13N-labeled ammonia was used to investigate 1) the cerebral extraction and clearance of ammonia, 2) the mechanism by which capillaries accommodate changes in cerebral blood flow (CBF) and 3) its use for the measurement of CBF. The unidirectional extraction of 13NH3 in rhesus monkeys was measured during PaCO2 induced changes in CBF and dog studies were performed using in vitro tissue counting techniques to examine 13NH3 extraction in gray and white matter, mixed tissue and cerebellum during variations in CBF produced by combinations of embolization, local brain compression, and changes in PaCO2. The single pass extraction fraction of 13NH3 varied from about 70 to 20% over a CBF range of 12 to 140 cc/min/100 g. Capillary permeability-surface area product (PS) estimates with a Renkin/Crone model show PS increasing with CBF. The magnitude and rate of increase in PS with CBF was highest in gray matter greater than mixed tissue greater than white matter. Tissue extraction of 13NH3 vs CBF relationship was best described by a unidirectional transport model in which CBF increases by both recruitment of capillaries and by increases of blood velocity in open capillaries. This saturable-recruitment model provides a possible explanation for the mechanism of flow changes at the capillary level. The net 13NH3 extraction subsequent to an i.v. injection increases non-linearly with CBF. Doubling or halving basal CBF produced from 35 to 50% changes in the 13N tissue concentrations with further increases in CBF associated with progressively smaller changes in 13N concentrations.