The volume of distribution in tissue (Vt) that can be achieved by direct interstitial infusion of therapeutic agents into brain is limited. The maintenance of a pressure gradient during interstitial infusion to establish fluid convection has been shown to increase the Vt of small, medium, and large molecules. We have used monocrystalline iron oxide nanocompounds, superparamagnetic particles of sizes the same order of magnitude as virions, to investigate the effect of dose, the volume of infusate, and the time of infusion on the distribution of large molecules in rodent brain. Our initial study in rats (n = 6) replicated the results of a previously described report of convection-enhanced delivery in cats. At a constant rate and concentration, the Vt increased in a linear fashion, proportional to the increases in time, volume, and dose. When using a constant rate and a constant concentration, however, it is unclear which variable or variables (dose, volume, infusion time) have the greatest influence on this effect. Therefore, we assessed each variable independently (n = 12). When the iron dose was increased from 5.3 to 26.5 micrograms, there was a three- to fivefold increase in the Vt, depending on the volume and time of infusion (2 Microliters/20 min, 24 microliters/20 min, or 24 microliters/120 min) (P < 0.001). When the volume of infusate was increased from 2 to 24 microliters, at an infusion time of 20 minutes and a dose of either 5.3 or 26.5 micrograms, there was a 43 or 52% decline in the Vt, respectively (P = 0.018). When the time for the infusion of 24 microliters was increased from 20 to 120 minutes, there was a 79% increase in the Vt at a dose of 26.5 micrograms but no change in the Vt at a dose of 5.3 micrograms. The effect associated with infusion time was not significant (P = 0.113). Magnetic resonance imaging was performed to document the distribution of monocrystalline iron oxide nanocompounds in vivo, and histochemical staining for iron was used to document the distribution of monocrystalline iron oxide nanocompounds in tissue sections. The Vt for both methods was calculated by computer image analysis, and the correlation between magnetic resonance and histological volumes was determined (r2 = 0.93). On the basis of this model, we suggest that dose, rather than convection, might be the most important variable in maximizing the Vt and improved distribution might be achieved by administering an increased concentration of agent.