This report analyzes departmental variations in time to degree and attrition in Ph.D. programs at Berkeley. An alternative hypothesis, the Ph.D. production function, is examined by cross-section econometric analysis of 28 departments. The inputs included in the production function were student variables--quality and percent males; faculty variables--quality and number; and stipend support variables--number of T.A.'s, R.A.'s and fellowships. Estimates are given for the relative importance of each variable in explaining length of time to degree and attrition. The author argues that fellowships and research assistantships reflect the external market demand for Ph.D.'s, while teaching assistantships reflect the university's internal demand for instructional support. The production function and behavioral hypotheses are integrated by relating departmental differences in resources to an index of excess demand for Ph.D.'s by field. The author suggests that the production function hypothesis may be misleading, for increased resources unaccompanied by increased market demand may not lead to increased Ph.D. production. The paper concludes with an examination of the national production of new Ph.D.'s during the period 1947-1948 to 1967-68, focusing on Berkeley's relation to total supply. (Author)