Overly large prediction errors reduce expectation change - A replication study

Mental disorders have been related to aberrations in using novel positive information to revise established negative beliefs. Previous research used to make a binary distinction between belief-confirming vs. -disconfirming information, but recently it has been examined how varying levels of positive information is used to update beliefs. The present study aimed to replicate a recent finding suggesting that positive information that deviates to a large extent from people’s prior expectations raises doubts about the credibility of new information and therefore hardly leads to change in expectations. In a heterogenous sample (N = 144), participants were provided with slightly positive, moderately positive, or extremely positive information in relation to their prior expectations about other people’s behaviour. Replicating previous research, the present study found that expectation change was greatest for moderately positive information. It also provided evidence for a possible mechanism underlying the inverse U-shaped relationship between the positivity of new information and change in expectations: Extremely positive information was devalued through defensive cognitive strategies, referred to as cognitive immunisation. Unlike traditional learning models, the present results suggest a tipping point above which the discrepancy between expectation and outcome is suspiciously large, so that the degree of expectation change decreases.