Bedömning av korrekthet och komplexitet i ett muntligt språkfärdighetstest

Sammanfattning: In this thesis, I investigate the relation between assessed speaking proficiency and measured complexity and accuracy in twenty test-takers’ performance in the oral part of the Swedish certificate test Swedex B2. Accuracy and complexity have been measured in seven different ways; both global and local measures from the CAF (Complexity–Accuracy–Fluency) triad have been used. Also, each test-taker’s level of processability capacity, according to the Processability theory (PT), has been estimated. Hence, I have tested correlations between in total eight measurements using Pearson’s correlation coefficient (r). Most measures co-vary with each other with some degree of significance. Furthermore, correlations between measurements and grades have been tested. The effect of accuracy and complexity on both a specific grade for ‘Formal accuracy and complexity’ and two holistic grades – one for the oral part of the test and one for the whole test, consisting of four parts – have been investigated. Results indicate that most measurements co-vary with the holistic grade for the oral exam. However, the specific grade co-varies with foremost global accuracy measures, indicating a halo-effect. In a final qualitative investigation, four test-takers are selected and examined closer. From this study, all seven CAF measures can be criticised; only PT level seems to capture the level of proficiency. Quantitative as well as qualitative results are considered in a discussion of both language testing and language proficiency measures.