The role of reduced working memory storage and processing resources in the associative memory deficit of older adults: simulation studies with younger adults

Abstract

Previous research indicates that relative to younger adults, older adults show a larger decline in long-term memory (LTM) for associations than for the components that make up these associations. The purpose of the present study was to investigate whether we can impair associative memory performance in young adults by reducing their working memory (WM) resources, hence providing potential clues regarding the underlying causes of the associative memory deficit in older adults. With two experiments, we investigated whether we can reduce younger adults’ long-term associative memory using secondary tasks in which either storage or processing WM loads were manipulated, while participants learned name–face pairs and then remembered the names, the faces, and the name–face associations. Results show that reducing either the storage or the processing resources of WM produced performance patterns of an associative long-term memory deficit in young adults. Furthermore, younger adults’ associative memory deficit was a function of their performance on a working memory span task. These results indicate that one potential reason older adults have an associative deficit is a reduction in their WM resources but further research is needed to assess the mechanisms involved in age-related associative memory deficits.

Keywords:

• associative memory deficit
• age and episodic memory
• working memory
• simulation in younger adults
• storage and processing resources

Acknowledgments

The authors would like to thank Angela Kilb for her comments and suggestions on earlier drafts, and members of the Memory and Cognitive Aging Laboratory at the University of Missouri for their contributions to these studies.

Notes

1. A related line of research is one that looked at the effects of age and of reduced attentional resources on features and their bindings in short-term memory and WM. In contrast to the above-mentioned studies, which dealt with binding in long-term memory, the studies used in this line of research were based on memory performance within a few seconds of the original study of the information. Several of these studies found that older adults show a binding/associative deficit in WM retention tasks (e.g., Chen & Naveh-Benjamin, 2012; Cowan et al., 2006; Hartman & Warren, 2005; Mitchell et al., 2000).

The research on the effects of divided attention on binding in WM resulted in somewhat less clear a picture. Whereas some studies show a larger decline in memory for associations (binding) than for the features in WM as a result of the introduction of a secondary task (e.g., Brown & Brockmole, 2010, Ex. 1; vanRullen, 2009), others have shown no such differential decline (e.g., Allen, Baddeley, & Hitch, 2006; Allen, Hitch, Mate, & Baddeley, 2012; Baddeley, Allen, & Hitch, 2011). These are interesting studies that created some theoretical controversy regarding the structure of WM (for example, the potential existence of an episodic buffer and whether or not binding requires effortful processes in the episodic buffer or takes place automatically prior to information entering the buffer). However, these studies are somewhat less relevant for the research question addressed in the current manuscript regarding whether interrupting different WM mechanisms affects differentially items and their binding in LTM, for several reasons. First, the materials used in these studies are quite different than those used in LTM research, including those one used in the current studies. Specifically, the WM studies mostly concentrate on memory for intra-item binding of features, for example, the binding of item and its color or spatial location, whereas the LTM studies use much more complex material that involve in many cases inter-item binding, for example the binding of face and name or of two unrelated words. Second, whereas LTM studies try to separate out the effects of DA at encoding from those at retrieval, in most WM studies the secondary task is used throughout the trial, affecting encoding, maintenance, and retrieval. Third, although both types of studies look at the effects of DA on memory, there are potentially important variables that might result in diverging results in the two domains. For example, any patterns of performance in memory for features and their bindings in WM may change considerably in LTM due to consolidation and other processes that occur during the retention interval, which is much longer in LTM studies.

2. The sample of younger adults used here is typical of studies comparing younger and older adults. The categorization of younger adults in these experiments as “low span” is based on a standardized automated version of the O-span task used in other studies, and the cutoff points we have used for defining “low span” and “high span” individuals, as well as the mean scores for each group, are very similar to those used in the literature, at least in samples of university students (Unsworth et al., 2005). Also, Unsworth, Redick, Heitz, Broadway, and Engle (2009) found that participants who processed information more successfully remembered items better than those who processed information less successfully. Their results showed that there was no trade-off between processing and storage, meaning that participants do not use a strategy to sacrifice their online math performance for better storage performance or vice versa. Thus they suggest that excluding data for participants who scored below 85% on O-span math accuracy is not necessary. Based on their finding, we kept those who scored lower than 85% on O-span math accuracy in our data for further analyses.

3. In the WM literature, participants that score below .85 on O-span math accuracy are typically excluded from the analyses. However, we were not concerned with the validity of the standard “O-span score” measuring stored items in Experiment 2. Instead, we used math accuracy as a tool for investigating item and associative performance in the name–face task, which is why we included all participants in the analyses.