The advantages of using magnetic resonance imaging (MRI) as opposed to computed tomographic (CT) scans or ventriculography in stereotactic surgery include the increased tissue contrast of the lesion or target, direct non-reformatted multiplanar imaging and target coordinate determination as well as reduced imaging artefacts produced by the stereotactic frame. One disadvantage of MR stereotaxis, however, is the potential for anatomic inaccuracy due to equipment-induced inhomogeneities of the magnetic field. The authors present an experimental study on an in vitro model to examine the accuracy of target localization using the Leksell stereotactic frame and MR imaging. Ten formalin-fixed brains taken from patients who had died of non-neurological diseases were sealed in a properly modelled plaster-cast shell simulating the skull bone. These models were fixed in the Leksell stereotactic frame and high-field MR images were performed (Siemens Magnetom SP 1.5 Tesla, T1-weighted spin echo sequences, TR/TE 600/15 ms, slice thickness 2 mm, FOV 300 mm). Following electrocoagulation of different targets on both lentiforme nuclei, the localization and extension of the lesions were controlled by MRI. A gross-/histopathological verification was performed. This model allows a good representation of the anatomic structures without any artefacts. The postoperative MRI control and the pathological examination of the lesions matched well with the preoperatively defined targets. The correlation of coordinates and measurements obtained with the pathological studies were within a +/- 2 mm range in all cases.