Towards a reconciliation of a Can Do and Can't Do approach to language assessment

(eng) The Common European Framework for Languages (CEF) is largely based on what has been called a 'Can Do' approach. This approach can be explained as follows: "The aim of the 'Can Do' project is to develop and validate a set of performance-related scales, describing what learners can actually do in the foreign language" (Common European Framework of Reference for Languages, p. 244). In line with this approach, the descriptors for each level of proficiency found in the CEF appendix emphasise the skills that learners have already acquired, while only providing an incomplete description of those they have yet to develop, i.e. their Can't Dos. Concerning written production, the descriptors overemphasise inappropriate use of connectors, punctuation and spelling and refer only implicitly to lexical and grammatical difficulties. With a view to making the language skills representative of each individual level of proficiency more explicit, we error-tagged learner texts extracted from the International Corpus of Learner English (Granger 2003; Dagneaux et al 1998), each of which had already been assessed independently on the basis of the CEF guidelines. We then compared the error-tagged learner texts with the level of proficiency each had been assigned. Our presentation aims at showing how a combined 'Can Do' and 'Can't Do' approach to language assessment will help to flesh out the CEF proficiency levels by making the descriptors more explicit.