Fig. 1.

Backward masking disrupts recognition of partially visible objects. (A and B) Forced-choice categorization task (n = 21 subjects). After 500 ms of fixation, stimuli were presented for variable exposure times (SOA from 25 to 150 ms), followed by a gray screen (A) or a noise mask (B) for 500 ms. Stimuli were presented unaltered (Whole; C, Left and D, Left), rendered partially visible (Partial; C, Right), or rendered occluded (D, Right) (SI Appendix, Fig. S1). (E) Experimental variation with novel objects (SI Appendix, Fig. S8). Behavioral performance is shown as a function of visibility for the unmasked (F) and masked (G) trials. Colors denote different SOAs. Error bars denote SEM. The horizontal dashed line indicates chance level (20%). Bin size = 2.5%. Note the discontinuity in the x axis to report performance at 100% visibility. (H) Average recognition performance as a function of SOA for partial objects (same data replotted from F and G, excluding 100% visibility). Performance was significantly degraded by masking (solid gray line) compared with the unmasked trials (dotted gray line) (P < 0.001, χ² test; df = 4). (I) Performance versus SOA for the occluded stimuli in D (note: chance = 25% here) (SI Appendix, Fig. S1). (J) Performance versus SOA for the novel object stimuli in E.