The effects of training and participation on sources of variance in a set of ratings of college classroom teaching effectiveness were compared. College students (N=96) were randomly assigned to four cells of the experimental design. Subjects in cells (A) and (B) participated in the construction of a set of behaviorally-anchored rating scales (BARS) of five aspects of college classroom teaching performance, while subjects in cells (C) and (D) performed a control task. Later, subjects in cells (A) and (C) were exposed to a rater training program, while subjects in cells (B) and (D) performed a control task. All subjects then evaluated five standardized simulated professors using the BARS. Training significantly reduced the overall elevation of the ratings; participation did not. Neither participation nor training significantly reduced the variance attributable to the category of behavior being evaluated. Both participation and training significantly reduced variance attributable to the professor being rated. Participation significantly increased the Category x Professor effect while training did not. There were no significant interactions among the treatments with regard to effects on any of the above characteristics of ratings. Findings suggest that, for these four characteristics of ratings, participation and training operate independently of each other. (Author)