Our objectives were to further characterize an artifact related to the localized failure of the frequency-selective (FATSAT) fat suppression magnetic resonance (MR) imaging technique. We constructed two phantoms simulating human orbital anatomy and imaged them on a 1.5-T MR scanner using (FATSAT) and short T1 inversion recovery (STIR) techniques of fat suppression. The first phantom resembled orbit structural configurations; it was imaged in coronal and axial planes and in varying orientations with respect to the main magnetic field (Z axis) to study the features of the artifact and to reproduce the asymmetry seen in clinical cases. We designed the second phantom to enable quantification of the change in artifact size with change in orientation. We imaged the orbits of a normal human volunteer in similar planes and orientations, and compared the results to clinical cases demonstrating the artifact and true orbital disease. The artifact identified with localized failure of FATSAT fat suppression manifested as regions of hyperintensity maximal at fat-air interfaces, with gradual fading of the increased signal with distance from the interfaces. The artifact was most prominent when the interfaces were perpendicular to the axis of the main magnetic field (Z axis). The regions of increased brightness obscured normal orbital structures but were not associated with alterations in the geometry of these structures. Changes in orientation of the interfaces with respect to the Z axis, both in the phantoms and normal volunteer, reproduced the asymmetry of fat suppression failure seen in clinical cases. The relationship of size of the artifact to change in orientation was nonlinear. The artifact was not seen on STIR images. We concluded that failure of FATSAT fat suppression may mimic orbital disease, particularly if asymmetric. As predicted by the Maxwell electromagnetism equation, slight variations in orientation of the fat-air interface to the Z axis may produce large asymmetries in fat suppression failure in the orbit. Confirmation may require either comparison with additional pulse sequences [T1-weighted spin echo (T1W SE) or STIR] or repositioning the patient's head to check for persistence of the finding with varying orientations.