The Brain Bank of La Salpêtrière Hospital (Paris) is implanted in a neuropathology laboratory. It is multipurpose, prospective, and "free of charge" for the users. Protocols are prospectively established, in collaboration with the neuroscientists. One of our major difficulties in the collection of cases concerns presently the controls: the neurological status of patients coming from Neurology departments has usually been correctly assessed but those patients are bad controls. The normality of the neurological status of patients dying in other departments is difficult to assess retrospectively. A general autopsy is performed in each case. Several systematic sampling and fixation procedures are currently in use; their pros and cons are discussed. The main safety problem we are confronted with is the risk of HIV and Jakob-Creutzfeldt transmission. We try to standardize our diagnostic procedures; criteria used in Alzheimer's disease, Parkinson's disease, Huntington's chorea are briefly reviewed. We plan, in the future, to standardize our procedures for control cases. The Brain Bank has had a very positive impact on the way this neuropathology laboratory works: it introduced new techniques; on the other hand, the adequate processing and diagnosis of the samples was, in many aspects, simplified by the collaboration with the neuropathology department. The demand for human brain samples is steadily increasing in Neuroscience, for at least 2 reasons: 1. some diseases are specifically human and lack adequate animal models (Alzheimer's disease, multisystem atrophy), or animal models may appear irrelevant in some aspects (multiple sclerosis) or finally, results obtained in animal models may have to be confronted with human pathology (AIDS ...) 2. many aspects of human neuroanatomy can not be extrapolated from animal data There are many ways of organizing a brain bank and no golden standard (Swaab et al., 1989): the neuroscientist himself may collect the samples in a given pathology or the neuropathologists may modify their practice to provide adequate samples to the neuroscientists. When the neuroscientist himself collects his own samples, he obviously proceeds more rapidly. However, he is confronted with the difficult problem of the controls, which require both a clinical follow-up and a pathological check up of the tissues, both of which may be difficult to obtain in a research unit. In our opinion, the neuropathologists are the natural "brain bankers": they are indeed naturally "rich", their job being precisely to collect human samples, in connection with the clinicians.(ABSTRACT TRUNCATED AT 400 WORDS)