Confocal laser-scanning microscopy was used to visualize subsurface cerebral microvessels labeled with intravascular fluorescein in a closed cranial window model of the anesthetized rat. In noninvasive optical sections up to 250 mu m beneath the brain surface, plasma perfusion and blood cell perfusion of individual capillaries were studied. Under resting conditions, in all cerebral capillaries the presence of plasma flow was demonstrated by the appearance of an intravenously injected fluorescent tracer within 20 seconds after injection. Plasma flow was verified even in capillaries that contained stationary erythrocytes or leukocytes; 91.1% of the capillaries contained flowing blood cells, 5.2% contained stationary blood cells, and no blood cells were seen in 3.6%. Mean blood cell velocity was 498.3+/-443.9 mu m/s, and the mean blood cell supply rate was 35.75+/-28.01 cells per second. When capillaries were continuously observed for 1 minute, ''on'' and ''off' periods of blood cell flow were noted. During hypercapnia (increase of PCO2 from 33.25 to 50.26 mmHg), mean blood cell flux increased from 38.6+/-17.2 to 55.5+/-12.2 per second (P<.005, paired t test of mean values in six animals), and blood cell velocity increased from 519.5+/-254.8 to 828.5+/-460.8 mu m/s (P=.074, paired t test of mean values in six animals). Homogeneity of blood cell flux increased as indicated by the coefficient of variation decreasing from 44.6% to 22.0%, and the portion of poorly perfused capillaries (blood cell flux, <40 per second) decreased from 59.2% to 22.4%. Capillary diameter increased from 5.33+/-0.25 to 5.66+/-0.29 mu m (P<.05, paired t test of means in five animals). Our results suggest that opening and closing of capillaries (capillary recruitment) in the classic ''all or none'' fashion is not a means of regulating cerebral blood flow. Rather, we suggest that alterations in heterogeneity of capillary perfusion accompanied by variations of the fraction of low-flow capillaries and capillary diameter may be important adaptory mechanisms.