Output register parallelism in an identical direct and semi-direct speaking test: A case study

The present trend in developing and using semi-direct speaking tests has been supported by test developers and researchers' claim of their increased practicality, higher reliability and concurrent validity with test scores in direct oral proficiency interviews. However, it is universally agreed within the language testing and assessment community that interchangeability must be investigated from multiple perspectives. This study compared test taker output from a computer-based Aptis General speaking test and a purposively developed identical face-to-face direct oral proficiency interview using a counterbalanced research design. Within subject analyses of salient output features identified in prior related research were completed. Results showed that test taker output in the computer-based test was less contextualised, with minimally higher lexical density and syntactic complexity. Given these findings, the indicated slight register shift in output may be viewed as non-consequential, or even as advantageous, for semi-direct speaking tests.