1. Hyaluronan is not only a lubricant but also enhances the synovial lining's resistance to fluid outflow. This finding led to the proposal that hyaluronan (> 2 x 10(6) Da, approximately 210 nm radius) may escape across the synovial lining less freely than smaller solutes (e.g. albumin, 6.7 x 10(4) Da, 3.6 nm radius) or water. Here multiple washouts were used to measure intraarticular hyaluronan mass and secretion rate in rabbit knees, leading to an estimate of hyaluronan turnover time. Plasma albumin permeation into the joint cavity was also measured to enable comparison of turnover times between molecules of very disparate size. 2. Endogenous hyaluronan mass in the joint cavity, analysed by high performance liquid chromatography of joint washes, was 182 +/- 9.9 micrograms (mean +/- S.E.M; n = 21). Since hyaluronan concentration in synovial fluid averages 3.62 +/- 0.19 micrograms microliters-1, the endogenous synovial fluid volume was calculated to be 50 microliters (mass/concentration), about double the aspiratable volume. 3. The hyaluronan secretion rate over 4 h was 4.80 +/- 0.77 micrograms h-1 (n = 5). The rate was significantly higher in contralateral joints expanded by 2 ml Ringer solution (5.80 +/- 0.84 micrograms h-1, n = 5, P = 0.01, Student's paired t test), indicating a stretch/hydration sensitive secretory mechanism. The newly secreted chains ((2.05-2.48) x 10(6) Da) were not significantly different in length from the endogenous chains (2.95 x 10(6) Da). 4. Hyaluronan turnover time, calculated as mass/secretion rate, was 31.4-37.9 h. This is more than an order of magnitude longer than turnover time for intra-articular albumin. The latter, determined from the intra-articular albumin mass and plasma-to-cavity permeation rate was 1.8 h (95% confidence intervals 1.2-3.5 h, n = 9). The big difference in turnover times support the view that, relative to albumin and water, hyaluronan is partially sieved out and retained in the joint cavity by the synovial lining. The lining cell layer is discontinuous, so it appears that interstitial matrix itself acts as a leaky size-selective molecular filter.