To systematically evaluate magnetic resonance (MR) imaging findings described as being indicative of spinal infection in patients with proven spinal infection. Contrast material-enhanced spinal MR images obtained in 46 consecutive patients (22 women, 24 men; mean age, 58.2 years) with culture or histologic examination results positive for spinal infection were systematically evaluated by two observers. Tuberculous and postoperative infections were excluded. Disk signal intensity and disk height, presence of the nuclear cleft, vertebral signal intensity alterations, endplate erosions on T1-weighted MR images, and presence of paraspinal or epidural inflammation were evaluated. Patient charts and surgical reports were reviewed. In the 44 patients with disk infection, MR imaging criteria with good to excellent sensitivity included presence of paraspinal or epidural inflammation (n = 43, 97.7% sensitivity), disk enhancement (n = 42, 95.4% sensitivity), hyperintensity or fluid-equivalent disk signal intensity on T2-weighted MR images (n = 41, 93.2% sensitivity), and erosion or destruction of at least one vertebral endplate (n = 37, 84.1% sensitivity). Effacement of the nuclear cleft was only applicable in 18 patients (n = 15, 83.3% sensitivity). Criteria with low sensitivity included decreased height of the intervertebral space (n = 23, 52.3% sensitivity) and disk hypointensity on T1-weighted MR images (n = 13, 29.5% sensitivity). Involvement of several spinal levels occurred in seven (16%) patients. Other spinal infections included isolated vertebral osteomyelitis (n = 1) and primary epidural abscess (n = 1). Most MR imaging criteria commonly used to diagnose disk infections offer good to excellent sensitivity. In atypical manifestations of proven spinal infections, however, some of the classically described MR imaging criteria may not be observed. Copyright RSNA, 2003.