Determining Canonical Views of Objects via Electrical Brain Imaging

S Sasane, L Schwabe

Institute for Computer Science, University of Rostock, Germany
Contact: lars.schwabe@uni-rostock.de

Some “canonical” object views are preferred compared to others. Psychophysical studies found that those views are recognized faster and recalled easier from memory (Blanz et al. 1999) but they activated other cognitive processes in addition to vision. We aimed at isolating the visual system’s contribution and performed two experiments while brain responses were recorded using electroencephalography with 64 channels. Experiment 1 (N=12) followed an oddball paradigm with computer-generated frequent target (a table), rare distractor (a car) and target stimuli (a chair), each in a priori determined canonical and non-canoncial views (1 stimulus/sec). Subjects were passively viewing and actively detecting the target. Experiment 2 (N=10) was a target-detection within rapid-serial visual presentation paradigm (RSVP, 10 stimuli/sec) with 40 systematically rendered views per object. Event-related potential analysis shows that i) in experiment 1 the early response (0..200 ms) distinguished between the 6 stimuli, but ii) the target-evoked P300 wave in the detection condition was indistinguishable between the two views. In experiment 2 the iii) P300 strength was affected by the target view: Apparently “canonical” views caused strongest P300 responses. Our results suggest RSVP as a paradigm to isolate the visual system’s contribution in mapping canonical views from other possibly confounding cognitive processes.

Up Home