Single Parameter Model for Free Recall And the Nature of Partially Filled Working Memory

I present a single parameter model of free recall and fit the one parameter, the probability per time unit of an item in working memory entering the next memory store (similar to Atkinson and Shiffrin, 1968), to the original Murdock (1962) data. Working memory is modeled as having space for a maximum of 4 items (Cowan, 2001). The first four probability values convey precise information about how items in the partially filled working memory enter the next memory store. In particular, one can distinguish between four separate “sub-stores ” and a single store model. Two alternatives follow for the probability of an item in the initially partially filled working memory store to enter the next memory store: the same probability as a single item in the filled working memory or a probability proportional to the total space that exists in working memory. I find that the latter alternative is a better fit (though not a perfect fit), suggesting that working memory is not divided into four separate sub-stores. It is suggested that new high statistics (low noise) experiments on short lists will either settle the issue or prove theory incorrect