Age-Related Differences According to the Associative Deficit and the Environmental Support Hypotheses: An Application of the Formal Charm Associative Memory Model

Abstract

Background/Study Context: According to both the associative deficit hypothesis (ADH; Naveh-Benjamin, 2000, Journal of Experimental Psychology: Learning, Memory, and Cognition, 26, 1170–1187) and the environmental support hypothesis (ESH; Craik, 1983, Philosophical Transactions of the Royal Society of London Series B, 302, 354–359), memory decline with aging can be seen as an impairment of the self-initiated associative memory processes such that supportive encoding and/or retrieval can reduce age-related differences. A formalization of relationships between the ADH and ESH was investigated using the distributed memory model “CHARM” (Composite Holographic Associative Recall-Recognition Model; Metcalfe, 1982, Psychological Review, 89, 627–661; Metcalfe, 1991, Psychological Review, 98, 529–543).

Methods: Empirical data were collected in young and elderly participants on cued recall and recognition tests according to both the level of processing (LOP: phonetic vs. semantic tasks) and the self-generated cueing (elaboration effect: provided vs. self-generated cue) manipulation. These data were compared with those from CHARM simulations that were designed to evaluate the impact of deteriorated associative processes (i.e., ADH) and the role of LOP and elaboration effects (i.e., ESH) in memory performance.

Results: The simulated data were largely consistent with the empirical data, showing that the impairment of associative processes in the CHARM model was accompanied by an increased need for environmental support at encoding (interaction between age, LOP, and elaboration) to reduce memory decline in cued recall tasks, which is somewhat observed in the recognition scores.

Conclusion: The overall results from CHARM simulations are in accordance with both the ADH and ESH hypotheses and provide discussion on the formal connections between these two main aging explanations.

Notes

¹From this, in numerous simulations by Metcalfe, items are composed of 63 features and pairs of semantically and phonemically related items are represented by pairs of items sharing 54 (or 36) and 27 common features, respectively (Metcalfe, 1982, 1985, 1990, 1991, 1993).

²For more rational arguments regarding the parallel between LOP and the associative strength principles, see Metcalfe (1985). Also see Craik and Lockhart (1986) as well as Lewandowsky and Hockley (1987) for discussions about this parallel and notably on the fact that there is no place in the CHARM model for the qualitative distinction between the different processing levels or record “domains.”