A model of the latent structure of performance ratings based on Wherry's (19.52) theory of rating was developed which sought to explain relationships between multiple ratings, each of multiple performance dimensions, obtained from multiple rating sources. The model was tested using data obtained from first-and second-level supervisors who rated the performance of approximately 700 entry level clerical and professional/technical employees of 96 insurance, financial, and service organizations. Results only partially confirmed the a priori model, supporting specifications of dimensional true performance and overall rater bias effects, but failing to support specifications of areal (dimensional) bias effects on ratings.