Previous literature on classroom testing of second language speech skills provides several models of both task types and rubrics for rating, and suggestions regarding procedures for testing speaking with large numbers of learners. However, there is no clear, widely disseminated consensus in the profession on the appropriate paradigm to guide the testing and rating of learner performance in a new language, either from second language acquisition research or from the best practices of successful teachers. While there is similarity of descriptors from one rubric to another in professional publications, these statements are at best subjective. Thus, the rating of learners' performance rests heavily on individual instructors' interpretations of those descriptors. An initial investigation of instructor assumptions was conducted regarding student performance on speaking tests in one program and identified several areas of discrepancy in instructor testing and rating practice. It is argued that faculty as a group must delineate more clearly their specific expectations by level for a number of rated features. The concerns identified in this study coincide with those discussed recently in the literature, suggesting that other programs might benefit from similar self-analysis. The instructor questionnaire is appended. Contains 17 references. (MSE)