This prospective longitudinal study evaluates the validity and reliability of joint indices (JIs) used to measure disease activity in patients with RA. From seven traditional JIs (Ritchie Articular Index (RAI), Modified RAI, Thompson score, 28 JI, 36 JI, total tender and total swollen joints) 37 'new' JIs were computed by considering three different characteristics of joint inflammation, tenderness, swelling and the combination of tenderness and swelling, and by grading for tenderness and/or weighting for surface area of the joints. Several aspects of validity were investigated, the construct (correlation with radiographic damage), correlational (correlation with ESR, general health) and criterion validity (correlation with a Health Assessment Questionnaire, discrimination between high and low disease activity). It was found that the validity and reliability of traditional JIs do not differ substantially. Graded JIs are almost always more valid than ungraded JIs. Weighted JIs are almost always less valid and reliable than unweighted JIs. Therefore no JI proved to be superior for measuring the disease activity under consideration. Taking simplicity into account the 28 JI, not graded and not weighted, was preferable.