Abstract
The face inversion effect (FIE) is a well studied phenomena in which inverted faces are recognized significantly slower and with less accuracy than upright faces (Yin, 1969). In the following study, we expose a differential time course to upright versus inverted face processing. We then leverage these results to indirectly quantify the relative contribution of nine different internal facial dimensions to the processing of upright and inverted faces.
Subject ability to detect differences between three morphed faces was measured using an adaptive threshold estimation procedure in a same-different delayed match-to-sample task. Test stimulus duration varied in five log-spaced steps from 25 ms to free viewing. Within each trial, both faces were either presented upright or inverted. At 25 and 53 ms, thresholds for upright faces did not differ significantly from inverted faces. When viewing upright faces at durations greater than 53 ms, however, performance increased relative to inverted faces, reaching a peak at 107 ms and gradually decreasing as duration increased. These results suggest a temporal dynamic to the FIE as face specific processes are engaged after the face has undergone some degree of initial processing.
In a second experiment, JNDs for individual part dimensions (eyebrow, eye, nose, and mouth shapes) and configural dimensions (eyebrow-eyebrow, eyebrow-eye, eye-eye, eye-nose, and nose-mouth distances) were measured using the same method described above for two subjects at 25 ms, 107 ms and free viewing. Upright and inverted JND measures were combined into a composite score and compared to a similar score derived from the whole face measures taken previously. When ranked by Euclidean distance from whole face scores, mouth shape, eye-eye distance and eye-nose distance were ranked highest suggesting that the differences in perceptibility between upright and inverted faces over test stimulus duration may be due to similar changes in perceptibility in these dimensions.