In vivo measurements are commonly used to determine the incorporated activity of radionuclides. However, for complex or old contaminations, the assessment of the activity distribution among organs is difficult to obtain from measurements only. A method is suggested to assess the activity of organs; it consists in separating the contribution of each contaminated organ in the experimental counts using organ-specific counting efficiencies. The feasibility of the method was first experimentally investigated using two torso phantoms containing radioactive organs. The first phantom contained two radioactive lungs and the second one two lungs and a liver. In the two cases the activity distribution was obtained using experimental specific efficiencies and the bias when compared with true activity was below 5%. The experiments were reproduced by Monte Carlo calculations in order to test the efficiency of the separation method with calculated organ-specific counting efficiencies instead of experimental ones. For the first experiment, the most contributing organ-specific counting efficiencies were correctly simulated but biases as large as 70% were found for the less contributing ones. Consequently, for one organ the activity was correctly assessed but for the other a 20% underestimate was found. For the second experiment, the calculated efficiencies were in better agreement with experimental values and the maximal difference between the calculated and the reference activity was 11%. The method's limits were investigated using perturbation point sources and considering wrong activity distribution hypothesis. The method gives correct results when the number of contaminated organs is overestimated by the startup hypothesis. The opposite situation can be identified only if more measurements than required by the startup hypothesis are carried out. © 2013 Elsevier Ltd. All rights reserved.