On the distinction between two tests as measures of the same construct and as interchangeable. By: Mark A. Fine Fine, M. A. (1992). On the distinction between two tests as measures of the same construct and as interchangeable. American Psychologist, 47(9), 1146. Made available courtesy of the American Psychological Association: http://psycnet.apa.org/journals/amp/47/9/1146a/ ***Reprinted with permission. No further reproduction is authorized without written permission from the American Psychological Association. This article may not exactly replicate the final version published in the APA journal. It is not the copy of record. *** Abstract: Comments on the debate between E. R. Harcum (see record 1989-31713-001) and W. Branch (see record 1990-13759-001) regarding the assumption that 2 tests measure the same construct because they correlate moderately highly with each other. Although 2 highly correlated measures do assess the same underlying construct, they may not be interchangeable due to their respective psychometric properties. psychology | academic research | tatistical correlations | test interpretation | statistical Keywords: significance Article: Walter Branch (February 1990) criticized as inappropriate the practice of assuming two tests measure the same construct simply because they correlate moderately highly with each other. A high correlation, he stated, does not ensure that the two tests are interchangeable, as the tests may have differences on other indices, such as obtained scores, means, and standard deviations. Branch raised an important point for users of educational and psychological tests, but he failed to consider the distinction between (a) two tests that measure the same or a similar construct, and (b) two tests that are parallel (alternate) forms of each other. Constructs (e.g., depression, self-esteem) are abstractions that are not amenable to direct observation; as a result, test (e.g., Beck Depression Inventory, Rosenberg Self-Esteem Inventory) are developed to operationally measure underlying constructs ( Murphy & Davidshofer, 1988). According to Cronbach and Meehl (1955), a test has construct validity when its “nomological network” supports the conclusion that it measures the construct it purports to assess. For a test to have construct validity, it should correlate highly with tests measuring theoretically similar constructs (convergent validity) and should not correlate with tests measuring theoretically unrelated constructs (divergent validity).