THE CONSTRUCT VALIDATION OF THE FSI ORAL INTERVIEW1

Abstract

The validation of oral language proficiency tests has long been problematical. The most common approaches to this problem have involved concurrent validation procedures for relating “indirect” measures to so‐called “direct” measures. The most frequently used criterion in such studies has been the Foreign Service Institute (FSI) oral interview. In validating the FSI interview itself, however, criterion‐referenced procedures are not possible since an adequate operationally defined criterion for comparison is not likely t o be found. Therefore construct validation procedures are necessary. In this study a multitrait‐rnultimethod matrix comprising six measures representing combinations of two traits (speaking and reading) and three methods (interview, translation, self‐rating) was used t o examine the construct validity of the FS1 oral interview. The six measures were administered individually to 75 native Mandarin Chinese speakers of English. Data were analyzed using the Campbell‐Fiske criteria for convergent and discriminant validity and using confirmatory factor analysis. The results indicate both convergent and discriminant validity for the FSI interview. Of the numerous factor models tested against the data, the model which provides the best fit comprises a general factor, two trait factors, and three method factors. A more parsimonious model, with two correlated trait factors and three method factors, fits the data nearly equally well and provides a more interpretable factor structure. These results provide strong evidence against a model of unitary language competence and support a model of partially divisible language competence.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 123 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.