Purpose: The hematopoietically active tissues of skeletal bone marrow are a prime target for computational dosimetry given potential risks of leukemia and, at higher dose levels, acute marrow toxicity. The complex three-dimensional geometry of trabecular spongiosa, however, complicates schema for dose assessment in such a way that only a few reference skeletal models have been developed to date, and which are based upon microimaging of a limited number of cadaveric bone spongiosa cores. The question then arises as to what degree of accuracy is achievable from reference skeletal dose models when applied to individual patients or specific exposed populations? Methods: Patient variability in marrow dosimetry were quantified for three skeletal sites -the ribs, lumbar vertebrae, and cranium -for the beta-emitters 45 Ca, 153 Sm, and 90 Y, and the alpha-particle emitters 223 Ra, 219 Rn, and 215 Po, the latter two being the immediate progeny of the former. For each radionuclide and bone site, three patient parameters were altered from their values in the reference model: (1) bone size as a surrogate for patient stature, (2) marrow cellularity as a surrogate for ageor disease-related changes in marrow adiposity, and (3) the trabecular bone volume fraction as a surrogate for bone mineral density. Marrow dose variability is expressed as percent differences in the radionuclide S value given by the reference model and the patient-parameterized model. The impact of radionuclide biokinetics on marrow dosimetry was not considered. Results: Variations in overall bone size play a very minor role in active marrow dose variability. Marrow cellularity is a significant factor in dose variability for active marrow self-irradiation, but it plays no role for radionuclides localized to the trabecular bone matrix. Variations in trabecular bone volume fractions impact the active marrow dose variability for short-range particle emitters 45 Ca, 223 Ra, 219 Rn, and 215 Po in the vertebrae and ribs, skeletal sites with small spongiosa proportions of trabecular bone. In the cranium, with its relative high proportion of trabecular bone, significant differences in marrow dosimetry from the reference model were noted for all radionuclides. Conclusions: Skeletal models of active marrow dosimetry should be more fully parameterized to permit closer matching to patient bone density and marrow cellularity, particularly when considering short-range particle emitters localized to either the bone trabeculae or active marrow, respectively.