Abstract
To explore the encoding of facial expression a facial expression ‘space‘ was constructed by averaging faces to produce a gender and identity neutral face for each expression, and an overall average face (the norm). We then made anti-expressions by morphing from an average expression through the norm, and investigated the aftereffects produced by these anti-expressions. We adapted participants to an anti-expression then showed them the norm and asked them to indicate which of the 6 expressions (excluding neutral) best described the norm. Participants consistently selected the expression that matched the anti-expression to which they had been adapted (i.e. adapting to anti-sad resulted in the norm appearing sad). These aftereffects may be the result of the way the stimuli were constructed, or may be an indication that facial expression is encoded with reference to a prototype, similar to the way recent evidence suggests we encode facial identity. To explore the latter possibility we adopted a technique previously used to study facial identity and measured the effect of adaptation along two trajectories. One trajectory went from an average expression (happy) to the matching anti-expression (anti-happy), passing through the norm. The other trajectory went from the same expression (happy) to a non-matching anti-expression (anti-fear), and therefore did not pass through the norm. Adaptation is believed to cause a shift in the norm, so a greater effect of adaptation observed for the trajectory passing through the norm would suggest facial expression to be encoded with reference to the norm. Our results showed no differences between the two trajectories. We are currently investigating additional trajectories within expression space.