Preliminary comparison of four factor structures of facial emotion recognition in individuals with schizophrenia

Individuals with schizophrenia tend to have moderate-to-severe deficits in facial emotion recognition. To characterize the deficits and inform future interventions, measures with robust structural validity are needed. Four structures have been proposed: the 7-domain structure (happiness, sadness, anger, disgust, fear, surprise, and neutral), the 1-domain structure (an overall ability), the 2-domain structure (threatening and non-threatening faces, with surprise excluded), and the 4-domain structure (positive, negative, unexpected, and neutral). This study compared the four candidate structures of facial emotion recognition in individuals with schizophrenia.

Data from 351 Asian adults with schizophrenia were obtained from projects validating social cognition measures. Confirmatory factor analysis was used to examine the four candidate structures using odd-even and first-and-second-half item parceling approaches. Robust maximum likelihood estimation was additionally used as a sensitivity analysis.

Under maximum likelihood estimation, only the 7-domain structure achieved satisfactory model fit across parceling approaches (χ²/df = 1.8 and 1.2; comparative fit index = 0.977 and 0.995; root-mean-square error of approximation = 0.049 and 0.023; standardized root-mean-square residual = 0.032 and 0.028). The preference for emotion-specific scoring remained unchanged under robust maximum likelihood estimation and in additional comparisons restricted to the six facial emotion recognition categories common to all candidate structures.

Separate domain scores for each emotion may be preferable. Facial emotion recognition measures could include six basic emotions and neutral faces to support comprehensive emotion-specific score interpretation. However, these preliminary, population-dependent findings require validation in non-Asian populations.