The effect of degree of prediction error elicited by retrieval on the reconsolidation of fear memory

The retrieval-extinction paradigm based on memory reconsolidation has been proposed as a non-invasive approach to attenuating fear, though its effectiveness remains debated due to inconsistent findings in the literature. Current research suggests that qualitative prediction error (“PE”) serves as a key boundary condition for determining whether a memory can undergo reconsolidation. However, these findings are based on qualitative (type and number) analysis, highlighting the need for further quantitative investigation. Drawing on previous experiments, this study manipulated prediction errors by altering the pairing patterns of CS-US during the retrieval and acquisition phases, setting up five groups with different prediction error manipulations. A reinforcement learning model was used to quantify prediction error (PE), fitting subjective expectancy ratings to a simplified Rescorla-Wagner model, and calculating the actual prediction errors elicited under different retrieval manipulations, combined with skin conductance response to reflect the intervention effects on fear memory in each group. Our findings indicate that different retrieval PE manipulations led to significant between-group differences in skin conductance response indicators during the fear reinstatement test, and the actual type and number of PE elicited were inconsistent with their operational definitions. The overall PE degree elicited during the retrieval phase may be a combination of the size, type, and number of PE. This study can help to further clarify the key role of PE in the retrieval-extinction paradigm, thereby promoting the clinical translation of this paradigm.