A Neural Implementation of MINERVA 2

Abstract

The MINERVA 2 (Hintzman, 1984) model of human memory has been used to simulate a variety of cognitive phenomena. These simulations, however, describe cognitive phenomena at Marr’s (1982) representation/algorithm level, with little effort to link the core assumptions of the model to an underlying neural implementation (however, see Kelly et al., 2017). This article describes a possible neural implementation of MINERVA 2—one that is simple and arguably biologically plausible. This implementation suggests a novel method for generating response latencies and provides a concrete example to support Marr’s claim that the representations and algorithms that mediate human performance in a variety of different cognitive tasks (e.g., decision making; Dougherty, Gettys, & Ogden, 1999) can be investigated and simulated without reference to their underlying neural implementation.