Oral proficiency tests were constructed for French, German, and English by the Dutch National Institute for Educational Measurement to determine how well students could cope linguistically with daily living situations, after 3 years of foreign language instruction. Originally the rating scale featured six categories. Because of difficulties in scoring, the rating categories were reduced to three: (1) an intelligent and appropriate response with minor mistakes; (2) an intelligible and appropriate response, but takes some effort to understand; and (3) response is not forthcoming/not intelligible/not appropriate in the context. Categories for fluency and pronunciation were good, adequate, inadequate, and weak. Consistency of raters and agreement among them increased with the second model, except with English test-raters. Analysis of the English test assignments, answers and ratings led to conclusions and suggestions with a view to the training of raters, the construction of future tests, and the definition of criteria for assessments. Clearer definition of categories, provision for positive rating of answers that are intelligible but display some lack of knowledge of vocabulary items, and considerations of social and contextual appropriateness are all taken into consideration in a new program of instructions for raters. (AMH)