The learning curve. Can the results of the grammar exam be predicted?

Abstract

This paper exploits the potentials of inferential statistics in its quest to answer two related questions, whether the exercises performed by students during their course in theoretical grammar really prepare them for the exam, and whether the students’ exam results can be predicted from their achievements in said exercises. This study is in the context of English Business Communication at Aalborg University, Denmark. Several statistical methods, various forms of regression analysis, are pursued in order to discover which one – if any – of them is best suited to make predictions as to future exam results. It is found that the exercises investigated do indeed contribute significantly to the students' learning process, and that the exam results are predictable within a reasonable margin of error from the results of the exercises. Somewhat surprisingly, the simplest forms of linear regression and 1st degree polynomial regression are found to be the best predictors of the exam results, not any of the elaborate methods also tested. As a side effect, the study also reveals that the students' level of knowledge of theoretical grammar prior to their entering the university has no appreciable influence on their exam results.