Maintenance of relational bindings: working memory or long-term memory?

Abstract

While there has been a wealth of research examining the effects of feature binding in working memory (WM), it remains unclear how relational bindings (pairings of items, or of an item to its presented serial position) are stored in WM. We proposed a hypothesis in which relational bindings are not maintained in WM, but rather pass directly to long-term memory (LTM), even over short spans. In order to test this hypothesis, we performed a series of experiments examining the effects of short-term memory conditions, as well as both temporally distinct and non-distinct LTM conditions, on serial position curves for words and for relational bindings. We predicted that, unlike in item memory, for relational bindings there would be no effect of either LTM condition on recency relative to the STM condition. In Experiment 1, we presented participants with sets of unrelated word pairs and measured performance on an associative recognition test after a short-term interval, a long-term interval, and a long-term interval in which interstimulus interval was also increased to maintain temporal distinctiveness of the items. In Experiment 2 we used a similar procedure but attempted to increase serial position effects by testing binding between word and serial position, rather than between items. In Experiment 3, the study procedure was again similar, but participants were tested on item recognition and on item-position binding using an order reconstruction test. While the experiments demonstrated either no effect of condition on recency, or evidence against such an effect, post-hoc analysis was able to demonstrate that many participants responded in a unique way to the STM test, with the unique pattern predicting success for this test. As such, we conclude that relational bindings are likely stored in WM, rather than being passed directly to LTM.