An Analysis of Immediate Memory: The Free-Recall Task

Performance in simple memory tasks is a standard part of intelligence tests. In spite of its ability to predict wide ranging aspects of intellectual activity, a detailed account of how people carry out memory tasks eludes current theory. In the free-recall version of the task, subjects are asked to study a list of items and then to report as many as possible in any order. Performance in the task is usually reported by tallying accuracy as a function of the position in which each item appeared in the study list, the so-called serial-position curve. We present representation, encoding, and retrieval equations that specify a stochastic model of the processing that people carry out when performing the task. One archival source lists data for undergraduate subjects with 80 trials per subject. We obtained excellent fits when we applied the encoding and retrieval equations to the archival serial-position data. In addition, we compared the model’s behaviour against the archival data on three additional measures: the order in which items are reported, the number of items reported correctly per trial, and the frequency of unstudied items reported in error. The model correctly anticipated the new measures without a change in its parameters, i.e., the new measures were obtained for free.