Rasmussen's encephalitis is a rare chronic neurological disorder, characterised by unilateral inflammation of the cerebral cortex, drug-resistant epilepsy, and progressive neurological and cognitive deterioration. Neuropathological and immunological studies support the notion that Rasmussen's encephalitis is probably driven by a T-cell response to one or more antigenic epitopes, with potential additional contribution by autoantibodies. Careful analysis of the association between histopathology and clinical presentation suggests that initial damage to the brain is mediated by T cells and microglia, suggesting a window for treatment if Rasmussen's encephalitis can be diagnosed early. Advances in neuroimaging suggest that progression of the inflammatory process seen with MRI might be a good biomarker in Rasmussen's encephalitis. For many patients, families, and doctors, choosing the right time to move from medical management to surgery is a real therapeutic dilemma. Cerebral hemispherectomy remains the only cure for seizures, but there are inevitable functional compromises. Decisions of whether or when surgery should be undertaken are challenging in the absence of a dense neurological deficit, and vary by institutional experience. Further, the optimum time for surgery, to give the best language and cognitive outcome, is not yet well understood. Immunomodulatory treatments seem to slow rather than halt disease progression in Rasmussen's encephalitis, without changing the eventual outcome.