The serial dependence of perceived emotional expression

People are not only able to quickly and accurately recognize emotional expressions, but are also highly sensitive to expression changes on short time scales. However, this sensitivity could result in unstable percepts of an individual’s emotional expression from moment to moment. The visual system must therefore balance the need to accurately recognize emotional expressions with maintaining perceived stability over time. A mechanism that could serve this purpose is the Continuity Field, a spatiotemporal region within which the visual system biases object identity towards stability (Fischer & Whitney, 2014; Liberman et al., 2014). The Continuity Field may stabilize the expression attributed to an individual face by making it dependent on recently seen expressions. To test this, we showed subjects a variety of facial expression morphs ranging from happy to sad to angry and found a serial dependence in perceived expression. Each facial expression (target emotion) was shown for 250 ms after which subjects matched the expression of a test face to the target emotion they had just seen. Subjects made consistent errors when reporting the perceived expression of the target face, seeing it as more similar to the facial expressions presented on the previous two trials (up to 12 seconds previously). In a second experiment, we showed subjects a variety of emotional expression morphs for different identities. On each trial, subjects were presented randomly with one of two possible identities (a male or a female face). Results showed stronger serial dependence in expression when the 1-back identity was the same as the current identity. Overall, the results indicate that perceived facial expressions at any one moment are biased towards recently seen expressions, especially within the same identity. The Continuity Field may therefore support apparent stability of emotional expressions by introducing perceived serial dependence in an identity-selective manner. Meeting abstract presented at VSS 2015