The development of novel, systemically administered radionuclide therapies (such as radioimmunotherapy) relies on the ability to predict dose-limiting toxicity to normal tissue. Where the kidney is the principal route of excretion of the radionuclide preparation and/or breakdown of products, nephrotoxicity may be the dose-limiting factor. Until recently, conventional (MIRD) dosimetry assumed the distribution in the kidney to be uniform. A new MIRD phantom of the kidney models it as a set of uniform suborgans. In the work described here, the assumption of uniformity of distribution and of heterogeneity of dose rate (and, thus, absorbed dose) was tested in the mouse model. In this paper, we examine the nonuniformity of distribution and the subsequent dose rate for 4 antibody preparations (IgG (150 kD), F(ab)'(2) (100 kD), Fab (50 kD) and sFv (27 kD)) labeled with 4 radionuclides ((125)I, (131)I, (186)Re, and (90)Y) of interest in radioimmunotherapy (RIT). The kidney was considered as a whole and as two suborgans (cortex and medulla), and the nonuniformity of the dose-rate distribution was measured by a correlation of modeled dose-rate distribution with the dose-rate distribution obtained for an equivalent uniform radionuclide distribution. The heterogeneity of distribution, the inter- and intra-suborgan, was seen to increase as the molecular weight of the antibodies decreased. The assumption of uniform activity distribution for the whole kidney gives a poor estimation of the distribution of the dose rate. In the cortex, the longer-range emitters smooth out the effect of heterogeneous distribution and, in mice, an assumption of uniform cortex self-dose distribution may be sufficient for simple calculations. It is unclear how much this smoothing would be relevant in the human kidney.