AN ANALYSIS OF COURSE GRADES

College grades suffer as a criterion for test development because of their unreliability and because of the difficulty of comparing the grades at one college with those at another. However, since they often constitute the best criterion we have, an optimum understanding of them is desirable. Toward this end, course grade criteria at two colleges were broken down into relatively homogeneous part‐grades. These were intercorrelated with the section scores of a special six‐hour aptitude test. In a number of cases final course grades are shown to be more heterogeneous than they should be to constitute a good criterion for test development, expecially where differential prediction is intended. The specific conclusions of this study are suggestive of the results that may be obtained from college course breakdowns of this type. They are of interest where specific course criteria, rather than scholastic averages are required. Some of the conclusions are: a) Unconventional parts of courses should be deleted from the course grade criterion. b) In English it is advisable to use a separate criterion for English, objective tests and English essays. c) Validation against part‐grades can reveal types of course material escaping present prediction.