Problems associated with criterion-referenced language testing are discussed in the context of both standardized proficiency testing and classroom assessment. First, different interpretations of criterion-referencing are examined. A range of approaches for defining criteria and performance levels in second language assessment are outlined, and some issues that have arisen in defining and applying these criteria are discussed, including the difficulties of defining the nature of proficiency and the failure of expert judges to agree on criteria. Finally, a discussion is given of research directions that might lead to language assessment criteria that incorporate multiple perspectives on learners' communicative needs and derive from empirical data on second language acquisition, variability in language use, and communicative competence. A 73-item bibliography is included. (MSE)