A reconstructive theory of serial-order memory

Theories of serial-order memory either include specialized mechanisms to represent order or claim that sequences of events are represented in memory by structures that are constructed during encoding. Unfortunately, both approaches have many shortcomings, and neither can explain how the order of arbitrary event sequences (i.e., sequences that are not intentionally learned; e.g., trips to the grocery store) are remembered. This paper presents an alternative approach to serial-order memory, one in which order is not directly represented in memory but is instead reconstructed during retrieval. This reconstructive theory is presented in two parts. The first demonstrates how the components of episodic memory (as implemented in the Search of Associative Memory model of Raaijmakers & Shiffrin\u27s, 1981b) that are necessary to encode, represent, and retrieve information about events in memory are sufficient to account for key aspects of serial-order memory. Several emergent properties of the reconstructive model allow it to explain phenomena that have been problematic for other theories of serial-order memory (e.g., recency effects). The second part of this paper focuses on how contextual information is used to reconstruct memory for order. Four experiments that examine the role of context in serial-order memory are presented. Experiment 1 evaluated three assumptions of the Raaijmakers and Shiffrin (1981b) model; the results support the model and indicate that people (1) attend to and remember the contexts in which items are presented, (2) represent contextual changes in memory, and (3) are better at remembering the contexts of items near the beginnings of sequences and following changes in context. Experiment 2 extends these results by showing that people use contextual information to make judgments about the relative order of two items. Experiment 3 replicates this result using a slightly different paradigm. Finally, Experiment 4 attempted to find out whether or not people use the most general contextual information available to make order judgments. The results of Experiment 4 are less conclusive than those of Experiments 1-3; several sources of this discrepancy are discussed