Abstract
Dynamic social signals are steadily stimulating the visual system during human interactions. A wealth of such signals are transmitted as facial expressions for communicating internal emotional states. Unlike static snapshots, routinely used in most of the experiments, dynamic facial expressions provide observers with richer and ecologically-valid signals. Common intuition would thus suggest an advantage for the recognition of dynamic over static inputs. However, while many studies reported an advantage in the recognition of dynamic over static expressions in clinical populations (Richoz et al., 2015), results obtained with healthy young adults are by far more contrasted. To clarify this issue, we conducted a large sample cross-sectional study to investigate facial expression recognition from early to elderly age. Over 400 observers (age range 5-100) performed recognition tasks of the six basic expressions in three conditions: static, shuffled (temporally randomized frames) and dynamic (Gold et al., 2013). We normalized the stimuli for their low-level properties and the amount of energy sampled over time, even for the static condition. Facial expression recognition profiles revealed a better performance for "happy" and typical confusions among expressions with similar morphology (fear-surprise), regardless of condition and age. We then applied a Generalized Additive Model with smoothing spline on an efficiency index to capture the nonlinear relationship between age and the experimental conditions. Overall, we observed strong efficiency in the recognition of dynamic facial expressions in the elderly population. This observation was driven by a suboptimal performance for static and shuffled expressions, a potential marker for impaired face processing that might be linked to other facets of general cognitive decline. Our findings also posit the use of dynamic stimuli as being critical in the assessment of facial expression recognition in elderly populations, inviting to caution when drawing conclusions from the sole use of static face images to this aim.
Meeting abstract presented at VSS 2017