Background There is a need for an accurate non-invasive diagnostic test for variant Creutzfeldt-Jakob disease (vCJD), We investigated the sensitivity and specificity of bilateral pulvinar high signal on magnetic resonance imaging (MRI) for the diagnosis of vCJD. Methods MRI from patients with vCJD and controls (patients with suspected CJD) were analysed. Scans were reviewed on two separate occasions by two neuroradiologists and scored for the distribution of changes, and likely final diagnosis. Scans from vCJD cases were reassessed to reach a consensus on all abnormalities. Findings We analysed 36 patients and 57 controls. vCJD patients were correctly identified based on bilateral pulvinar high signal in 29 of 36 and 32 of 36 cases on the first assessment by the two radiologists, and 32 of 36 and 31 of 36 on their second assessment. Bilateral increased pulvinar signal was identified in one of 57 and one of 57 controls on the first assessment and two of 57 and three of 57 controls on the second assessment. These reported changes in controls were graded as minimal/equivocal in six of seven patients and moderate in one (<0.5% of all control assessments). 80% of the assessments in vCJD cases were graded as moderate or substantial. On consensus review, 28 of 36 cases and none of 57 controls had prominent bilateral pulvinar signal-sensitivity 78% (95% CI 60-90%) and specificity 100% (95% CI 94-100%), Other common MRI features of vCJD were medial thalamic and periaqueductal grey matter high signal, and the notable absence of cerebral atrophy. Pulvinar high signal correlated with histological gliosis. Interpretation In the appropriate clinical context the MRI identification of bilaterally increased pulvinar signal is a useful non-invasive test for the diagnosis of vCJD.