in the right middle temporal gyrus.

The researchers then tested whether gender information carried over from one type of image to another. For example, they trained the classifier on brain responses to male and female faces and tested whether it could tell the gender of bodies. The only region where this worked for all three pairings of faces, bodies, and objects was a small cluster in the right middle temporal gyrus, and a second type of analysis pointed to the same area.

This region was not dedicated to gender alone. Like most of the gender-related regions the researchers identified, it also carried information about whether a person was viewing a face, a body, or an object.

To probe what kind of visual information this region might be using, the researchers compared its activity with an artificial neural network trained to classify images by gender. Activity in the right middle temporal gyrus most closely resembled the network’s middle layers, suggesting that it relies on mid-level visual features. Mid-level visual features are intermediate representations such as contours, textures, shapes, and spatial arrangements that lie between basic features like color or orientation and the recognition of whole objects or categories.

The researchers also examined how gender-related brain regions communicated with one another. The pattern of these connections was modestly similar when people viewed faces and bodies, but not when they viewed objects.

“Together, these findings identify a category-general neural hub for gender perception and illuminate how mid-level visual features and distributed networks support the abstraction of gender across heterogeneous visual inputs,” the authors concluded.

The study contributes to the scientific understanding of the neural basis of social perception. However, eLife’s reviewers rated the evidence as incomplete, citing conceptual concerns, weak statistical methods, and possible confounds from basic differences between the images, such as their size. They suggested the authors temper their claims that the region holds an abstract representation of gender, and the preprint has not yet been revised in response.

The study was also small, with 22 young adults, and participants did not need to attend to gender during the task. Because the faces were of Chinese actors and the gender associations of objects are shaped by culture, the results might differ in other populations.

The paper, “A shared neural code for gender across faces, bodies, and objects in the human brain,” was authored by Wenjie Liu, Xiqian Lu, Yuhui Cheng, Ruidi Wang, Xinquan Lu, Xiangyong Yuan, and Yi Jiang.