Visual and Auditory Object Representations in Ventral Visual Cortex After Restoring Sight in Humans

Abstract

Visual category-selective representations in human ventral occipital temporal cortex (VOTC) seem to emerge early in infancy. Surprisingly, the VOTC of congenitally blind humans features category-selectivity for auditory and haptic objects. Yet it has been unknown whether VOTC would show category-selective visual responses if sight were restored in congenitally blind humans. Assuming competition for synaptic space during development, cross-modal activation of VOTC as a consequence of congenital blindness might interfere with visual processing in sight-recovery individuals. To test this hypothesis, we investigated adults who had experienced a transient phase of congenital blindness due to bilateral dense cataracts before their sight was restored by cataract-removal surgery. In a functional magnetic resonance imaging (fMRI) study, participants watched movies of faces, scenes, body parts, and other objects in the visual condition, while in the auditory condition they listened to the corresponding sounds. The most prominent group difference was a reduced face-selectivity in individuals with reversed congenital cataracts compared with age- and sex-matched normally sighted individuals. In addition, a double dissociation was found: only sight recovery individuals demonstrated significant decoding accuracy of visual categories based on auditory category representations in VOTC, while only normally sighted individuals' VOTC decoded auditory categories based on visual category representations. The present results uncovered the neural mechanisms of previously observed face processing impairments in individuals with reversed congenital blindness. We suggest that lower face-selectivity in the sight recovery group might arise from selective deficits in the cortical representation of the central visual field in lower-tier visual areas. Additionally, we speculate that in higher-order visual areas cross-modal activity might facilitate-rather than interfere-with visual functional recovery after congenital blindness.

Keywords: category‐selective representations; fMRI; sensitive periods; sight recovery.

FIGURE 1

Experimental design (here illustrated for the visual condition). Stimuli consisted of short movie or audio clips representing four different categories—body parts, scenes, faces, and objects—presented in separate visual and auditory runs (see Van Den Hurk et al. 2017). Each clip (video or audio) lasted 1800 ms with a 200 ms interstimulus interval. Stimuli were presented in blocks of 8 clips per category lasting for 16 s in total. Blocks were separated by a 12 s long interval. Each of the 8 runs consisted of 4 blocks of each of the four categories and lasted approximately 9 min. Four runs were visual, and the other four were auditory runs. Note, that schematic representations of the natural image clips are used in this figure; for the original movies see Van Den Hurk et al. (2017).

FIGURE 2

Whole‐brain analysis. (A) Statistical maps for each category‐selective contrast were calculated for visually presented movie clips of: Faces (vs. all other categories), scenes (vs. all other categories), bodies (vs. all other categories) and other objects (vs. all other categories) in normally‐sighted individuals (the SC group) and in congenital cataract reversal individuals (the CC group). (B) Interaction between group and condition. The voxel‐wise threshold was set to p < 0.001 uncorrected and the resulting statistical maps were corrected for multiple comparisons using cluster‐wise FWE‐correction at p < 0.05. (C) Group‐overlap maps for face‐selectivity, body‐selectivity, scene‐selectivity and other object‐selectivity are shown separately for the SC and the CC group. Individual subject maps of category‐selective contrasts were thresholded (p < 0.0001, voxel‐wise, uncorrected) and the resulting statistical maps were corrected for multiple comparisons with cluster‐wise FWE‐correction at p < 0.05. Each voxel was consecutively assigned the color, which represents the number of the participants who showed an above threshold activation in this voxel. For better visualization, group‐overlap maps were constrained by the ventral occipital temporal mask (VOTC; see Methods: Within‐Subject Classification). L = left hemisphere; R = right hemisphere.

FIGURE 3

ROI analysis in the regions functionally defined by Rosenke et al. (2021). Mean signal change in category‐selective regions is shown for bodies, scenes, faces and other objects in the SC and the CC group. First, to independently define the ROIs the 30 most active voxels in each of the participants were identified within the functional mask, as defined in the visual atlas of Rosenke et al. (2021), in the first visual run. Next, in these voxels we extracted the beta activation parameters from the remaining three visual runs for each participant and each experimental condition. Finally, the beta‐values were averaged across all 30 voxels in each participant. CC = congenital cataract reversal individuals, SC = normally‐sighted individuals.

FIGURE 4

Classification results. Within‐subject decoding results for visual (A) and auditory (B) categories in four ROIs in the CC and the SC group. Cross‐modal classification results for (C) visual‐to‐auditory (D) and auditory‐to‐visual decoding in four ROIs in the CC and the SC group. The values were tested against chance level = 25% and between the groups. *p < 0.05, **p < 0.01, ***p < 0.001. undec = undecided cases. CC = congenital cataract reversal individuals. SC = normally‐sighted individuals.