Associative independence revisited: competition between conflicting associations can be resolved or even reversed in one trial

Abstract

In one type of association-memory paradigm, after studying pairs of the form AB, AC, participants must recall both B and C in response to A. Counterintuitively, yet often replicated, recall probabilities of B and C are typically uncorrelated (“associative independence”). This face-value independence is now understood to reflect a negative correlation due to AB and AC competing, approximately offset by a positive correlation produced by subject- and item-variability. The outcome might vary with stimulus material; for noun-pairs, and with a single study trial per pair, AB and AC have been found to be positively correlated. We replicated the positive correlation between AB and AC for noun-pairs, but this did not differ from the correlation expected for independent memory tests, suggesting that for noun pairs, AB and AC are independent on average. In Experiment 2, participants instructed to form separate images for AB and AC again produced an independence pattern, but participants instructed to combine AB and AC into an integrative image produced a facilitation pattern. Thus, the relationship between AB and AC varies, and can be influenced by study strategy. Association-memory models may need to accommodate a diverse range of AB–AC relationships, and studies that build on AB/AC learning may need to consider whether AB/AC start out with a competitive, facilitatory or independent relationship.