In recent years the Graduate Record Examinations (GRE) files have become recognized as a valuable source for studies of talent flow. When test takers register to take the GRE, they complete a background questionnaire that asks, among other things, their undergraduate major field and their intended field of graduate study. With this information, along with subsequent test scores, grades, parents' education, age, gender, ethnic group, citizenship, and other variables, we are able to examine patterns that relate major field choices, and changes in those choices, to background data and academic ability. Studies using the GRE files had been costly because of the length and complexity of the files and because of the careful and tedious programming that has had to be done to accommodate the coding changes that took place as the questionnaire was revised over the years. If we were to conduct further talent flow studies using the GRE data, it became clear that we needed a data base designed especially for that purpose. In this project we designed a special GRE talent flow data base from 1978 to 1987 having three different structures. The first is an individual examinee file in which one record exists for each test taker. The record is short and identically formatted from year to year. The second structure is a matrix in which rows correspond to intended graduate fields of study and columns correspond to undergraduate majors. One hundred major fields are arranged in a rational order. Contained within the cells of the matrices are statistics such as numbers of test takers, mean GRE verbal score, percentage planning to earn a doctorate, numbers of minorities, and so forth. The third matrix structure collapses the 100 specific fields into 10 general fields. All three data bases are on computer files for public use; the matrices are available in hard copy as well. During the project, we demonstrated some uses of the data base and devoted considerable effort to the design of graphic representations that could clearly express various aspects of talent flow. Some of the findings from our examples (on U.S. citizens only) were as follows: • Engineering, physical sciences, and mathematics showed steady growth as graduate fields of study, whereas all other broad fields showed declines through about 1984 and then increases. • The common belief that as numbers entering a field decline, student quality rises, and as numbers rise, quality declines, was not supported. • The percentage of examinees over age 30 increased from 15% to 28%; the greatest percentage were in the service professions. • The percentage of female test takers grew only slightly; physical sciences and engineering showed very little increase in female representation. • Consistent with the decline in doctorates earned by Black students was the number of Black test takers, which also declined. • Among test takers with degrees in humanities, and physical, biological, and social sciences, a higher percentage of women than men changed fields for graduate work. Among those in health sciences, education, and business, men were more likely to change than were women. • Test takers appeared to choose graduate specialties consistent with their relative verbal and quantitative skills, as reflected in their GRE scores: low-scoring students in demanding undergraduate fields appeared to move to less demanding graduate fields, and high-scoring individuals in less demanding fields appeared to move to more demanding ones.