| A procedure preference reversal occurs when different methods for measuring a preference yield different results. A robust finding is that these reversals occur with regularity across a number of different measurement methods, yet no satisfactory explanation of this phenomenon exists. This dissertation provides evidence that one type of procedure preference reversal, between individual ratings and pairwise choices, is due to the interaction of choice alternatives during choice but not during ratings. These results suggest that the current lack of a satisfactory explanation is due to reliance on the common assumption that alternatives are evaluated independently of each other during choice. A contrary assumption, that alternatives are interdependent, was made and implemented in a parallel constraint satisfaction model. Experiments 1--3 tested the model, resulting in the identification of three cognitive effects underlying multiattribute choice: (1) a discrimination effect: high variances between choice attributes result in faster response times (RTs) than low variances, (2) a tradeoff effect: negatively covarying valences of choice attributes (i.e., compensatory choices) result in longer RTs than positively covarying valences (i.e., dominated choices), with maximum RTs occurring at the choice indifference point, and (3) an opportunity cost effect: for compensatory choices, but not for dominated choices, RTs are greater and confidence is lower for larger variances of choice attributes.; The model was applied to the ratings versus choice preference problem by activating alternatives individually for ratings and in pairs for choices. Three different cognitive factors that could lead to preference reversals during choice were identified from the model. These factors manifest themselves psychologically as a contrast effect, an attribute discrimination effect, and an attentional focus effect. Experiment 4 supported all three effects, but did not discriminate effectively between the latter two effects. Experiment 5, designed to discriminate between the two, suggests that the dominant effect, and the cause of preference reversals identified in previous research, is due to differences in discrimination time between different attributes. Results are discussed, focusing on the assumption of interdependent choice alternatives, use of a parallel constraint satisfaction model to make predictions, and the preference reversal phenomenon. |
