Evidence of absence: no relationship between behaviourally measured prediction error response and schizotypy

Abstract

Introduction: The predictive processing framework has attracted much interest in the field of schizophrenia research in recent years, with an increasing number of studies also carried out in healthy individuals with nonclinical psychosis-like experiences. The current research adopted a continuum approach to psychosis and aimed to investigate different types of prediction error responses in relation to psychometrically defined schizotypy.

Methods: One hundred and two healthy volunteers underwent a battery of behavioural tasks including (a) a force-matching task, (b) a Kamin blocking task, and (c) a reversal learning task together with three questionnaires measuring domains of schizotypy from different approaches.

Results: Neither frequentist nor Bayesian statistical methods supported the notion that alterations in prediction error responses were related to schizotypal traits in any of the three tasks.

Conclusions: These null results suggest that deficits in predictive processing associated with clinical states of psychosis are not always present in healthy individuals with schizotypal traits.

Keywords: Sensory prediction; associative learning; prediction error; psychosis continuum; reversal learning.

Figure 1.

Comparison between mean difference of active and passive forces in the Finger and Slider conditions. The dotted line indicates perfect performance.

Figure 2.

Low and high group ratings for each stimulus in the test stage for O-LIFE unusual experiences (A) and introvertive anhedonia (B).

Figure 3.

Low and high group ratings for unusual experiences across learning stages. + and − refer to the presence or the absence of the outcome, respectively.

Figure 4.

Low and high group ratings for introvertive anhedonia cross learning stages. + and − refer to the presence or the absence of the outcome, respectively.

Figure 5.

Accuracies of true reversal trials (A) and probabilistic error trials (B).

Figure 6.

Results from Bayesian Correlation Pairs analyses. Panels A, B and C show results for the force-matching, Kamin blocking and reversal learning tasks, respectively. CI: credibility interval; BF: Bayes factor.