A new model for retrieval from memory: Iterative resonance applied to recognition memory and serial reaction-time tasks

Douglas J. K. Mewhort

I will sketch a new theory of retrieval based on a resonance metaphor and show that it captures data from recognition memory and from serial reaction-time tasks. According to the model, when a retrieval probe is presented, it resonates with items stored in memory in proportion to their similarity to the probe, and an echo of the resonant information is formed. If the echo does not provide clear evidence, further comparisons are calculated. In the subsequent comparisons, information taken from memory is sharpened, and the process cycles until clear evidence is obtained. RT is a function of the number of iterations required to provide the necessary evidence. Using the model, I will show, by simulation, that complex rule-like behaviour can be produced from a structured record of events without applying formal rules.