Abstract
We hypothesize that the visual system uses the temporal continuity of object identity to build position invariant object representation. This predicts that varying an object’s identity while maintaining its continuity of movement will alter position invariance. Here, we tested this prediction for human facial gender perception. During an Exposure Phase, sixteen subjects covertly tracked smiles/blinks of a face (3 degrees in size) that continuously orbited around a fixation point (10 sec period, 5 degrees eccentricity). While orbiting, facial gender morphed alternately between female and male, becoming maximally female when crossing the position to the left of fixation (female pole) and maximally male at the position to the right (male pole). Subjects performed this task for 30 minutes broken into 6 minute blocks. Sixteen other subjects participated in a control condition in which the orbital path was broken along the vertical meridian into two alternating semi-circular paths. To test for the predicted experience-induced changes in position invariance, subjects performed a Discrimination Task judging whether the gender of two simultaneously presented faces – one at each gender pole – was the same or different (120 trials, 50ms presentations, eccentricity and size same as in Exposure Phase). We found that following the Exposure Phase, subjects tended to perceive faces as female at the male pole, and male at the female pole. This experience-induced perceptual bias was significantly stronger in the continuous condition (Exposure Phase) compared to the control condition, ruling out a simple adaptation explanation. The induced bias was attenuated but still significant 24 hours after the Exposure Phase. This suggests that the temporal continuity experience is strong enough to alter position invariant object judgments in <1 hour and this effect persists for at least 24 hours. We propose that the natural temporal continuity of objects in motion builds perceptual tolerance to translation.
Meeting abstract presented at VSS 2012