Abstract
Previous work has shown that trait anxiety is associated with interpretative biases in the perception of facial expressions, specifically an increased tendency to judge ambiguous facial expressions as fearful. Using functional magnetic resonance imaging and both univariate and multivariate analysis techniques, we investigated the neural correlates of these biases in perception. We focused specifically on neural regions previously implicated in face processing: the superior temporal sulcus (STS), amygdala and fusiform face area (FFA). Subjects were presented with pictures of faces that showed one of three pure emotional expressions (fear, sad, surprise) or intermediate morphs between these same expressions. These expressions were selected based on previous research indicating that these dimensions are the ones where anxiety-related perceptual biases are most likely to be observed. BOLD response parameter estimates were calculated using univariate regression and multivariate pattern analysis. No anxiety-related differences were observed in the univariate analysis. For the multivariate analysis, a linear classifier was used to quantify (dis)similarities between neural representations of intermediate morphs and those of end-point pure expressions. It was predicted that for the two continua containing fear, individual differences in trait anxiety would modulate the extent to which neural representations of intermediate morphs showed greater similarity to pure fear than the other constituent expression. Bilateral regions of interest were investigated focusing on the superior temporal sulcus (STS), amygdala and fusiform face area (FFA). Results from this preliminary study indicated that anxiety-related biases in the neural representations of ambiguous expressions containing some percentage of the expression fear were predominantly observed in STS, with high trait anxious individuals showing reduced distances between the neural representations for these morphs and those for pure fear. Additional data from a follow-up experiment will be presented. Results are discussed in the context of content-based models of anxiety.