Cognitive dissociation of memory systems: Implicit and explicit memory in pigeons (Columbia livia)

Four experiments investigated implicit and explicit memory in pigeons. Several ways of defining implicit memory were considered, and 2 varieties of tasks widely agreed on as requiring implicit memory in humans, involving grammar learning and sequence learning, were modified to be performed by pigeons. In experiments 1 and 2, birds learned to categorize strings of alphanumeric according to their adherence to the rules of synthetic grammars. In Experiment 1 birds categorized character strings that had been generated by a finite state Markovian grammar and character strings which broke the rules of this same grammar. In Experiment 2, birds categorized character strings that had been generated by the grammar used in Experiment 1 and character strings that had been generated by a second artificial grammar. The results of Experiments 1 and 2 indicate that birds learned to accurately categorize character strings generated by artificial grammars, but performed poorly at categorizing nongrammatical character strings, a form of the feature-positive effect. A connectionist model having 2 memory systems which mimic the characteristics of implicit and explicit memory simulated the general pattern of responding displayed by birds in Experiments 1 and 2. The categorization performance of birds was most closely reproduced by the portion of the model designed to reflect important characteristics of implicit memory. Experiments 3 and 4 demonstrated implicit memory based on the learning of sequences of responses. In Experiment 3, birds responded faster to cues which appeared as part of a repeating pattern than to randomly generated cues. In Experiment 4 birds performed a task involving both implicit and explicit measures of performance. Implicit memory was involved in responding to repeating cue locations, as in Experiment 3, and explicit memory was involved in a second phase of the task which required memory for a specific response within the previously executed sequence of pecks. Furthermore, 2 task manipulations in Experiment 4 produced a double dissociation, affecting 1 of the 2 memory systems more than the other. Under a randomly structured condition, birds displayed slower response times, yet retained the ability to recall responses at greater than chance levels. Conversely, shorter exposure durations impaired recall of previous responses but did not affect response times. These experiments suggest that separate implicit and explicit memory systems are required for a comprehensive theory of nonhuman animal memory.