Validity and generalizability of social dance performance ratings.

The purpose of three related studies was to develop a social dance instrument that had logical validity and generalizable performance ratings. Three social dance experts critiqued the instrument and concluded the items described the characteristics of good social dance performance. Twelve couples from a social dance class were videotaped for 5 min while performing the fox trot. Their performances were evaluated by a different set of coders in each of the three generalizability (G-) studies conducted (Subjects x Coders x Days ANOVA design). Dance performances were held constant across coding occasions so the day facet represents degree of consistency in coding, not stability of subjects' performances. Decision (D-) studies were also conducted for various measurement conditions. Results of the G-studies indicated that coders with several years of experience teaching and evaluating dance had less systematic coder bias and less systematic coder bias for a subset of dancers than coders who had little or no experience evaluating movement of any form. The performance ratings given by novice coders could be generalized for any randomly selected coder evaluating on any given day (G = .79). It was concluded the social dance test was valid and coders could be trained to obtain generalizable performance ratings.