Modeling and data collection efforts needed to develop a test of the effectiveness of adult literacy and language training programs in increasing the earnings of trainees are reviewed. After reviewing the framework developed for the Manpower Development and Training Act and the Comprehensive Employment and Training Act (CETA) evaluations and the data sets used to test the models, literacy and language programs are considered in an attempt to determine the information needed for evaluation. Problems addressed include: (1) the post hoc nature of evaluations, precluding the possibility of a true experimental design; (2) the need for large sample sizes to detect small anticipated program effects on earnings of program participants; (3) the necessity for longitudinal data on earnings; and (4) the need to contain the costs of an evaluation. An interesting recent approach is the framework using recursive earnings equations developed by O. Ashenfelter (1975, 1978, 1979). A model, developed by N. Kiefer (1978) to extend Ashenfelter's framework, allows for the possibility that language training might alter the functional form relating human capital and earnings. The Continuous Longitudinal Manpower Survey, a study of demographic characteristics of CETA participants and program impacts on employment and earnings, is reviewed. The use of Current Population Survey data in combination with Social Security Administration data is also covered. Recommendations are provided for the selection of control groups. (TJH)