Deconstructing the construct, one gap at a time: knowledge and automaticity in canonical versus speeded C-Tests

Despite extensive use of the C-Test as a measure of global language proficiency, its underlying performance components and sensitivity to timing constraints remain under-specified. This study addresses both issues by integrating two methodological strands. First, it relates language knowledge (KNOW) and efficient speeded deployment of knowledge (AUTO) to performance on canonical (unspeeded) and speeded English C-Tests of 237 adult L2 learners. Second, it analyzes screen-recorded test-taking processes of 25 English learners working on C-Tests under different timing conditions. Results converged across analyses. Within each format, KNOW was a strong, format-general predictor, whereas AUTO made a small but significant unique contribution only in the speeded C-Test. Across formats, the knowledge effect was essentially unchanged, whereas automaticity played a larger role in the speeded format; the two predictors also overlapped substantially. Process evidence aligned with this pattern: under time pressure, participants produced fewer fast, correct first-pass responses and fewer total correct responses, while backtracking and revision rates remained similar. We conclude that timing does not appear to alter knowledge demands substantially, but is associated with a greater predictive role for the speed and efficiency with which knowledge is accessed and executed. The speeded C-Test therefore appears more sensitive to efficiency-related aspects of performance, whereas the canonical format remains more strongly driven by knowledge accessible under negligible time pressure. Practically, using both versions yields complementary diagnostics, reflecting both “how much is known” and “how efficiently it can be deployed”, to inform better assessment designs and applications.