The problem of severe chronic pain following spinal cord injury (SCI) has been well delineated for many years. However, progress has been slow in determining which treatment procedures work for which type of SCI pain in a particular individual at a particular time. Reasons for such slow progress include the low incidence and prevalence of SCI making large series of patients unavailable for trials in many centers. Lack of a widely accepted SCI pain classification scheme has also limited the comparability of similar interventions carried out by different investigators. Design limitations of many existing studies that can be remedied in future investigations are reviewed. Finally, discussion of two conceptual models of SCI pain is presented and an argument raised for the utility of both.