Abstract
Numerous behavioral studies have demonstrated that face recognition depends on holistic processing, based especially on the famous composite face effect. The decline in hit rate when composite halves are aligned compared to when they are misaligned is taken to reflect interference from the whole-face configuration. However, whether this interference is due to a perceptual bias or reduced discriminative efficiency is still a matter of debate. To clarify the nature of this effect, we devised a new composite paradigm using ambiguous stimuli. We generated three series of morphed faces derived from merging pairs of celebrities. Ambiguous morphed top-halves were coupled with unmorphed bottom-halves, the two halves being aligned or misaligned. The task was to say which celebrity the ambiguous top-half most resembled. We plotted the frequency of response and response time as a function of the morphing level of the top-half and the identity of the unmorphed bottom-half. We examined whether, compared to the misaligned condition, the data in the aligned condition showed a decrease in slope of the psychophysical function, which would indicate reduced discriminative efficiency, or a change in intercept alone, which would indicate a perceptual bias. 30 healthy participants performed this task. Our results showed that the bottom-half altered responses to the top-half when faces were aligned but not when they were misaligned. More crucially, the change was due to a shift in the position of the psychophysical function (p<0.001) without a significant change in the slope of these functions. These results demonstrate that the composite illusion originates in a perceptual bias introduced from one face-half that affects the decision about the identity of the other face-half when they are aligned, and is not from decreased discriminative efficiency of judgments about that other face-half, due to reduced attention, interference, or other possible mechanisms that can degrade perception.
Meeting abstract presented at VSS 2012