When two events occur closely in time, an “association” exists between memories for those events. When a pair of associ- ated events is semantically similar, it is easier to recognize the complete pair and easier to tell the complete pair apart from pairs of events that did not co-occur; there is also, however, a bias to report that similar events had co-occurred, even when they had not. A new experiment shows that these phenomena occur whenever two events share features, whether those fea- tures are perceptual or conceptual in nature and whether the events themselves are verbal or non-verbal. We present a dy- namic model for storage and recognition of associations that shows how all these results can be explained by the princi- ple that shared features lead to correlated processing of similar events, which in turn increases capacity to process associative information.

Using hierarchical Bayesian estimation of RT distributions, wepresent a novel application of Systems Factorial Technology(Townsend & Nozawa, 1995) to the retrieval of item and asso-ciative information from episodic memory. We find that itemand associative information are retrieved concurrently, withpositive memory evidence arising from a holistic match be-tween the test pair and the contents of memory, in which bothitem and associative matches are pooled together into a sin-gle source. This retrieval architecture is inconsistent with bothstrictly serial processing and independence of item and asso-ciative information. Pooling of item and associative matchesimplies that while item and associative information may beseparable, they are not qualitatively different, nor are quali-tatively different processes (e.g., familiarity vs. recollection)used to retrieve these kinds of information.