Serial Order Depends on Item-Dependent and Item-Independent Contexts

We address four issues in response to Osth and Hurlstone's (2022) commentary on the context retrieval and updating (CRU) theory of serial order (Logan, 2021). First, we clarify the relations between CRU, chains, and associations. We show that CRU is not equivalent to a chaining theory and uses similarity rather than association to retrieve contexts. Second, we fix an error Logan (2021) made in accounting for the tendency to recall ACB instead of ACD in recalling ABCDEF (fill-in vs. in-fill errors, respectively). When implemented correctly, the idea that subjects mix the current context with an initial list cue after the first order error correctly predicts that fill-in errors are more frequent than in-fill errors. Third, we address position-specific prior-list intrusions, suggesting modifications to CRU and introducing a position-coding model based on CRU representations to account for them. We suggest that position-specific prior-list intrusions are evidence for position coding on some proportion of the trials but are not evidence against item coding on other trials. Finally, we address position-specific between-group intrusions in structured lists, agreeing with Osth and Hurlstone that reasonable modifications to CRU cannot account for them. We suggest that such intrusions support position coding on some proportion of the trials but do not rule out CRU-like item-based codes. We conclude by suggesting that item-independent and item-dependent coding are alternative strategies for serial recall and we stress the importance of accounting for immediate performance.

Automated summary

In response to Osth and Hurlstone's (2022) commentary, this article addresses four issues related to the context retrieval and updating (CRU) theory of serial order proposed by Logan (2021). The authors clarify the relationships between CRU, chains, and associations, highlighting that CRU relies on similarity rather than association for context retrieval. They also correct an error in Logan's (2021) explanation for the tendency to recall ACB instead of ACD in the ABCDEF sequence. Furthermore, the article discusses position-specific prior-list intrusions and provides modifications to CRU, introducing a position-coding model based on CRU representations. The authors suggest that position-specific between-group intrusions in structured lists cannot be accounted for by reasonable modifications to CRU, but these intrusions still support position coding on some trials without ruling out CRU-like item-based codes. The article emphasizes the importance of considering immediate performance and presents item-independent and item-dependent coding as alternative strategies for serial recall.