Trace and delay eyeblink conditioning: Contrasting phenomena of declarative and nondeclarative memory

We tested the proposal that trace and delay eyeblink conditioning are fundamentally different kinds of learning. Strings of one, two, three, or four trials with the conditioned stimulus (CS) alone and strings of one, two, three, or four trials with paired presentations of both the CS and the unconditioned stimulus (US) occurred in such a way that the probability of a US was independent of string length. Before each trial, participants predicted the likelihood of the US on the next trial. During both delay (n = 20) and trace (n = 18) conditioning, participants exhibited high expectation of the US following strings of CS-alone trials and low expectation of the US following strings of CS-US trials-a phenomenon known as the gambler's fallacy. During delay conditioning, conditioned responses (CRs) were not influenced by expectancy but by the associative strength of the CS and US. Thus, CR probability was high following a string of CS-US trials and low following a string of CS-alone trials. The results for trace conditioning were opposite. CR probability was high when expectancy of the US was high and low when expectancy of the US was low. The results show that trace and delay eyeblink conditioning are fundamentally different phenomena. We consider how the findings can be understood in terms of the declarative and nondeclarative memory systems that support eyeblink classical conditioning.