Assessment center dimensions have often been found to be low in convergent and discriminant validity (Hinrichs & Haanpera, 1976; Sackett & Dreher, 1982; Sackett & Hakel, 1979; Turnage & Muchinsky, 1982). Assessors' use of prototypes may interfere with assessment center ratings. Reliance on prototypes may be especially pronounced when dimensions are abstract. In this study, the influence of concrete dimensions on assessors observations, classifications, rating accuracy, and convergent and discriminant validity was investigated in an assessment center simulation. Sixty-six university students were trained as assessors. Using either concrete or abstract dimensions, they then evaluated the performance of confederates in three situational exercises. Subjects who rated concrete dimensions classified behaviors more accurately, rated dimensions more accurately according to two accuracy measures, and produced somewhat better convergent and discriminant validity than did subjects who rated abstract dimensions. Subjects who rated abstract dimensions had more accurate ratings according to one accuracy measure than did subjects who rated concrete dimensions.