Reinterpreting selective impairments in memory: computational and empirical simulations of dissociations in amnesia

Abstract

By a dominant account, memory is composed of multiple storage systems, each operating according to unique principles. By an alternative account, memory is a single storage system and operates according to a single set of principles. Selective memory impairments in amnesia serve as the primary evidence for the multiple-system perspective. This thesis reports a critical appraisal of the multiple-system perspective using a combination of computational and empirical methods. In the computational analysis, I adopt the Holographic Exemplar Model, a single-system model of memory based on Hintzman’s (1986) classic MINERVA2 model. I simulate amnesia by manipulating the quality with which items are encoded in memory. In the empirical analysis, I simulate amnesia by manipulating peoples’ quality of encoding by limiting the time given to study stimuli. Simulations 1-2 and Experiments 1-2 simulate a dissociation between classification and recognition. All four analyses are consistent with the original results. Simulation 3 and Experiment 3 simulate single and double dissociations between tachistoscopic identification and recognition. The analyses were consistent with the single but not double dissociation. Simulation 4 and Experiment 4 simulate a dissociation among word-stem completion, cued recall, and recognition. Both analyses were only partially consistent with the original results, representing a failure overall. Simulation 5 and Experiment 5 derived a novel prediction from artificial grammar learning, predicting a non-dissociation between string completion and recognition. The mixed results provide some support for a single-system account of memory and opens opportunities for future work. I argue that the analysis is best considered in convergence with previous work moving toward a more integrated account of memory