Bridging the Gap Between Measurement Models and Theories of Human Memory

Abstract

Two types of formal memory models can be distinguished: measurement models and computational models. Measurement models aim at measuring memory processes. In contrast, computational models focus on explaining the structures and processes underlying memory performance. Whereas measurement models involve a statistical framework for testing the model's assumptions, computational models often cannot be applied to empirical data directly. On the other hand, measurement models usually lack psychological theory, whereas computational models are formalized theories allowing for precise predictions. Using the computational model MINERVA2 as an example, it is shown that the gap between both types of formal models can be bridged quite easily by developing closed-form equations for the model's predictions. The resulting multinomial model combines the advantages of measurement models and computational models. The benefits are demonstrated by applying the model to two experiments on the global similarity effect and the list-length effect. The MINERVA 2-inspired multinomial model provides an excellent fit to the data, a coherent explanation of the effects, and statistically sound measures of the processes involved.