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.
Participants examined sets of apartments described along four dimensions. Attribute values were manipulated to provide a way to infer strategy from response patterns. Experiment 1 established baseline behavior in unconstrained search, whereas Experiments 2-4 constrained participants to search either by alternative or by dimension. Dimensionwise presentation resulted in higher accuracy and reduced looking times. In three alternative choice, there was no evidence that strategy use depended on constraint condition. Evidence for possible strategy differences across constraint conditions was found when either multiple judgments rather than a single choice had to be made or the number of alternatives was increased to five. These results supported features of the adaptive decision maker hypothesis (Payne, Bettman & Johnson, 1988) but suggested that strategy use is not always strongly linked to acquisition pattern.
Four experiments assessed preferences for schematic faces. In Experiment 1, eye gap and nose width were manipulated separately, and effects of shifting the range of values were assessed. Descriptive ratings of width showed contrast effects in accordance with A. Parducci's (1995) range-frequency theory. Evaluative ratings of pleasantness showed reversals of preference ordering that were modeled as shifts in ideal points toward the means of the contextual distributions. In Experiments 2 and 3, similar effects of context on preference were demonstrated in a trinary-choice task in which faces varied only in eye gap. In Experiment 4, eye gap and nose width were manipulated together, resulting in systematic contextual shifts of the ideal face within the 2-dimensional attribute space. The results demonstrated the pervasive effects of context on the construction of ideals determining preference and underlying attitudes.
In three experiments, students completed sets of three pairwise choices in each of several domains. The first two pairs in each set were contextual pairs, and the third was the target pair. Context was manipulated by widening the range of values on one dimension and narrowing it on the second dimension. Consistent with previous research, participants on target trials more often chose the alternative whose poorer value was on the widened dimension. Four models explained this effect as (a) decreased weighting of the wide dimension, (b) contextual shift in values, (c) a tendency to equalize responses, or (d) a tendency to add value based on dominance relationships. Path models provided evidence against weight change and response equalization, and manipulation of dominance did not reduce the effect. Results supported the value-shift explanation of trade-off contrast.
Students indicated which of 2 options they would choose or reject in a between-subjects task-frame manipulation. Alternatives had either high or low variability of feature values, corresponding to enriched and impoverished alternatives, respectively. Previous research has yielded mixed results of task framing, with E. Shafir (1993) demonstrating greater preference for the enriched alternative in choice than rejection but Y. Ganzach (1995) demonstrating the opposite result. An accentuation model explained these differences by postulating that the greater demands for justification in the choice task lead to accentuation of differences between alternatives in choice. The accentuation model was tested against weight-change models using 250 Ss in 2 experiments, 1 using various decision scenarios and the other using 4-trait adjective descriptions of potential roommates. Results were consistent with accentuation theory and inconsistent with a systematic change in weighting of positive and negative attributes across choice and rejection tasks.
Clarified the relationship between behavioral looking measures with the theoretical construct of weight. A sampling model was proposed in which the weight given to a piece of information corresponds to the amount of sampling of that information in either a continuous, discrete or strategic manner, which were related to process tracing measures of initial and additional time per and frequency of acquisition. Ss were 215 college students who uncovered information corresponding to verbal and math aptitude scores of hypothetical applicants. Ss judged the likelihood of success in a designated major or chose 1 from pair of applicants who they believed to be more likely to succeed. Judgment data demonstrated large effects of task focus on the weighting of verbal and math scores and corresponding increases in number of acquisitions and time per acquisition on the information receiving more weight. Analyses of choice proportions revealed effects of task focus on weight and bias parameters. Initial looks reflected orientation and screening functions and additional looks reflected more evaluative processes. Similarities and differences among groups of Ss who were classified as following different identifiable choice strategies were also explored.
Conditions under which pairwise dissimilarity ratings should reflect manipulations of the stimulus distribution were outlined by a model that proposed these effects. These conditions arise from either a context dependent process for constructing implicit scale values or a process that uses previously established stimulus-response associates. Consistent with the model, results from 3 experiments using unidimensional psychophysical stimuli demonstrated disordinal context effects on pairwise dissimilarity ratings when (a) there was a 3-s delay between presentation of pair members or (b) a unidimensional rating task preceded the pairwise dissimilarity ratings. Global effects of density were fit well by a model that extended A. Parducci's (1983) range-frequency theory to dissimilarity ratings. Local density effects were generally consistent with predictions from C. L. Krumhansl's (1978) distance-density theory.
To examine 3 contrasting models of decoy effects on choice behavior Ss were presented choice triads from several different domains, with alternatives described along 2 dimensions. In Exp 1, with 158 college students, the decoy alternative in each set was dominated by only 1 of the 2 other alternatives in the set. In Exp 2, with 77 Ss, the decoy alternative was dominated by both of the other alternatives in the set. Within different blocks of trials Ss rated overall attractiveness of each alternative, importance of the different dimensions, attractiveness of each attribute value, and the justifiability of each alternative. Significant effects of manipulating the decoy were found for justifiability ratings and value ratings, with these combining to predict effects on attractiveness ratings. Results argue against a weight-change model of decoy effects and support value-shift and value-added models.
Participants made paired comparisons of square sizes, with blue squares presented on the left side of a computer screen and red squares on the right. Context was manipulated by varying the distributions of blue and red squares separately. In Exp 1, target squares from low-range distributions were judged larger than the same-size squares from high-range distributions. In Exp 2, ranges were equated but ranks manipulated between distributions, pairing bell with U-shaped and positively skewed with negatively skewed distributions. Results provided little support for a rank-dependent valuation model. Exp 3 used distributions that were designed to test between adaptation-level and response-equalization models. Both models received support, with response latencies constituting an important moderating variable. Response patterns for short latency participants were consistent with a stimulus-based adaptation-level process, and those for longer latency participants were consistent with a response-equalization process.
The use of qualitative evaluations and comparisons entailed by heuristic choice strategies may account for the occurrence of decision anomalies such as certainty effects, possibility effects, and preference reversals. The strong reduction of these effects when gambles are represented as being played many times suggests that the multiple-play representation leads to an increase in the use of integration strategies that are more quantitative. 161 undergraduates were presented with a hypothetical gambling situation and asked if they would choose or refuse to play a gamble 1 or 100 times and to justify their decisions in writing. Results supported the hypothesis. Justifications for single-play choices focused on single attributes of the gamble, and for multiple play choices reflected strategies that integrated probabilities and outcomes.
Two experiments examined at what stage contrast effects occur within the impression-formation process. University students rated the likability of persons described by 5 trait adjectives, with the distribution of trait adjectives manipulated between Ss. Theoretically, contextual stimuli may affect the valuation of component traits before integration, a preintegration model, or they may affect the valuation of the composite impression after the components have been integrated, a postintegration model. Context was manipulated so that the preintegration model predicted an interaction between context and target, but the postintegration model did not. The predicted interaction was replicated several times, supporting preintegration processing of contextual information. However, when subjects were divided into fast and slow judges (on the basis of judgment latencies), the pattern of results supported preintegration contrast for fast judges and postintegration contrast for slow judges. These results support the conclusion that contrast effects may operate at different levels in impression formation.
Assessed the effects of stimulus context on similarity judgments of emotion words in a series of 4 experiments involving 90, 74, 71, and 85 undergraduates, respectively. In each experiment, the context either uniformly represented locations on J. A. Russell's (see PA, Vol 66:5062) 2-dimensional circumplex emotion space or emphasized emotions lying within particular quadrants of that space. The manipulation of context led to systematic changes in the similarity relations among the emotion words. Multidimensional scaling of the similarity data modeled these changes in terms of an emerging 3rd dimension that teased apart distinctions among the emotion words added in each context. Contextual effects were strongest when similarities were assessed via a sorting task and were greatly attenuated when similarities were assessed via direct pairwise ratings of similarity.
Participants who differed in the extremity of self-definition ("own position") with respect to a given trait (sociability, independence, or patience) made trait and evaluative judgments of behavior stimuli that varied in their descriptive implications for that trait. Across four experiments, individual differences in trait ratings of unambiguous information were mediated largely by differences in Ss' affective reactions to these stimuli rather than by direct use of own position as a judgmental anchor. When the target information was ambiguous, however, own position influenced trait judgments independently of Ss' affective reactions to these stimuli. These latter effects were moderated by either encoding or informational mechanisms. A theoretical framework is presented that accounts for these results and predicts how effects of self-knowledge on judgments of others should vary across different trait dimensions.
Previous research by Huber, Payne, & Puto (1982) and by Tyszka (1983) 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. By manipulating decoys within-subjects, significant preference reversals have been obtained. The processes of 3 classes of models can account for these reversals: (1) The decoy affects the weights assigned to different dimensions, (2) the decoy produces range-frequency effects on the dimensional values of the alternatives, and (3) 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 choice set, asymmetrical dominance was demonstrated to be a necessary and sufficient condition for producing the predicted preference reversals.
This study explored how preference reversals, induced by changes in response mode (choice vs pricing), are moderated by how lotteries are represented, as being played 1, 10, or 100 times. Ss chose which of a pair of gambles they preferred to play and determined the minimum selling price of each gamble. Replicating previous research (e.g., A. Tversky and M. Bar-Hillel, 1983), the preference reversal was obtained for single-play representations: Participants tended to choose the gambles with the higher probabilities of winning, but priced them lower. However, for multiple-play representations, preference reversals were reduced, and consistency between pricing and choice behaviors was increased. Both response modes were sensitive to differences in the expected values of gambles, but sensitivity did not vary significantly with the number of plays. These results support the hypothesis that violations of expected utility theory are reduced for multiple-play gambles.
Two experiments explored methods for standardizing ratings of the psychopathology of clinical case histories. In both experiments, the same case histories were rated as more pathological when mostly mild rather than severe cases were presented as the immediate context. Psychometric analyses demonstrated that this type of contextual effect is a potentially important source of unreliability in clinical judgment. In Experiment 1, increasing the number of points in the rating scale from 3 to either 7 or 100 significantly reduced the effects of the immediate context. In Experiment 2, providing verbal anchors in the form of either detailed DSM descriptions for each rating category or sample case histories for the two end-categories increased the reliability of the ratings by reducing the effects of the immediate contexts; however, these reductions occurred only when the ranges of the immediate contexts had been severely restricted. According to the range-frequency analysis, verbal anchors served to equate the endpoints of the subjective range for the different contextual conditions.
Examined whether intrapersonal comparisons and social comparisons operate in similar ways to determine ratings of happiness. Events were varied to create positively and negatively skewed distributions. The events in each distribution were ascribed to either a single person or a group of people; Ss rated how happy they would feel if they experienced specific events within the distribution. Ratings for both intrapersonal and social comparisons were fit well by Parducci's (1984) range-frequency theory. Individual events received higher ratings when presented within the positively skewed context. Overall happiness, as measured by both the mean of the happiness ratings as well as direct ratings, was highest for the negatively skewed distributions. The effects of skewing were more pronounced for intrapersonal comparisons, but ratings were more closely defined by the range of experimental stimuli for social comparisons.
In 2 experiments, 120 university students were instructed to assign grades as fairly as possible to different hypothetical distributions of exam scores (bell, U, positively skewed, and negatively skewed). Experiment 1 demonstrated significant distribution effects that were quantitatively consistent with A. Parducci's (1965) range-frequency theory: Grading reflected a roughly equal compromise between a tendency to assign grades to equal subranges of exam scores (e.g., A's to the top fifth of the range) and a tendency to assign an equal number of scores to each grade (e.g., A's to the top 20% of scores). Results of Exp 2 supported a range-frequency model in which different grades tend to be used with equal frequency over a modified model in which different grades tend to be used with fixed but unequal frequencies.
Three experiments demonstrated the applicability of a range-frequency analysis to social judgments. Subjects rated the happiness of either (a) schematic drawings of faces or (b) life events as expressed in short verbal descriptions. The relative frequency of these stimuli was manipulated experimentally, as was the number of rating categories. Consistent with psychophysical research, ratings became less sensitive to differences in the frequencies of contextual stimuli as the number of categories increased (the category effect). With more categories, ratings also showed less adjustment to the range of stimuli actually presented. The reduction in adjustment was greater when stimuli were presented successively and when the experimental set covered a limited range. These effects of varying the number of categories were interpreted as reflecting changes in the effective context for judgment: With more categories, the differences between the effective frequencies of contextual stimuli are reduced and new, more extreme comparison values are evoked. The implications of using coarse versus fine scales of judgment are discussed in terms of the dynamics of social judgment as well as choice of the appropriate number of categories in social research. We argue that selection of the number of rating categories should be guided by research objectives rather than by a search for "true" judgments.
Presented photographs of faces in a series, either singly or in pairs, for ratings of physical attractiveness. In Experiment 1, faces were presented singly to 281 undergraduates; the same face elicited higher ratings when less attractive faces predominated in the experimental series (successive contrast). Increasing the number of available categories resulted in higher ratings but did not reduce the amount of successive contrast. In Experiment 2, faces were presented in pairs to 251 undergraduates; the same face elicited lower ratings when presented simultaneously with a less attractive face (simultaneous assimilation). A model that uses the judgments resulting from a range-frequency compromise as the stimulus values for integration within pairs provides the best account of how both contrast and assimilation occur within the same session. Alternative interpretations of the observed contrast and assimilation are discussed.
Four experiments were conducted, using approximately 500 undergraduates, to compare category ratings of sets of stimuli with different skewing (i.e., positive and negative) under different conditions. Results indicate that squares received higher category ratings when the smaller sizes were presented more frequently than the larger sizes. This shift in the rating scale was greater when there were 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 Ss were not told how many categories to use until after the contextual stimuli had been presented. It is suggested that 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. It is concluded that this elaborated model correctly predicts the disappearance of the category effect when contexts are manipulated by varying the spacing of stimulus values rather than by varying their relative frequencies.
Copyright 1995-1998 by the University of South Carolina
Department of Psychology: Experimental Psychology Program
Webmaster:
Jonathan C. Pettibone