Working memory and episodic long-term memory

When we test a person’s working memory (WM) by asking them to maintain some information for a few seconds, we must assume that episodic long-term memory (LTM) also lays down a record of the information. Therefore, many have argued that performance in tests of WM relies in part on episodic LTM. How can we distinguish the contribution of WM and of episodic LTM to performance in a memory test? I propose an approach derived from the function of WM as a representational medium for holding the currently most relevant information available. This requires that WM protects its contents against proactive interference from outdated information. Episodic LTM, by contrast, is vulnerable to proactive interference. I will present three series of experiments in which I used proactive interference as an indicator for a contribution of episodic LTM to performance in three kinds of WM tests. In a continuous-reproduction test of simple visual features, proactive interference affects performance for sets that exceed estimates of WM capacity, but only when the items are presented relatively slowly. In a test of pair-wise bindings of words and objects, proactive interference emerged already for set sizes of more than 2 pairs. By contrast, in forward serial recall of verbal lists, I found only sparse evidence of proactive interference even with list lengths of 7 to 9 items. The contribution of episodic LTM to performance in WM tests appears to depend strongly on the test.