The vasculature of the mammalian renal medulla is complex, having neither discrete input nor output. There is also efficient countercurrent exchange between ascending and descending vasa recta in the vascular bundles. These considerations have hampered measurement of medullary blood flow since they impose pronounced constraints on methods used to assess flow. Three main strategies have been used: (i) indicator extraction; (ii) erythrocyte velocity tracking; and (iii) indicator dilution. These are discussed with respect to their assumptions, requirements, and limitations. There is a consensus that medullary blood flow is autoregulated, albeit over a narrower pressure range than is total renal blood flow. When normalized to gram tissue weight, medullary blood flow in the dog is similar to that in the rat, on the order of 1 to 1.5 mL X min-1 X g-1. This is considerably greater than estimated by the radioiodinated albumin uptake method which has severe conceptual and practical problems. From both theoretical and experimental evidence it seems that urinary concentrating ability is considerably less sensitive to changes in medullary blood flow than is often assumed.