This report describes efforts made in 1981-82 to develop an error-prone model (EPM) to help judge the extent of misreporting of income and family size on applications for government-sponsored school meal benefits. (EPM's are statistical formulas that produce scoring systems used to distinguish applications likely to result in excess benefits from applications likely to be accurate.) A sample of 2,480 cases from the National Evaluation of the School Nutrition Programs (NESNP) database was used in the analysis: 1,984 cases were used to develop the model, and 486 were set aside to test the model after its development. The statistical procedure used to generate the initial EPM, sequential search, made use of the Automatic Interaction Detector program. Although analysis of the NESNP data successfully produced an EPM, the model had two major deficiencies. Therefore, data from the 1981-82 Income Verification Pilot Project (IVPP) were used to continue exploring EPM development. The random sample consisted of 741 families from 54 schools participating in Phase 1 of the IVPP. Results showed that the model developed from the NESNP database was validated on in-home audit data. The goal of producing a simple but effective EPM also was achieved. The resulting scoring profile was found to be capable of selecting a subgroup of applicants who have a 70 percent probability of receiving excess benefits. (RH)