This study reports validity evidence for a large-scale and "low-stakes" per
formance assessment involving 905 Grade 8 students. A subsample of 198 stud
ents was used to study the relationship of performance measures with conven
tional achievement and affective measures. Confirmatory factor analysis ind
icated that the eight math performance tasks were unidimensional. Generaliz
ability and dependability coefficients were .72 and .68, respectively. Also
provided is other empirical validity evidence. Performance scores produced
large and significant correlations with the achievement variables. Gender
differences were significant for the total performance score as well as for
the two components: concepts, procedures, and relationships, and applicati
ons and problem solving. Results are related to theory and practice.