Twenty-six monolingual and 46 bilingual college students were assigned to 2 groups on the basis of their performance on a complex relational task, an empirical model of instructional control (O'Hora, Barnes-Holmes, Roche, & Smeets, 2004). The subjects were then exposed to the vocabulary, arithmetic, and digit-symbol encoding subtests of the WAIS-III. Subjects (N = 31) who successfully completed the relational task performed significantly better on the vocabulary and arithmetic subtests than those subjects (N = 44) who failed to do so. No significant differences in relational task performances of these 2 groups were obtained on the digit-symbol encoding subtest. In post-hoc statistical analyses, a low but significant correlation was obtained between the vocabulary and arithmetic scores and the percentage of correct responses emitted in 1 particular training phase of the relational task. Monolingual and bilingual subjects' performances were not significantly different in either the relational task or the WAIS subtests. These findings support the position that derived relational performances may provide a behavioral approach to human language abilities.