ERIC Number: ED223107
Record Type: RIE
Publication Date: 1981
Reference Count: N/A
Convergent and Discriminant Validation of Integrated and Unitary Language Skills: The Need for a Research Model.
Clifford, Ray T.
Correlational studies establishing the validity of language skill tests have traditionally described how well the tests have converged, that is, yielded equivalent results. Convergent and discriminant validation is a logical extension of these traditional procedures. However, since it requires evidence of discriminant as well as convergent validation it is ideal for the more rigorous, and functionally more important, problem of establishing the construct validity of language skill tests. A re-examination of examples drawn from the literature shows that studies claiming convergent test validity consistently fail to demonstrate evidence of discriminant or construct validity for the traits they purport to measure. These failures may be the result of error variance introduced by testing and rating methods and/or by attempts to measure skills which are based on shared rather than unique contributing elements. Suggestions are given for minimizing both method and specification error variance in convergent and discriminant language validation studies. (Author)
Descriptors: Language Research, Language Skills, Language Tests, Literature Reviews, Second Language Learning, Test Validity, Testing
Not available separately; see FL 013 318.
Publication Type: Reports - Research
Education Level: N/A
Authoring Institution: N/A
Note: In its: The Construct Validation of Tests of Communicative Competence.