Research into brain imaging of pain is largely dominated by experimental acute-pain studies. Applied study paradigms have evolved a lot over past years and the ensuing results have furthered enormously our understanding of acute-pain processing. In sharp contrast, published work on brain-imaging in chronic pain remains scant. Furthermore, the results of these studies are highly incongruent, which could be explained by the fact that patient populations studied varied largely in terms of pain history, pain distribution, cause of pain, and psychological set-up. To circumvent these problems, several investigators have used surrogate models of neuropathic pain, but the validity of these models is highly questionable. In this Review we critically discuss the problems and shortcomings of most published reports on chronic pain and we propose some strategies for future studies. We argue that the post-operative pain model is highly appealing since it opens perspectives for prospective longitudinal studies with repeated assessments and it enables control for many confounding factors, which hamper the interpretation of most current studies. We also plead for a multimodal imaging approach in which classic brain-activation studies are supplemented with genetic, neurochemistry, brain morphometry, and transcranial magnetic stimulation studies.