The evaluation of rates of tumor blood flow with small, rapidly diffusing tracers requires an accurate model for mass transport within the tissue and tracer biodistribution.It is generally assumed that the whole tumor or several tumor regions act as well–mixed compartments, an assumption that has never been evaluated in tumors. The purpose of this study was to assess the accuracy of compartmental flow models in tissue * isolated tumors. We measured the residence time distributions of various tracers with the use of ex vivo perfusion of tissue–isolated rat R3230AC mammary tumors. This approach permits simultaneious, independent measurements of total blood flow and tracer concentrations in afferent and efferent vessels. The isolated tumors were perfused with Krebs-Henseleit solution, to which could beadded D 2 O saline and either 3% by volume F44-E (a perfluorocarbon emulsion) or 1% by weight fluorescein isothiocyanate (FITC)–albumin. A pulse of D 2 O and one of the other tracers was added to the perfusing liquid, and the relative concentrations of both D 2 O and perfluorocarbon or FITC–albumin were measured in the tumor effluent. D 2 O and the perfluorocarbon were measured with an imaging spectrometer tuned to either 2 H or 19 F. FITC–albumin concentrations were measured by luminescence spectrometry. The results were analyzed using various compartmental models. The tracer residence time distribution deviated from that expected for a single well–mixed compartment. Only half of the D 2 O left the tumor with a time constant consistent with the known perfusate flow. The remainder exited the tumor more rapidly than expected, and neither vascular shunting nor macroscopic flow heterogeneity accounts for this component of the D 2 O flow. However, two–compartment models provide an improved fit to the data. Conclusions: Our experiments demonstrate that the simple compartmental model used to estimate blood flow with diffusible tracers is not accurate. The nonideal blood flow found in our experiments reflects phenomena that may have important effects in the development of pharmacokinetic models of drug delivery to tumors. The accuracy of blood flow measurements, made with such techniques as nuclear magnetic resonance, positron–emission tomography, and computed tomography, may also be affected when they rely on the assumption that the tumor is a collection of well–mixed compartments. [J. Natl Cancer Inst 86:293–299, 1994]