Super and deepened-extinction in human predictive learning and a comparison of associative models.

Abstract

Cue-exposure is a treatment (e.g. for addictions and phobias) that aims to extinguish conditioned responses to target cues. However, especially in the case of addiction, relapse still occurs after cue-exposure and this may be due to recovery of conditioned responses outside of the extinction context. Super-extinction and deepened-extinction are two compound-cue extinction procedures which have been assessed for their capacity to produce more robust extinction than standard single-cue extinction procedures. We carried out further assessment of super and deepened-extinction protocols but found no evidence that they produced less response recovery compared to single-cue extinction. Contrariwise, super-extinction actually produced more recovery than the other two conditions. These results can be understood in terms of configural associative models (configural Rescorla-Wagner and Pearce configural model) but not in terms of the simple elemental Rescorla-Wagner model. Furthermore, the configural models provided better fits to overall data, and the Pearce configural model was better than the configural Rescorla-Wagner model.