Despite the current enthusiasm for neuroimaging as a key method in translational neuroscience, there is a lack of debate about the nosological framework within which neuroimaging measures should be related to diagnostic categories. Here, the aim was to stimulate a debate about the role of cognitive neuroscience and neuroimaging in mediating between molecular/genetic, clinical diagnostic, and symptom-based descriptions of neuropsychiatric disorders. The diagnostic role of neuroimaging in translational neuroscience is stressed, namely, to be combined with cognitive measures to define cognitive-anatomical syndromes as an intermediate diagnostic category that mediates between clinical diagnoses and psychoreactive as well as neurobiological etiologic factors. This multilevel approach will be illustrated by reviewing recent insights into the cognitive-anatomical basis of inappropriate social behavior and social knowledge in frontotemporal dementia and by discussing its implications for the study of neuropsychiatric disorders such as major depressive disorder in which neuroanatomical abnormalities are more subtle.