When does relational information influence evaluative conditioning?

Building on the intriguing findings of Peters and Gawronski (2011, Experiment 3), we examined the conditions under which relational information (i.e., information about how two stimuli are related) moderates evaluative conditioning (EC; i.e., the effect of stimulus pairing on liking). In Experiment 1, participants saw stimulus pairs that consisted of a novel nonword (conditioned stimulus; CS) and a known positive or negative word (unconditioned stimulus; US). Before or after the pairings, participants were told that the nonword had the opposite meaning to the word with which it was paired. Subsequent measures of the implicit evaluation of the nonwords revealed that nonwords paired with positive words were liked less than nonwords paired with negative words, but only when the relational information was presented before the CS-US pairings. In a second experiment, participants were first informed that the CS and US of each pair were related in one way (e.g., that they have the same meaning). Afterwards, this information was either confirmed (e.g., that they indeed have the same meaning) or reversed (e.g., that they actually have an opposite meaning). Whereas the first relational information had more impact on implicit evaluations than on explicit evaluations, the reverse was true for the second relational information. Moreover, informing participants that CS and US were equivalent produced the same effects as pairing CS and US without providing explicit relational information, thus suggesting that the mere co-occurrence of CS and US is treated as a cue for equivalence of CS and US. Implications for mental process models of EC are discussed.