Empirical and theoretical limits on lag recency in free recall

One widely accepted empirical regularity in free recall holds that when people successively transition from report of one list item to another, they prefer transitions across short lags (e.g., by reporting items from adjacent serial positions) to transitions involving large lags. This regularity has provided crucial support for the temporal context model (TCM), a model of the evolution of temporal context in episodic memory (Howard & Kahana, 2002a). We report a reanalysis of 14 data sets that shows that, contrary to the presumed preference for short lags, people often produce transitions with larger lags during recall. We show that these data cannot be accommodated by the TCM. We furthermore show that existing applications of the model have, for mathematical convenience, introduced assumptions that have circumvented its core principle of context evolution. When we instantiated the TCM as it was actually described, with a gradually evolving context, we found that its behavior qualitatively departed from that of the version currently implemented, but that the model was still unable to capture the nature of transitions in free recall. We conclude that the TCM requires further modification and development before it can explain the data that constitute its main source of support. Supplementary materials relevant to this article can be downloaded from the Psychonomic Society’s Norms, Stimuli, and Data Archive, www.psychonomic .org/archive.