Routine evaluation of axial MR images of the cervical spine with high-intensity CSF (long TR/TE spin-echo or gradient-echo images) revealed apparent narrowing of the cord's anteroposterior diameter when these images were compared with corresponding postmyelography CT scans. This discrepancy was believed to be due to the truncation artifact at the CSF-cord boundary. To examine the truncation effect, we compared cord diameters in 12 patients on postmyelography CT scans and MR images and then compared these with MR scans of normal volunteers and of an agar-saline spine phantom. There was an artifactual diminution of the cord diameter in the 128-step phase-encoding axis of the 128 x 256-matrix MR scan as compared with the diameter of the cord in the patients' postiohexol CT scans and in the 256 phase-encoded axis MR scan in the volunteer study. A similar discrepancy was noted in the spine phantom study, in which the cord diameter in the 256-step phase-encoded MR scan, the CT scan, and direct measurement exceeded that in the 128-step phase-encoded axis MR scan. The range of differences between the measurements was as large as 2.3 mm (patients), 1.7 mm (volunteers), and 1.8 mm (phantom) for the three studies. In all three studies, varying the photographic window width and level produced variation in the apparent cord diameter of up to 1.5 mm. To eliminate this effect, the cord diameters in the phantom and the normal control subjects were measured at identical window levels. The truncation artifact, coupled with standard window settings used in photography, may lead to inaccurate display of the diameter of the cervical spinal cord.