Exploring the Interaction between Working Memory and Long-Term Memory: Evidence for the Workspace Model

During the past three decades, the field of working memory has made significant progress, which is reflected in the large range of models of working memory, each with different scopes and emphases. It is now clear that working memory is fundamental for complex cognitive tasks such as learning, reasoning and comprehension. Current interest focuses strongly on the interaction of working memory with long-term memory, as it has become clear that models of working memory alone are incapable of capturing some of our complex cognitive abilities. Most models of working memory acknowledge a close relationship between working memory and long-term memory, but have contrasting views about how this interaction is implemented. In my PhD research I have compared three broad classes of models, each proposing a different flow of information through the memory system. The first model proposes that working memory acts like a gateway for perceptual input on its way to long-term memory. In the unitary model, working memory is seen as comprising the activated portion of long-term memory. The workspace model views working memory as a workspace that is separate from, and deals with the activated contents of long-term memory. The main aim of my thesis was to address the differences between these three models experimentally.

Dual-task experiments

In one set of experiments we employed a dual-task paradigm to compare the three models by testing their predictions about the effects of irrelevant visual input on different visuo-spatial working memory tasks. Two important main findings emerged from these experiments. (1) The maintenance of visual images in working memory was largely insensitive to the effects of concurrent visual perceptual input. This insensitivity was not due to the immunity of visual image maintenance to any kind of interference, as demonstrated by the contrasting disruptive effects of other secondary tasks that require visuo-spatial working memory. (2) Visual mental imagery was susceptible to interference from irrelevant visual input. This interference effect was selective, as demonstrated by a lack of disruption of visual imagery by other secondary tasks. The workspace model appears best able to account for these results. The first main result supports this model’s main assumption that image maintenance in working memory is independent of the perceptually driven activation of long-term memory. The second observation can be accounted for by interpreting the interference as a competition for the activation of long-term memory, as required for the generation of images in visual working memory for imagery, as well as for perceptual processes.

Hemispatial neglect syndrome

In a second set of experiments, we further tested the three models by investigating implicit processing of visual information by neglect patients. One patient demonstrated high-level implicit processing, and an alternative account of implicit processing - in terms of the automatic processing of emotional content by the amygdala - was ruled out. The demonstration of implicit semantic processing was replicated with a different paradigm and patient, and in this experiment a specific assumption of the workspace model was confirmed: implicit processing appears to depend on the activation of long-term memory representations, in the absence of a visual representation in working memory. Both the gateway model and unitary model have difficulty accounting for these results.

Conclusion

The results of both sets of experiments thus converge to support the workspace model, and suggest a view in which perceptual input activates the contents of long-term memory, prior to these activated representations being made available in a functionally separate working memory system for further processing. The gateway model and unitary model are unable to accommodate all findings.