Figure 4.

fMRI BOLD Signal Change

BOLD signal changes in each adapt-test condition were extracted from voxels of retinotopic cortex that represented the location of the peripheral stimulus. Activity in response to the change-same and change-different conditions has been compared to that of the no-change condition, used as a baseline here for clarity of data presentation.

(A) All trials of each condition were modeled, regardless of whether the observer detected a change or not (see Figure S1A).

(B) Each individual's behavioral responses were used to model the crowded percept. Only the trials in which the change-same switch was undetected (∼86% of trials) and the change-different switch was detected (∼84% of trials) have been modeled. All false-alarm trials were also rejected (∼10% of trials) (see Figure S1B).

(C) We hypothesized that the difference in activity between crowded and uncrowded states would be greatest for brain areas driving the crowded percept. Accordingly, we computed a PI for each visual area by subtracting the activity for crowded trials from the activity for uncrowded trials to yield a single value for each visual area, representing the increase in activity in uncrowded trials compared to crowded trials. Brain areas that preferentially modulate with the crowded percept have a higher (positive) PI, whereas areas more responsive to physical changes in the stimulus have a PI close to zero. There is a clear trend for the crowded percept to be increasingly represented from early to late visual areas. Error bars indicate +/− SEM in all cases.