Concern over costs associated with mental disorders has led to an increase in the number of economic evaluations of treatment interventions; unfortunately, methods for measuring resource use have not kept pace with this concern. Although it is well-known that a significant proportion of the costs associated with mental illness are for resources other than treatment, program evaluators and researchers often count only treatment costs in cost-effectiveness comparisons. Further, existing methods for measuring resource use are plagued by faulty assumptions about resource use, poor validity and reliability, and difficulties quantifying resource use. The authors discuss these problems and suggest five ways of improving measurement of nontreatment resources: clarifying assumptions, using multiple data sources, flexible data collection strategies, methods for improving the accuracy of recall, and an episodic approach to measurement.