The longitudinal method aims to point out phenomena that are bound with time. It is thus particularly fit for the study of ageing as it is proved by many large American research programs. It is also useful to follow during a limited period (short term) the natural history of pathologic entities as degenerative illnesses. After mentioning the fundamental conditions of any longitudinal approach, the program of this study on senile dementia is described. Each case is followed for 2 years, with an examination every 6 months. This later is based on anamnestic, psychiatric, neurologic and functional items as well as on intellectual functions and EEG. A sample of 50 cases should be collected. After a period of 3 years prospection 91 cases were found of which on 31 fitted. The difficulty of recruitement is bound to several factors: the little knowledge of senile dementia frequency in the general population, the poor criteria used in their choices by non-specialists as the directors of homes, the refusals and the withdrawals. Difficulties in establishing contact with patients showing behavioral troubles are underlined as well as cooperative problems with the surroundings. The repetition, at regular periods, of psychological tests sets the problem of learning. In the way of senile dementia it should decrease with the illness evolution thus bringing the overestimation of initial capacities, the gap becomming artifically greater between performances at the beginning and at the end of the observation.