Microvascular hematocrit and its possible relation to oxygen supply were systematically examined. We studied the red cell volume fraction (hematocrit) in arterial blood and in capillaries under a variety of circumstances. Control capillary hematocrit averaged 10.4 +/- 2.0% (SE) and arteriolar (14.2 micrometer ID) hematocrit averaged 13.9 +/- 1.2% in cremaster muscles of pentobarbital-anesthetized hamsters. Carotid artery hematocrit was 53.2 +/- 0.6%. The low microvessel hematocrit could not be entirely explained by a high red cell flux through arteriovenous channels other than capillaries (shunting). Hematocrit was not only low at rest, but varied with physiological stimuli. A 1-Hz muscle contraction increased capillary hematocrit to 18.5 +/- 2.4%, and maximal vasodilation induced a rise to 39.3 +/- 9.5%. The quantitative relations between capillary red cell flux, arterial hematocrit, and total blood flow could be explained by a two-element model of microvascular blood flow that incorporated a relatively slow-moving plasma layer (1.2 micrometer). Such a model would generate a low microvessel hematocrit and might reduce the diffusion capacity of individual capillaries, but would not reduce time-averaged red cell flux or alter steady-state vascular oxygen supply.