Journal of psychophysiology vol:9 issue:2 pages:127-141
Extinction of a Pavlovian conditioned response has often been interpreted as reflecting a weakening or disappearance of a previously established associative link between the conditioned (CS) and the unconditioned stimulus (US). In this article, we first review different sources of independent evidence which allow us to definitively reject this ''extinction-implies-unlearning'' position. Some recent phenomena from the animal laboratory (e.g., renewal, reinstatement) suggest that the original CS-US relation is probably never unlearned, but is rather supplemented by new, additional knowledge implying that in some contexts or at some moments in time, the CS-US relation does not hold. This conclusion is corroborated and supplemented by neurophysiological evidence on the un-erasability of subcortically established CS-US connections. Secondly, in some paradigms an extinction procedure does not result in any demonstrable behavioral extinction, including an animal conditioned flavor preference paradigm, the human evaluative conditioning, and the human hidden-covariation detection paradigm. It is argued that this resistance-to-extinction phenomenon probably requires some form of a dual-process account of Pavlovian conditioning. In this context, the merits and shortcomings of a distinction in terms of intrinsic/extrinsic changes in CS value and in terms of sensory/affective levels of US representation are critically discussed. As an alternative approach, we propose a difference between two Pavlovian functional systems: the Expectancy System and the Referential System. Finally, a series of guidelines are provided for the construction of future theories of Pavlovian extinction.