Mathematical and connectionist models of human memory: a comparison

Associative symmetry and memory theory

This article reexamines the theory and data concerning two opposing views of episodic association. The independent association hypothesis (IAH) sees associations as unidirectional and separately modifiable links between individual item representations. The associative symmetry hypothesis (ASH) sees an association as a holistic conjunction of the constituent items, blending the elements of their representations into a new one. In early tests of these hypotheses, experimenters compared forward and backward recall of paired associates, looking for asymmetries. Although some studies showed significant differences between forward and backward recall, the vast majority did not. The author of this study used a mathematical analysis of distributed memory models to reexamine this classic question. These analyses revealed that symmetric and asymmetric models can mimic each other, offering identical predictions regarding forward and backward recall. To distinguish these models, the author examined the correlation between forward and backward recall at the level of individual pairs of items. Both this correlation and the correlation between recall of pairs tested in the same direction were near unity. These results provide new evidence favoring the ASH over the IAH.

An autoassociative neural network model of paired-associate learning

Hebbian heteroassociative learning is inherently asymmetric. Storing a forward association, from item A to item B, enables recall of B (given A), but does not permit recall of A (given B). Recurrent networks can solve this problem by associating A to B and B back to A. In these recurrent networks, the forward and backward associations can be differentially weighted to account for asymmetries in recall performance. In the special case of equal strength forward and backward weights, these recurrent networks can be modeled as a single autoassociative network where A and B are two parts of a single, stored pattern. We analyze a general, recurrent neural network model of associative memory and examine its ability to fit a rich set of experimental data on human associative learning. The model fits the data significantly better when the forward and backward storage strengths are highly correlated than when they are less correlated. This network-based analysis of associative learning supports the view that associations between symbolic elements are better conceptualized as a blending of two ideas into a single unit than as separately modifiable forward and backward associations linking representations in memory.