In a quest for a objective assessment of synovitis of the knee, the value of circumferential measurements of the knee at two levels has been tested. Measurement 1 cm above the patella was shown to be more precise and to correlate better with the quantity of synovial fluid aspirated than measurement at mid-patellar level. The size of change necessary for statistical significance frequently exceeded the change likely to be found in clinical practice. The precision of measurement is greater in normal than in rheumatoid knees, and can be increased by using the mean of duplicate measurements rather than a single measurement. In clinical trials, the use of one observer throughout will increase the reliance that can be placed on the changes in measurements.