Abstract
A stream of face images can give rise to a variety of visual interactions, including adaptation, serial dependence, ensemble coding, and contrastive effects. We compared how the rate of presentation impacted two visual judgments – extracting the average expression across a set of faces vs. distortions in expression induced by the preceding faces (as in the flashed face distortion effect, FFDE). The faces were 5 different identities from the Karolinska Directed Emotional Faces, morphed between neutral and happy expressions (Experiment 1) or shown with neutral expressions (Experiment 2). The images subtended a width of 4 degree with the closest edge at 6 degree to the left and right of fixation, and were presented at a range of temporal frequencies (1.2Hz, 4.8Hz, 24Hz, and 60Hz). The total duration of the trial was fixed at 4.17s. 14 participant completed two sessions (one session for each experiment) in counterbalanced order. In the first experiment, the averaged facial expression presented in one visual field was the 50% morph level, and the averaged expression in the other visual field was either 35% or 65% morph level. A two-alterative forced choice task was used to identify which visual field had the stronger expression. In the second experiment, participants were asked to judge how distorted the neutral faces appeared on average. Over the range tested the FFDE decreased with temporal frequency while the accuracy of expression averaging increased. Our results indicate a transition from averaging to contrastive interaction (distortion) as the temporal frequency decreases.