Perception of Dynamic Facial Expressions Probed by a New High-Level After-Effect

High-level after-effects have been reported for the recognition of static faces [1,2]. It has been shown that the presentation of static ‘anti-faces’ biases the perception of neutral test faces temporarily towards the perception of specific identities. Recent studies have demonstrated high-level after-effects also for point-light walkers, resulting in shifts of perceived gender [3,4]. We present an experiment showing for the first time high-level after-effects in the recognition of dynamic facial expressions. Facial expressions were generated as a morph animation based on a weighted sum of 3D shapes derived from scans of facial action units [5]. With this technique we were able to define a metric space of dynamic expressions by morphing, similar to face spaces for static stimuli. Morphing between prototypical expressions (happy and disgust) and a neutral face without intrinsic facial motion we generated ‘anti-expressions’ by choosing negative weights for the prototypes. In addition, for testing we generated expressions with reduced recognizability choosing small positive weights of the prototypes. The morphing space was equilibrated for recognizability by measuring the psychometric functions that map the morphing weights on the recognition rates of the two expressions (happy and disgust) in a 2 AFC task. Only the non-rigid intrinsic face motion was morphed. In addition, a meaningless 3D head motion was added in order to minimize the influence of low-level adaptation effects. Subjects were adapted for 8s with 5 repetitions of the anti-expressions. They were tested with happy and disgust expressions with reduced expression strength. Adaptation stimuli were simulated with 2 facial identities and were shown either in forward or reverse time order. We found strong expression-related after-effects (increased and decreased recognition for matching and non-matching expression, respectively, p < 0.05, N=13). We investigated the influence of static vs. dynamic representations in the observed after-effect. The temporal order of the adapting stimuli does not have a significant influence on the strength of the observed after-effect. The analysis of the 2D optic flow patterns of adaptation and test stimuli rules out the possibility that the observed after-effects reflect classical low-level motion after effects. Instead, the results seem compatible with the adaptation of neural representations of ‘snapshot keyframes’ [6] that arise during the presentation of dynamic facial expressions