In this non-experimental study, a model was developed for portfolio assessment based on definitions and applications in the assessment literature. This model describes portfolio components, scores to be computed, and uses to be made of the scores. The literature was then reviewed to find examples of actual applications that would provide realistic estimates of statistical characteristics of assessment results. Estimates of statistical characteristics were also obtained from performance assessments similar to portfolios. The model and estimates were then used to estimate characteristics of an operational large-scale portfolio assessment program. Estimates were obtained of reliability of results, score distribution characteristics, and the validity of the procedure for writing evaluations for 12th graders. The analysis suggests that a well-structured and carefully scored portfolio assessment has the potential to provide scores that meet standards of reliability required for use with individual students. By developing a score that is the sum of the scores on the various portfolio entries, useful discriminations can be made among students. A factor that has not been considered is the cost of obtaining the expected results, which would undoubtedly be high. One table presents analysis results, and one graph shows composite score reliabilities. (SLD)