The applicability of a recognized construct validation procedure, the nomological network, to a sixth-grade criterion-referenced mathematics test developed in a metropolitan school system was investigated. Two samples were tested. The first sample was used to investigate two hypotheses: items within each objective will constitute a Guttman scale, and a principal-factor analysis will reduce the test items to groupings similar to the groupings on objectives set up by the test developers. Both hypotheses were supported. Results from the second sample indicated: (1) total test scores were higher on posttest than on pretest as a consequence of instruction on the objectives; (2) mean scores on objectives which received the most time-on-task showed the greatest gain when teachers knew the pretest objectives; (3) gain differences between males and females were not significant; (4) most criterion-referenced test scores were correlated with norm-referenced test scores on similar objectives; and (5) posttest variance was greater than pretest variance, contrary to the original hypothesis. Correlations of socioeconomic status and IQ with test scores were no greater for norm-referenced tests than for criterion-referenced tests. (BW)