Brain imaging provides ever more sensitive measures of structure and function relevant to human psychology and has revealed correlates for virtually every psychiatric disorder. Yet it plays no accepted role in psychiatric diagnosis beyond ruling out medical factors such as tumors or traumatic brain injuries. Why is brain imaging not used in the diagnosis of primary psychiatric disorders, such as depression, bipolar disease, schizophrenia, and attention-deficit hyperactivity disorder (ADHD)? This article addresses this question. It reviews the state of the art in psychiatric imaging, including diagnostic and other applications, and explains the nonutility of diagnostic imaging in terms of aspects of both the current state of imaging and the current nature of psychiatric nosology. The likely future path by which imaging-based diagnoses will be incorporated into psychiatry is also discussed. By reviewing one well-known attempt to use SPECT scanning in psychiatric diagnosis, the article examines a real-world practice that illustrates several related points: the appeal of the idea of image-assisted diagnosis for physicians, patients and families, despite a lack of proven effectiveness, and the mismatch between the categories and dimensions of current nosology and those suggested by imaging.