Smell what you hardly see: Odors assist categorization in the human visual cortex
Visual categorization is the brain ability to rapidly and automatically respond to widely variable visual inputs in a category-selective manner (i.e., distinct responses between categories and similar responses within categories). Whether category-selective neural responses are purely visual or can be influenced by other sensory modalities remains unclear. Here, we test whether odors modulate visual categorization, expecting that odors facilitate the neural categorization of congruent visual objects, especially when the visual category is ambiguous. Scalp electroencephalogram (EEG) was recorded while natural images depicting various objects were displayed in rapid 12-Hz streams (i.e., 12 images / second) and variable exemplars of a target category (either human faces, cars, or facelike objects in dedicated sequences) were interleaved every 9th stimulus to tag category-selective responses at 12/9 = 1.33 Hz in the EEG frequency spectrum. During visual stimulation, participants (N = 26) were implicitly exposed to odor contexts (either body, gasoline or baseline odors) and performed an orthogonal cross-detection task. We identify clear category-selective responses to every category over the occipito-temporal cortex, with the largest response for human faces and the lowest for facelike objects. Critically, body odor boosts the response to the ambiguous facelike objects (i.e., either perceived as nonface objects or faces) over the right hemisphere, especially for participants reporting their presence post-stimulation. By contrast, odors do not significantly modulate other category-selective responses, nor the general visual response recorded at 12 Hz, revealing a specific influence on the categorization of congruent ambiguous stimuli. Overall, these findings support the view that the brain actively uses cues from the different senses to readily categorize visual inputs, and that olfaction, which is generally considered as poorly functional in humans, is well placed to disambiguate visual information.