Some limits to appropriate application and the characteristics of uncertainties in damage function analysis (DFA) for breeder reactor spectra were investigated by means of computer experiments. Simulated irradiations in available neutron spectra were evaluated in terms of simple damage models and were used to study (a) the existence of damage functions, (b) the uniqueness of damage function solutions, (c) data error propagation, and (d) procedures for combining various errors to provide a total lower-bound fluence limit for a specified property change. An important factor in achieving a successful DFA was found to be the similarity between the experimental spectra used to generate the damage function and the spectra in which it was applied.