For each participant, four ROIs were generated using data from the second fMRI experiment (one table column per ROI). AV ROIs were created from pSTS regions that showed a significant positive response to blocks of AV speech (q < 0.01). A∩V ROIs were created from pSTS regions that showed a significant positive response to both unisensory A and unisensory V stimulation (both q < 0.05). The percent BOLD signal change in each ROI was calculated for all trials, mouth-viewing trials, and eye-viewing trials in the first fMRI experiment (one table row per trial type). For each participant, the fraction of trials on which a McGurk fusion percept was reported in response to a McGurk stimulus was calculated. Across 34 participants, we correlated each ROI fMRI value with each participant’s McGurk fraction to obtain r and p values. No correction for multiple comparisons was performed.