A model was validated wherein local cerebral blood flow (LCBF) in humans was quantified by single-photon emission computed tomography (SPECT) with i.v. injected N-isopropyl-P-[123I]iodoamphetamine (IMP) combined with a modification of the classic method of arterial input sampling. After i.v. injection of IMP in rat, autoradiograms of the brain showed activity distributions in the pattern of LCBF. IMP was nearly completely removed on 1st pass through monkey brain after intracarotid injection (CBF = 33 ml/100 g per min) and washed out with a half-time of approximately 1 h. When the modified method of arterial input and tissue-sample counting applied to dog brain, there was good correspondence between LCBF based on IMP and on that by microsphere injection over a wide flow range. In applying the method to human subjects using SPECT, whole-brain CBF measured 47.2 .+-. 5.4 ml/100 g per min (mean .+-. SD, n = 5), stable gray-white distinction persisted for over 1 h, and the half-time for brain washout was approximately 1 h. Perfusion deficits in patients were clearly demonstrated and quantified, comparing well with results now available from positron ECT.