BACKGROUND. The two major uncertainties associated with absorbed dose calculations involve: (1) measurement errors from assessment of radioactivity in specific organs and tissues by direct counting; and (2) application of standard anthropomorphic and biokinetic models for dose assessment. Uncertainties in direct counting result from the inherent difficulty of measuring radioactivity inside the body. Although the system recommended by the Medical Internal Radiation Dose (MIRD) Committee of the Society of Nuclear Medicine provides a general framework and conceptual basis for the dosimetry of administered radiopharmaceuticals, it does not provide complete methods for assessing some of the more important quantities of interest in radioimmunotherapy, such as dose to tumors and descriptions of spatial dose distributions within tissues. Current MIRD anthropomorphic models are only crude representations of the human body. Generalized biokinetic models used in the MIRD system may vary considerably from the actual biokinetics of radiolabeled compounds in the body. This review describes limitations of the present MIRD system for radioimmunotherapy; they include assumptions used in treatment planning and the lack of specific methods for tumor dosimetry, multi-cellular dosimetry, microdosimetry, small animal dosimetry, and uncertainty analysis. CONCLUSIONS. Treatment planning for radioimmunotherapy requires patient-specific organ models and customized biokinetic parameters. Improvements are also needed in marrow dosimetry to account for the amount and distribution of red marrow relative to that found in adjacent source regions, skeletal structures, and circulating blood. Simplified assumptions with regard to the locally absorbed fraction of beta-particle energy in tissues adjacent to source regions should not be used when depth-dose profiles are needed; for example, radiation absorbed doses to intestinal walls should be calculated over the entire mass of tissue or described by absorbed-dose distributions. Additional research is needed to develop improved measurement techniques and computational methods to assess more accurately internal dose distributions within tumors and normal tissues.