The effect of the chemical shift artifact, resulting from misregistration or phase cancellation at the interface between compact and trabecular bone, on apparent bone thickness was quantified in six isolated equine limbs. Sagittal T1-weighted spin echo (SE) and in-phase three-dimensional spoiled gradient echo (SPGR) images were acquired twice with a 1.5 T magnetic resonance (MR) unit, switching the frequency encoding direction between acquisitions. Out-of-phase SPGR images were also obtained. MR images with different frequency encoding directions were compared with each other and to radiographs made from corresponding 3-mm-bone sections. Compact bone thickness was significantly different when comparing images acquired with different frequency encoding directions for both SE and SPGR sequences. Significant differences were identified in the frequency but not the phase encoding direction when measurements of compact bone in MR images were compared with measurements obtained from thin section radiographs for the majority of surfaces studied (P<0.05). Correction of MR measurements with the calculated chemical shift abolished these differences (P>0.05). Measurements of compact bone from out-of-phase SPGR sequences were significantly different than from in-phase sequences (P<0.001) with out-of-phase measurements greater than in-phase measurements by an average of 0.38 mm. These results indicate that the chemical shift artifact results in errors in MR evaluation of compact bone thickness when measurements are performed in the frequency encoding direction or in out-of-phase images. For better accuracy, measurements should be performed parallel to the phase encoding direction and avoiding out-of-phase gradient echo sequences.