Writing fits well within the realm of outcomes suitable for observation by performance assessments. Studies of the reliability of performance assessments have suggested that interrater reliability can be consistently high. Scoring consistency, however, is only one aspect of quality in decisions based on assessment results. Another is generalizability. Research suggests that if the number of ratings per task could be increased, it may yield an increase in "task" generalizability without a dramatic increase in the actual number of tasks. Multitrait analytic scoring strategies for writing performance assessments may increase "task" generalizability over a single holistic score. Research undertaken by G. Roid (1994) supports the potential usefulness of analytic scores as effective sources for feedback to students and as bases for meaningful discussion on the writing process. Work at the Center for the Study of Evaluation at the University of California, Los Angeles, has expanded on the development of methodology and uses of analytic scoring. Work on narrative-writing-specific scoring rubrics has shown promising evidence of reliability and validity. Training in and use of these rubrics has also increased participating teachers' understanding of the quality components of writing. (Contains 3 figures and 15 references.) (SLD)