How much ‚general language proficiency’ is there in different language tests?

The current research addresses the seemingly contradiction between the assumption that language proficiency as considered by language testing researchers is multi-dimensional on the one hand, and the multiple findings of C‑tests loading on a single “general language proficiency” (GLP) factor on the other hand. Research on the structure of language proficiency and in psycholinguistics suggests that GLP may best be represented as the common core across diverse language measures. In the present research, it is hypothesized that C‑tests are excellent measures of this common core. In contrast, other language measures, beyond putting demands on GLP, are assumed to tap unique processes, explaining why multi-dimensionality often best reflects the structure of language measures. These hypotheses are addressed by examining structural equation models that evaluate alternative assumptions about the dimensionality of language proficiency in general, and the construct validity of C-tests in particular. 222 students from the highest school track in Luxembourg completed a French C‑test, as well as the Test de Connaissance du Français (TCF), encompassing measures of reading, listening, speaking and writing. The findings point out that the four TCF measures put extra demands on unique processes, whereas the C-test measured GLP only. We conclude that using the term “general language proficiency” to express what C-tests measure, risks insinuating that C-tests also assess other aspects of language proficiency. However, C-tests should not be expected to replace other language measures when a clear diagnostic of language proficiency in particular domains is needed.