Previous research has demonstrated that adding a 3rd (decoy) alternative that is dominated by only 1 of the other 2 alternatives in a choice set increases the preference for the dominating alternative (J. Huber, Payne, & Puto, 1982;Tyszka, 1983). By manipulating decoys within subjects, significant preference reversals have been obtained. The processes of 3 classes of models can account for these reversals: (a) The decoy affects the weights assigned to different dimensions, (b) the decoy produces range-frequency effects on the dimensional values of the alternatives, and (c) the perception of dominance directly increases the attractiveness of the targeted alternative. The results of 3 experiments designed to test these models favored a direct effect of dominance. When a dominated decoy was added to a choice set, asymmetrical dominance was demonstrated to be a necessary and sufficient condition for producing the predicted preference reversals.Preference reversals have intrigued decision scientists in much the same way that visual illusions have captured the attention of researchers investigating visual perception. In each case, the respective cognitive system appears to operate reasonably well in the normal environment. Yet, when brought into a laboratory situation in which stimulus information can be selectively manipulated, the systems can be made to break down, resulting in behaviors (or perceptions) that sometimes defy common sense. If one focuses only on these results, it is tempting to conclude that the cognitive system is poorly designed. However, the functional adequacy of the system under normal conditions implies that this is not the case. Instead, these systemic breakdowns provide theoretical leverage for better understanding the processes that govern decision making under normal conditions. Three broad classes of preference reversals have been reported in the literature. By far, the most extensively researched of these is the task-induced preference reversal, in which preference ordering is demonstrated to differ depending on the mode of response required by the task. Much of this research has involved comparison of how gambles are evalu-

Squares receive higher category ratings when the smaller sizes are presented more frequently than the larger sizes. This shift in the rating scale is greater when there are either fewer categories (the Category Effect) or more stimuli. Similar shifts were obtained whether the stimuli were presented successively for judgment or simultaneously. The Category Effect also occurred when subjects were not told how many categories to use until after the contextual stimuli had been presented. A simple range-frequency model describes most of the shifts in scale by variations in a single weighting parameter. However, these shifts are predicted by an elaborated model in which the number of representations of any stimulus in working memory is limited by a principle of consistent assignment of each stimulus to a single category. This elaborated model correctly predicts the disapearance of the Category Effect when contexts are manipulated by varying the spacing of stimulus values rather than by varying their relative frequencies. Category ratings are probably the most frequently used of dependent variables in psychological research, perhaps because they reflect the way people ordinarily express value judgments.

Three experiments explored cognitive models of inferior, compromise, and phantom decoy effects in both judgment and choice. Participants made judgments of attractiveness, justifiability, and evaluation anxiety associated with each alternative in the set, along with judgments of the attractiveness of each alternative's dimensional values. In another session, they also chose the alternative they most preferred. Results were analyzed in terms of the degree to which decoy effects reflected shifts in dimensional values or reflected emergent values based on relationships with other alternatives in the set. Both emergent-value and value-shift models of inferior decoy effects were supported, but only the emergent-value model of compromise decoy effects was supported. Results for the phantom decoy indicated that this effect was choice-based and did not occur in judgment. Thus, although decoy effects were largely similar in choice and judgment, they also differed in important ways. ᭧

Reference price effects on consumer price perceptions are often explained by Helson's adaptation-level theory, in which the cognitive representation of reference price is the prototype of the relevant category. However, recent conceptualizations and empirical evidence suggest the possibility of an exemplar model, which may be specified using Volkmann's range theory or Parducci's range-frequency theory.In two experiments, these three contextual models of reference price effects are pitted against one another. Based on the MANOVA and model fitting, range-frequency theory accounted for reference price effects that the other theories could not, suggesting that consumers compare the target price against specific members of the category rather than the category prototype. A third experiment demonstrated that range and frequency effects are moderated by the stimulus presentation condition, suggesting that consumers place greater weight on extreme prices anchoring the range for internal reference prices than for external reference prices.

This article highlights the theoretical differences between the Likert and Thurstone approaches to attitude measurement and demonstrates how such differences can lead to discrepant attitude estimates for individuals with the most extreme opinions. Both simulated data and real data on attitude toward abortion are used to demonstrate this discrepancy. The results suggest that attitude researchers should, at the very least, devote more attention to the empirical response characteristics of items on a Likert attitude questionnaire. At most, these results suggest that other methods, such as the Thurstone technique or one of its recently developed item response theory counterparts, should be used to derive attitude estimates from disagree-agree responses.Introductory texts often portray the Thurstone (1928) and the Likert (1932) approaches to attitude measurement as though both methods provide equally valid scores measuring attitude when individuals respond to a set of questionnaire items using a (binary or graded) disagree-agree response scale (Mueller, 1986;Petty & Cacioppo, 1981). This overly simplistic portrayal is fostered by studies that indicate that Likert and Thurstone attitude scores typically are correlated to at least a moderate degree (.60 ≤ r ≤ .95), regardless of whether responses to the same set of items are scored with the two proce-