It is only quite recently, among other things as a result of the increasing demands for learner-centered education, that, particularly in the field of foreign language teaching to adults, an interest in the use of self-assessment techniques has begun to emerge.
However, for an appropriate use of self-assessment as a means of evaluation of foreign language proficiency, an insight into the validity and reliability characteristics of the learners' judgements is needed.
This paper discusses some findings of an experiment in which Dutch adult learners were asked to assess their proficiency in French by filling in two evaluation forms. In one form, the questions included aim at a global assess-ment of the learners' command of the French language both with regard to each of the four skills and to pronunciation, grammar, vocabulary and orthography.
The other form used covers only the four skills. The questions included in this form present, for each skill, a number of specific situations of language use. The learner is asked to indicate how well he thinks he can handle the French language in each situation described.
In order to determine the reliability of the self-ratings the learners were asked to fill in, after an interval of about two weeks, the latter forra a second time. Criteria for the validity of the learners' judgements were based on their scores on language-proficiency tests. Besides this, data on a number of personal characteristics were obtained.
The relationships between all these variables were analyzed. The research findings indicate among other things
1. that self-ratings which are based on a specification of language proficien-cy in terms of language activities are reliable to a fairly high degree;
2. that these self-ratings are better predictors of the test-scores than the ratings obtained by means of a global evaluation form;
3. that self-ratings of listening comprehension are less valid than self-ratings of speaking or reading comprehension.
Diverging task orientations in L2 oral proficiency tests – a conversation analytic approach to participant understandings of pre-set discussion tasks
