Associations compete directly in memory

Abstract

Associations are confusable when they share an item. For example, double-function pairs (with the form AB, BC) are harder to remember than control pairs. Although ambiguous pairs are more difficult on average, it is not clear whether memories for associations compete directly with one another (associative competition hypothesis), as assumed by models that incorporate associative symmetry (bidirectional associations). Alternatively, associative interference results might be explained away by: (a) item suppression hypothesis: competition only between memory for the two target items (A and C are both targets of B); (b) candidate competition hypothesis: The cue (B) retrieves two potential targets, A and C, which compete to be output. These alternative hypotheses could explain previous results in the related, AB/AC learning procedure. Our procedure included a large amount of interference that had to be resolved within a single study set. Participants studied sets of control (single-function) and double-function pairs and were asked to produce one or two associates, respectively, to cue items. Recall of AB and BC were negatively correlated and could not be explained away by item suppression or competition between simultaneously retrieved candidate items. Thus, competition can occur at the level of representation of associations, regardless of which item is the cue, consistent with associative symmetry.

Keywords:

• Paired-associate learning
• Double-function lists
• Interference
• Associative symmetry
• Associative interference

We thank Isabel Lek for help testing participants, and Michael Kahana for valuable feedback on the manuscript. Supported in part by the Natural Sciences and Engineering Research Council of Canada and the Alberta Ingenuity Fund.

Notes

¹ Note that one could equally conceive of a facilitatory relationship between memory for the two associates of a given item. For example, during study, if PARENT–MONKEY is encoded with high strength, and later, MONKEY–CARPET is presented to the participant, the participant's ease in retrieving the prior pair associate of MONKEY (PARENT) could have a positive consequence for encoding of the new associate item (CARPET). The critical point, though, is that this would predict a positive correlation between memory for PARENT and memory for CARPET whether tested with a common probe (MONKEY; “same-probe” condition in Figure 3) or tested with different probe items (“yoked” correlation in Figure 3); thus, this kind of “item facilitation” effect would only weaken any observation of a negative correlation due to associative competition and therefore could not explain away any empirical support for the associative competition hypothesis.