Prediction error generated by extinction attenuates overshadowing in humans

Abstract

Prediction error is a discrepancy between expected and actual outcomes in each situation (Torrents-Rodas et al., 2021). Experiencing extinction causes a rise in prediction error, which produces attention to the context, and thus context dependency. Behavioral measures of attention, such as responses to specific elements of a compound stimulus (Reynolds, 1961), have been previously proposed and can add another measure of the effect of extinction on attention and an insight into how attention changes when extinction occurs. The present experiment aimed to study the effects of prediction error produced by extinction on overshadowing through behavioral measures. Forty-eight participants were randomly assigned to four groups (n = 16) and trained to respond to Sample Stimulus (SS): Y, in a matching to sample task during a first phase. In the second phase, two groups experienced the extinction of SS: Y, while the other groups did not. All groups were trained to emit R1 responses to either compound SS: AX or SS: X during this phase. The test phase presented elements A and X separated. Results showed overshadowing when the extinction of SS: Y was not experienced, where the percentage of R1 responses was greater for A than X. However, when the extinction of SS: Y was experienced, both elements received similar responses showing an overshadowing attenuation. These results suggest that extinction produces an attentional change to X. This effect appears to be related to a rise in prediction error during extinction, which produces changes in attention.