Skip to main content
. 2020 Aug 3;30(15):3039–3044.e2. doi: 10.1016/j.cub.2020.05.071

Figure 1.

Figure 1

Experimental Design and Classification Results

(A) Congenitally blind individuals (n = 8) participated in an fMRI experiment in which they were listening to three natural sounds interleaved with silent periods (apart from MRI scanner noise). Multi-voxel pattern analysis (MVPA) was used to decode the sounds from participants’ early visual cortex activity patterns.

(B) In a separate fMRI session, retinotopic mapping was performed for a group of sighted participants to individually define early visual areas V1, V2, and V3. These retinotopically defined regions of interest (ROIs) were then mapped onto a cortical reconstruction of each blind participant using cortex-based alignment. Aligned ROIs were converted into maximum probability maps that were then used in the data analysis as early visual ROIs for blind participants. Additionally, auditory cortex and motor cortex ROIs were defined for each participant using brain atlases.

(C and D) Mean classification accuracy of the classifier distinguishing the three natural sounds in the different ROIs in (C) blind participants and (D) sighted participants. The data for sighted participants (n = 10) were taken from [1] (the results for auditory cortex and motor cortex were recalculated within the same ROIs that were used for blind participants). Early visual cortex (EVC) contains V1, V2, and V3. AC, auditory cortex; MC, motor cortex. Chance level (one out of three) is marked with a black line. Error bars indicate SEM. Testing against chance level was performed with a permutation analysis. Results for V1, V2, and V3 were corrected for multiple comparisons, within each group, with a single threshold test. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p = 0.001.

(E and F) Mean classification accuracies for all visual ROIs divided into three eccentricities (fovea, periphery, and far periphery) in both (E) the blind and (F) sighted group. Results were corrected for multiple comparisons within each group using the false discovery rate. ∗p < 0.05; ∗∗p < 0.01.

See also Figures S1 and S3 and Table S2.