We now know that a single gene mutation may present with multiple phenotypes, and vice versa, that a range of genetic abnormalities may cause a single phenotype. As a result, our traditional approach to determining disease nosology, though it has served medicine and therapeutics well in the last century and a half, is now outdated. Under that traditional model, the collection and aggregation of data are subjective, depending as it does on patient-doctor interactions. The interpretation of this information is through experience in the absence of a theoretical appreciation of the structural and functional architecture of the human brain. That may also be why it has not generated fundamental breakthroughs in our understanding of the pathophysiology of many psychiatric and certain neurological diseases. The application of modern computerized analytic techniques is leading to real advances in the application of advanced human neuroimaging to clinical science. The variety of ways in which brain tissues can be characterised by modern MRI is having an impact on diagnosis and on understanding disease mechanisms. The development of sophisticated, rule-based, image classification and data-mining techniques is a potential further advance. So what lies ahead? Is it time to radically overhaul our epistemological approach to brain disease? We know a great deal about brain structure and function. From genes, through functional protein expression, to cerebral networks and functionally specialised areas we have accumulated a mass of knowledge about the brain that defies easy interpretation. Advances in information technologies, from supercomputers to distributed and interactive databases, allied to new mathematics, knowledge of the human genome, and the increasingly sophisticated brain imaging methods make it possible to federate and integrate existing and future clinical and neuroscientific data for a more biologically based, mechanistic approach to brain disorders. This is the ambition of the Human Brain Project.