Creating stereotypes and capturing their content

In two studies, attributes of target groups were manipulated and then stereotypes of those groups were assessed using multiple measurement techniques. A variety of trait ascription (e.g. Likert-type scales) and perceived variability (probability of differentiation, standard deviation) indexes were used. The measures were compared with regard to their sensitivity to experimentally created differences between the social groups. The data suggested that all of the mean level assessment techniques were fairly sensitive to actual differences between groups, but Likert-style trait ratings were the best-performing measures, in that they most consistently produced the proper ordering of means, and were most sensitive to our manipulations. The variability measures were generally unresponsive to marked differences in group variability across experimental conditions. Suggestions for researchers concerning appropriate use of measures and conceptions of stereotyping are provided.