A numerical exploration of a stochastic model of human list learning

Abstract

A simple network which shows primacy and recency effects is presented. The model uses stochastic updating of clipped weights to produce a range of different memory behaviours. The model, originally proposed by Kahn, Wong and Shewington (1991, 1995), shows a much wider range of behaviours than originally predicted. These behaviours depend on the probability of updating weights, initial non-zero weights, type and degree of dilution.