Psychosis Risk Is Associated with Increased Multimodal Reports of Socially Salient Content in Degraded Stimuli

Background and Hypothesis Humans are adaptively tuned to detect faces and speech in noisy environments. While such detection is generally adaptive, we hypothesize that hallucinations and perceptual distortions arise from overdetection of such features under uncertainty. Study Design To test this hypothesis, we measured face and speech detection rates in over 800 adolescents and young adults, oversampled for individuals at risk of developing psychotic disorders. Study Results Individuals who reported seeing more faces in degraded two-tone images tended to report hearing more speech in degraded filtered audio clips (r(765) = 0.32, p < .001). Recognition rates from both tasks independently predicted clinician-rated positive symptoms (face detection: r(784) = 0.19, p < .001, 95% CI [0.12,0.26]; speech detection: r(818) = 0.16, p < .001, 95% CI [0.1, 0.23]). However, a composite score exhibited a stronger association than either task individually (r(760) = 0.23, p < .001, 95% CI [0.16,0.29]). This composite score was uniquely predictive of perceptual abnormalities over-and-above other types of positive symptoms such as unusual thought content, suspiciousness, grandiose ideas, and disorganized communication (b = 0.066, t(781) = 3.01, p = 0.003, partial r^2 = 0.011). Conclusions Our results suggest that a supramodal tendency to report socially salient signals under uncertainty may contribute to positive perceptual symptoms and index clinical risk for psychosis. We place these findings within the broader literature and discuss the potential utility of these highly scalable behavioral measures for better identifying individuals at risk for developing psychosis.