Individuals with schizophrenia demonstrate atypical facial emotion recognition. However, inconsistent findings in the literature highlight the limitations of standard paradigms that rely on fixed, stereotypical facial configurations. The present study employed a novel computational tool to examine internal representations of facial emotions — defined as individual expectations of how emotions appear on the face. Twenty-eight patients with schizophrenia and 25 healthy controls generated facial expressions of happiness, fear, and anger on a photorealistic avatar through an iterative selection process, converging on ideal expressions for each target emotion across 8 iterations of 10 samples per iteration. Individualised models capturing the range of facial configurations associated with each emotion category were constructed from the selected expressions. No significant group differences were observed in the number of expressions selected, the breadth of expressions deemed representative of each emotion (spread), or the discriminability of emotion categories (d-prime). Both groups demonstrated greater difficulty distinguishing fearful from angry expressions relative to distinguishing either from happy expressions. Notably, patients exhibited significantly greater within-group centroid dispersion, indicating that their internal representations were more variable and less similar compared to controls. This suggests that patients' internal representations of facial emotions are more heterogeneous, potentially reflecting less shared understanding of facial features that define each emotion category. These findings offer a novel, representation-based account of emotion recognition in schizophrenia. Rather than a uniform perceptual deficit, the results indicate greater variability in internal representations of facial expressions, which may disrupt emotion recognition and contribute to social communication difficulties.
Song et al. (Mon,) studied this question.