A Short TI Inversion Recovery (STIR) sequence with spin-echo data collection was compared to a conventional cardiac gated long TR spin-echo (SE) sequence for detecting intramedullary signal abnormalities. The cervical (n = 48), cervico-thoracic (n = 1), and thoraco-lumbar (n = 18) spinal cord was imaged in a sagittal plane with a 0.5 T (n = 61) or 1.5 T (n = 6) MRI unit in 67 patients with clinical evidence of myelopathy of different etiologies (e.g., multiple sclerosis, trauma, herniated intervertebral disk, spondylosis, etc.). In all patients, ungated double or quadruple echo STIR images (TR 1000-1400 msec, TI 100 msec, TE 30-60 or 30-60-90-120 msec) were compared with cardiac gated long TR (1400-2100 msec), double echo (30-100, 50-100, or 50-150 msec) SE images with first order flow compensation for the second echo. Although STIR images appeared "noisier" than long TR SE images, they showed fewer ghost artifacts. In 55 patients, single or multiple, focal or diffuse, hyperintense areas within the spinal cord were observed on both long TR SE and (magnitude reconstructed) STIR images. Lesion conspicuity was better on the STIR images in 25 patients, better on the SE images in 14 patients, and equal in 16. STIR sequence provides a valuable alternative to gated long TR SE sequence for the MRI investigation of intramedullary spinal lesions.