Abstract
The effect of the number of behavior categories systematically observed on the reliability and accuracy of judges' ratings was determined. Ss representing two gymnastic judging systems rated 24 filmed routines under standardized conditions. The results showed that the judges rating only one category had significantly less variance from absolute ratings, less intravariance about their own mean, and higher reliability than those rating three categories. When the individual judge's ratings were combined into group ratings, no significant differences were found between observation systems on variance from absolute ratings; however, the groups rating only one category had significantly less intravariance than groups rating more than one category.