Testing the predictions of a distinctiveness model of memory: The production effect in backward recall

Individuals can engage in a variety of study strategies that influence later recall. The effects can be complex and are often diagnostic of underlying memory processes. For example, producing items by reading them aloud yields a crossover pattern: Early list items are recalled more accurately when studied silently, whereas later items benefit from being read aloud. When aloud and silent items alternate within a list, this contrast produces a characteristic sawtooth serial-position curve in which aloud items form the peaks and silent items the troughs. According to the revised feature model, these interactions arise from encoding mechanisms: Production both disrupts rehearsal-primarily affecting early serial positions-and adds distinctive features that benefit later items. Crucially, because these mechanisms operate prior to retrieval and do not depend on how recall is initiated, the revised feature model predicts that the interaction between production and serial position should appear regardless of whether recall proceeds forward (from first to last) or backward (from last to first). That is, changing the output order should not undo encoding-based differences that have already shaped the memory traces. In six experiments, we tested this a priori prediction by manipulating recall direction in immediate serial recall. Consistent with expectations, the interactions associated with the production effect were observed in both forward and backward recall, despite the substantial performance differences typically observed between the two tasks. These findings are consistent with the encoding-based predictions of the revised feature model, and the model provides a principled account of the patterns observed across the experiments. (PsycInfo Database Record (c) 2026 APA, all rights reserved).