An empirical investigation into the construct validity of tests of speaking and reading English as a second language (ESL) was performed using the multitrait-multimethod convergent-divergent design of Campbell and Fiske. Interview, translation, and self-rating tests of the two hypothesized traits, "speaking ability" and "reading ability," were administered to a population of 75 native speakers of Mandarin Chinese at the University of Illinois. The hypothesis of convergent validity was supported for all of the tests. The two hypotheses of discriminant validity were supported in enough instances to provide some evidence of this type of validity and, thus, evidence of the independence of the speaking and reading traits. An analysis of variance was also performed that supported the hypothesis that speaking and reading abilities are independently measurable. In addition, it provided evidence that the method of testing has a significant influence on the test scores. The Campbell-Fiske design for collecting data is endorsed, but newer ways of formulating and testing the hypotheses used in evaluating the data are advanced. (Author)