Radionuclide identification from a measured gamma-spectrum is an iterative process, where the analyst aims to find correct nuclides by decreasing the amount of possibilities by trial and error. Although the process of identification is quite complex, it can be formulated using rules of thumb combined with exact mathematical analysis. Thus, an expert system can be built, where the knowledge of a human expert is converted to explicit rules. In this paper expert system SHAMAN is presented, which carriess out the qualitative nuclide identification and activity determination with minimum of user intervention. The reasoning process is performed by an inference engine written in C-language. The system uses a database containing over 2000 radionuclides with about 48 000 gamma-transitions. Spectra are provided in preprocessed format, where peak energies, intensities and backgoounds with respective error estimates have been calculated by a separate analysis program.