Although quality of life assessments have been employed successfully in descriptive and evaluative studies in oncology, their use in cancer clinical trials has, to date, been limited. A range of issues have impeded the conduct of clinical trial-based quality of life investigations. These include: the absence of theoretical models to guide the development of quality of life measures; over-reliance on ad hoc approaches to quality of life assessment; and insufficient attention to the practical constraints operating in clinical research settings. Of primary importance is the need to develop multidimensional quality of life instruments that are brief and psychometrically robust. It is suggested that future work on instrument development focus on refining currently available generic or cancer-specific measures, and on developing new diagnostic-specific questionnaire modules. This psychometric work should be guided by appropriate theoretical models of the relationship among health-related quality of life domains. Although it is widely accepted that the patient represents the most appropriate source of quality of life data, it is suggested that efforts also be directed toward improving the validity and reliability of physician-generated assessments of patients' performance status and of treatment toxicities, and toward determining the feasibility of employing family members as proxy raters of the psychologic and social health status of patients who are unwilling or unable to provide such information. Additional attention should be paid to the many logistical problems that arise in clinical trial-based quality of life investigations. In particular, research designs and data collection procedures should be selected that minimize patient, medical staff, and institutional burden.