Abstract The true uncertainties in estimates of body organ absorbed dose and dose equivalent, from exposures of interplanetary astronauts to large solar particle events (SPEs), are essentially unknown. Variations in models used to parameterize SPE proton spectra for input into space radiation transport and shielding computer codes can result in uncertainty about the reliability of dose predictions for these events. Also, different radiation transport codes and their input databases can yield significant differences in dose predictions, even for the same input spectra. Different results may also be obtained for the same input spectra and transport codes if different spacecraft and body self-shielding distributions are assumed. Heretofore there have been no systematic investigations of the variations in dose and dose equivalent resulting from these assumptions and models. In this work we present a study of the variability in predictions of organ dose and dose equivalent arising from the use of different parameters to represent the same incident SPE proton data and from the use of equivalent sphere approximations to represent human body geometry. The study uses the BRYNTRN space radiation transport code to calculate dose and dose equivalent for the skin, ocular lens and bone marrow using the October 1989 SPE as a model event. Comparisons of organ dose and dose equivalent, obtained with a realistic human geometry model and with the oft-used equivalent sphere approximation, are also made. It is demonstrated that variations of 30–40% in organ dose and dose equivalent are obtained for slight variations in spectral fitting parameters obtained when various data points are included or excluded from the fitting procedure. It is further demonstrated that extrapolating spectra from low energy (⩽30 MeV) proton fluence measurements, rather than using fluence data extending out to 100 MeV results in dose and dose equivalent predictions that are underestimated by factors as large as 2–3. Finally, it is also demonstrated that the use of equivalent sphere approximations to represent body organ self-shielding distributions results in organ doses and dose equivalent predictions that are 2–3 times larger than values obtained with anthropomorphic shielding configurations.