In this study we attempted to develop a profile that could be used prospectively to identify veterans over 64 years of age who might be at risk for medication noncompliance. Male veterans (N = 249) having from one to seven oral daily prescribed medications were studied. Instruments administered to determine their relationship to compliance were the Paired-Associate Test, the Mini Mental State, the Multidimensional Health Locus of Control, and the Standard Depression Scale. Compliance was determined by a pill count made at two home visits. Seventy-three percent of the subjects were noncompliant. Variables significantly related to noncompliance by bi-variate analysis were included in the stepwise logistic regression to develop a predictive model of noncompliance. Ethnicity and number of daily prescribed pills were the only significant variables in the final model. The model correctly classified the subjects as compliant/noncompliant in 77% of the cases. Discussions during home interviews revealed a number of common problems.