We propose a generative model of PA learning based upon the distributed memory model frameworks proposed by the matrix model [3, 4] and convolution-correlation memory models [5] of associative learning. To further define our model we employ the brain-state-in-a-box model [6, 7] as our deblurring mechanism to increase ecological validity of our model, as opposed to a heuristic such as the winner-take-all choice rule. Here we model how item- and association-memory manipulations may modulate within memory performance in a cued recall task. In particular, we ask whether manipulations of material-type can passively result in stronger associations (i.e., without requiring the participant to vary their strategy). For example, current modeling results demonstrate how items that are learned stronger during study can result in better association-memory.

This modeling approach represents a framework for a range of PA learning effects that have already been reported (e.g., [1, 2]) as well as predicting as-of-yet unobserved patterns. Simulations of how manipulations of material-type can modulate item- and association-memory have not yet been theoretically explored and can have profound implications to current memory models.