Severe mental illnesses such as schizophrenia and mood disorders have a major impact on public health. Disease prevalence and phenotypic expression are the products of environment and gene interactions. However, our incomplete understanding of their aetiology and pathophysiology thwarts primary prevention and early diagnosis and limits the effective application of currently available treatments as well as the development of novel therapeutic approaches. Neuroimaging can provide detailed in vivo information about the biological mechanisms underpinning the relationship between genetic variation and clinical phenotypes or response to treatment. However, the biological complexity of severe mental illness results from unknown or unpredictable interactions between multiple genetic and environmental factors, many of which have only been partially identified. We propose that the use of epidemiological principles to neuroimaging research is a necessary next step in psychiatric research. Because of the complexity of mental disorders and the multiple risk factors involved only the use of large epidemiologically defined samples will allow us to study the broader spectrum of psychopathology, including sub-threshold presentation and explore pathophysiological processes and the functional impact of genetic and non-genetic factors on the onset and persistence of psychopathology.