Many indices of interrater agreement on binary tasks have been proposed to assess reliability, but none has escaped criticism. In a series of Monte Carlo simulations, five such indices were evaluated using , an unbiased indicator of raters' ability to distinguish between the true presence or absence of the characteristic being judged. and, to a lesser extent, coefficients performed best across variations in characteristic prevalence, and raters' expertise and bias. Correlations with for , Scott's, and Gwet's were markedly lower. In situations where two raters make a series of binary judgments, the findings suggest that researchers should choose or to assess interrater agreement as the superiority of these indices was least influenced by variations in the decision environment and characteristics of the decision makers.

In the first of 2 telephone-survey studies, factor analysis of the attitudes of 159 respondents revealed a general conservatism factor and two forms of liberalism, traditional and radical. Conservatism increased with age, traditional liberalism was strongest in women and middleaged persons, and radical liberalism was stronger in men and decreased with age. In the second study, 240 respondents estimated the attitudes of a young, middle-aged, or old male or female target. Evidence of an "old-is-conservative" stereotype was clearest among young participants. Among old participants, the stereotype was evident only when the target was male. People associated traditional liberalism more with women than with men and radical liberalism more with men than with women. Both kinds of liberalism were expected to decrease with age. The authors conclude that age plays as important a role as gender in the attitude impressions people form during initial encounters.

Studies were undertaken to assess the accuracy of people's estimates of the attitudes of men and women. In the first study, we assessed attitudes of male and female university students on a broad range of social and political issues. In the second, we asked new participants to estimate the attitudes of typical males and females on the same statements. These estimates were used to select a set of stereotypic male statements and a set of stereotypic female statements. In the third study, participants estimated the attitudes of male and female students on the two sets of statements. The results of the first two studies indicated that both men and women expected larger gender differences in attitudes than actually exist. In the third study, we confirmed this result and found that people were least accurate in their estimates of the attitudes of men. Partial support was obtained for the hypothesis that in-group stereotypes would be more accurate than out-group stereotypes. Men were more accurate than women in estimating the attitudes of men but men and women were equally accurate in estimating the attitudes of women. The picture of stereotype accuracy for gender-based judgments of attitudes that emerges from our research is more complicated than earlier research would suggest.

Although variables related to attitude change have been heavily studied, there has been surprisingly little emphasis on variables related to attitude stability.In the present study we examined whether or not measures of attitude extremity, latency of agreement responses, and issue importance could be used to predict attitude stability. In a computerized testing procedure that made it possible to record response latencies, undergraduates (N = 50) indicated the extent of their agreement with 40 controversial statements as well as the importance they attached to each issue. Retest data on the same variables were obtained two to three weeks later. The results indicated that all three predictor variables were inter-related but each contributed independent information in predicting attitude stability. We discuss the theoretical aspects of the results in terms of Fazio's (1986Fazio's ( , 1989Fazio's ( , 1990) notion of attitude accessibility. From a practical perspective, we stress the importance of gathering stability-related information whenever it is feasible to do so. ResumeOn a beaucoup etudte les variables liees au changement d'attirude, mais il est etonnant de constater le peu d'importance qu'on semble accorder a la stabilite de l'attitude. Dans la presente £tude, nous avons tent6 de determiner si les mesures de rextremisme dans l'attitude, de la latence des reponses positives et de l'importance de la question controversee permettent de predire la stabilite de l'attitude. Lors de tests sur ordinateur qui permettaient d'enregistrer les latences de la reponse, des etudiants de premier cycle (N = 50) ont indique dans quelle mesure ils elaient d'accord avec 40 enonces sur des questions controversies ainsi que l'importance qu'ils attachaient & chaque question. De deux a trois semaines plus tard, des retests ont 6t6 effectues au moyen des memes variables. Les r£sultats obtenus indiquent que les trois variables pr£dicteurs sont interrelies, mais que chacune apporte de 1'information distincte pour ce qui est de prevoir la stabilite de l'attitude. Nous examinons les aspects theoriques des resultats en fonction de la notion d'accroissement de l'attitude de Fazio (1986, 1989, 1990). D'un point de vue pratique, nous faisons ressortir l'importance de recueillir de l'information liee a la stabilite lorsqu'il est possible de le faire.

No abstract