Temporal stability of semantic predictions in subclinical autistic and schizotypal personality traits

Language impairments are core symptoms of both schizophrenia and autism spectrum disorders and have been linked to deficits in predictive language processing. While altered use of semantic predictions have been reported in both conditions, little is known whether semantic predictions are stable over time. The goal of this study was therefore to investigate the temporal stability of semantic prior beliefs focusing on individuals with schizotypal and autistic traits. 115 participants, assessed for subclinical schizotypal (SPQ_5ls; mean = 77.99, SD = 39.31) and autistic traits (AQ; mean = 15.67, SD = 6.01), completed an auditory stability paradigm at two timepoints to investigate the temporal stability of semantic predictions. At timepoint one, consisting of one session, participants listened to 240 sentence beginnings varying in predictability (e.g., high: "The swimmer jumped into the…"; low: "The child hid the toy under the…") and provided a prediction for each sentence-final word. Timepoint two, consisting of two sessions, each session comprising of 120 old and 120 new sentences. In addition to final-word predictions, sentence recall was assessed to examine the influence of memory on prediction stability. Generalized linear mixed models revealed that higher predictability led to greater temporal stability of semantic predictions. Importantly, increasing schizotypal and autistic traits were associated with reduced stability, particularly in highly predictable contexts where stable predictions typically facilitate efficient language processing. While poorer sentence recall was linked to greater instability, especially in medium- and low-predictability contexts, it did not account for the reduced stability observed in relation to schizotypal and autistic traits. These findings suggest that individuals with higher schizotypal and autistic traits struggle to form stable, lasting semantic predictions, which may contribute to difficulties in efficient language processing.