The kinetics of transvascular fluid transport across fish capillaries and redistribution of fluids between intravascular compartments in intact fish are unknown. Cannulae were placed in the dorsal aorta (DA) and caudal vein (CV) of rainbow trout Oncorhynchus mykiss (mass 0.45-0.85 kg) and the fish spleenectomized. The following day a peristaltic pump was fitted to complete the extracorporeal arterio-venous circulation. Hematocrit (Hct) was monitored in unanesthetized fish either manually, by collecting blood from the extracorporeal loop at 5 min intervals for a period of 1 h (groups 1 and 2), or continuously (instantaneously) with an impedance flow-cell inserted in the aortic cannula (group 3). Fish in group 1 were volume expanded by injecting a volume of saline (0.9 g% NaCl; SI) or trout plasma (PI) equivalent to 40% of the plasma volume. In group 2, 20% or 35% of the blood volume was removed, and in group 3, 35% of the blood volume was removed. Plasma volume (Vp) was calculated from an assumed blood volume of 35 ml kg(-1) and the Hct. Vp declined mono-exponentially after SI with a half-time of 6.8 min and Vp reached a new steady state at 28.1 ml kg(-1); 30% of the injected volume remained in the vasculature. Volume recovery after PI was also mono-exponential, but slower (half-time=15.4 min) than SI, whereas the steady-state Vp (27.3 ml kg(-1)) was similar and 30% of the injected volume remained in the vasculature. Thus the presence of plasma proteins delayed fluid efflux from the vasculature, but did not affect the volume lost. Transvascular fluid filtration coefficients calculated from this data were 5.5 (SI) and 4.5 ml mmHg(-1) kg(-1) min(-1) (PI), and interstitial compliance was 11.8 (SI) and 9.7 ml mmHg(-1) kg(-1) (PI). The rate of volume recovery after 20% or 35% hemorrhage was independent of the hemorrhage volume (half-time=13.3 and 15.1 min, respectively) and similar to the half-time of PI, indicating that protein-rich interstitial fluid is returned to the vasculature. There is a nearly instantaneous change in Hct that occurs during the hemorrhage period; it is dependent on hemorrhage duration and volume and not associated with the subsequent mono-exponential recovery. This initial response is best explained by a rapid fluid shift from a large-volume (approximately 40% of total blood volume), low-hematocrit (less than half of systemic Hct) microcirculation into the higher-hematocrit macrocirculation. These studies are consistent with transcapillary fluid flux across a barrier that is highly permeable to protein, and cannot be explained by fluid shift between primary and secondary circulations, or by transcapilllary flux across a capillary bed that is impermeable to plasma proteins. The results support the hypothesis that whole-body reflection coefficients in trout are very low and that plasma oncotic pressure is not a determinant of transcapillary fluid balance. They also show that both transvascular and intravascular fluid movements are important effectors of central volume homeostasis.