Quality of life measures take into account the patient's perception of health. Many generic or specific instruments are available. The psychometric properties of such measures should allow for adequately testing the hypothesis of an investigation. We studied the properties of three quality of life measures: the Health Assessment Questionnaire (HAQ) specific for rheumatic diseases, the Nottingham Health Profile (NHP)--a generic measure--and the General Health Questionnaire (GHQ) which measures psychological dimensions. They were applied in a one year cohort study of 111 French rheumatoid arthritis patients set up for determining prognosis factors of quality of life. Criterion validity was established on high correlation between Ritchie index and physical dimensions (r = 0.29 to 0.58, p < 0.01). Internal consistency was good with Cronbach's alpha coefficients over 0.8 for all dimensions. Reproducibility was studied for physical dimensions with Intra-class Correlation Coefficient (ICC) for patients clinically unchanged. It was excellent for the HAQ (ICC = 0.89), good for the NHP (ICC = 0.57 to 0.73) but weak for the GHQ (ICC = 0.13 for somatic dimension). After one year follow-up, a significant change in quality of life could only be evidenced by the HAQ (Standardized Response Mean = 0.4, p < 0.05) in patients with clinical significant change. So among the three instruments, the HAQ should be preferred for longitudinal studies, possibly supplemented with a generic instrument that investigates more dimensions of quality of life.