Co-occurrence and relational information in evaluative learning: A multinomial modeling approach.

Dual-process theories of evaluative learning suggest that evaluative representations can be formed via two functionally distinct mechanisms: automatic formation of associative links between co-occurring events (associative learning) and nonautomatic generation and truth assessment of mental propositions about the relation between stimuli (propositional learning). Single-process propositional theories reject the idea of automatic association formation, attributing all instances of evaluative learning to propositional processes. A central question in the debate between the two theories concerns the mechanisms underlying unqualified effects of stimulus co-occurrence when the relation between the co-occurring stimuli suggests an evaluation that is opposite to the one implied by the observed co-occurrence (e.g., sunscreen prevents skin cancer). Addressing interpretational ambiguities in previous research on the differential impact of co-occurrence and relational information on implicit and explicit measures, the current research used a multinomial modeling approach to investigate the functional properties of the effects of co-occurrence and relational information on a single measure of evaluative responses. Although the moderating effects obtained for relational information are consistent with the predictions of the two theories, the obtained properties of co-occurrence effects pose an explanatory challenge to both dual-process and single-process propositional theories. The findings demonstrate the value of multinomial modeling in providing deeper insights into the functional properties of the effects of co-occurrence and relational information, which impose stronger empirical constraints on extant theories of evaluative learning.