The neural signature of language-based prediction errors in individuals with psychotic-like experiences

Background:
Hallucinations occur in many psychiatric disorders, such as psychosis or schizophrenia. Still their underlying mechanisms pose many questions. The theory of Predictive Coding (PC) offers an explanation. PC assumes an imbalance in the weighting of prior knowledge and incoming sensory (i.e. prior weight) information in psychosis, which leads to an aberrant prediction error processing. Former studies revealed that the neural signature of reward-based prediction errors is different in psychosis patients compared to healthy population. Although language functions are disrupted in individuals with psychosis, it is unclear whether the processing of language-based prediction error is altered.

Methods:
We used an fMRI language paradigm, in which sentence entropy, intelligibility and surprisal of the sentence final word was varied. This allowed us to measure the prior knowledge and the prediction error, as well as the computational analysis of the prior weight. We collected data form individuals with (N=10) and without (N=10) psychotic-like experiences (Data collection ongoing).

Results:
The preliminary results revealed group differences in language-based prediction errors in the left superior temporal gyrus (STG). The activation in the STG was negatively linked to symptom strength and to prior weight, indicating that those individuals with stronger symptoms show a reduction in the language-based prediction error signal and an overweighting of prior information.
Taken together, these preliminary results show for the first time alteration in language-based prediction error processing in the STG in individuals with psychotic-like experiences.