A synovial cavity is separated from plasma by synovial intima in series with capillary endothelium. Because 20% of the intimal surface is bare interstitium, the system is a convenient model for the study of passive transport through serial endothelial and interstitial layers. Here hydraulic flow across the composite barrier was investigated in forty-seven knees of isolated, blood-perfused rabbit hindquarters, at intra-articular pressures between 4 and 30 cmH2O. In order to measure barrier conductance at constant intra-articular pressure, pressure on the opposite side of the barrier was varied, i.e. capillary blood pressure (PC). Capillary pressure was changed by alteration of vascular perfusion pressures, and the resulting changes in rate of absorption of Krebs solution from the synovial cavity (QS) were recorded. Trans-synovial absorption was a negative linear function of PC at each joint pressure, in verification of the applicability of Starling's hypothesis to this system. The hydraulic conductance of the blood-joint barrier was calculated as dQS/dPC. Conductance was independent of intra-articular pressure below 9 cmH2O and was 0.12 +/- 0.015 microliter min-1 mmHg-1 (mean +/- S.E. of mean). Barrier conductance increased as a curvilinear function of intra-articular pressure above 9.4 cmH2O (yield pressure). At 30 cmH2O conductance averaged 0.60 +/- 0.06 microliter min-1 mmHg-1, a 5-fold increase. A hyperbolic curve relating net barrier conductance to joint pressure was predicted from the hypothesis that interstitial conductance increases as a monotonic function of intra-articular pressure above yield pressure (Appendix). The data were in reasonable agreement with the theoretical hyperbola. Interstitial conductivity (3 X 10(-7)-7 X 10(-7) cm4 s-1 N-1 below yield pressure) and mean endothelial conductance (1.1 X 10(-4)-1.4 X 10(-4) cm3 s-1 N-1) were evaluated and compared with values in other tissues. Synovial endothelium contains on average 0.25 fenestrae micron-1 circumference. The conductance of a single fenestra was calculated to be 2.3 X 10(-13) cm5 s-1 N-1. Interstitial resistance accounted for roughly half the total resistance below yield point: therefore dQS/dPC should not be equated with 'capillary filtration capacity' in tissues with dense or fenestrated capillary beds. Large inconsistencies between interstitial conductivity and glycosaminoglycan concentration are noted, and mechanistic explanations of increases in conductivity with joint pressure are offered.