In this study the relationship between underachievement in anatomy and spatial ability (both geometrical and anatomical) is investigated. Subjects were second-year medical students at the University of Cape Town from 1980 to 1983. Geometrical spatial ability was measured using a battery of three-dimensional exercises involving the sectioning, joining, translation, rotation and visualization of simple solid objects. Anatomical spatial achievement was measured using university practical examination scores of April, June and October, as well as students' scores on those items in the April, June and November MCQ anatomy examinations which were classified as spatially three-dimensional by a panel of lecturers in anatomy. Non-spatial anatomical achievement was measured using the university essay examination scores of April, June and November, together with students' scores on the non-spatial items in the MCQ anatomy examinations. From 1980 to 1983 it was found that students who failed the battery of geometrical spatial exercises and/or recorded large, persistent deficits on spatial MCQs relative to non-spatial MCQs, scored significantly lower marks in practical anatomy examinations throughout the year than those students who scored well in the battery of geometrical spatial exercises. Spatially competent and spatially inept students performed equally well on the non-spatial MCQs and the non-spatial essay examinations. Borderline and failing students recorded the greatest deficits in anatomical spatial scores (for whom losses of approximately 18% occurred in practical examinations in anatomy). Since potential failures with persistent spatial handicaps can be readily identified by mid-year, a programme of differentiated teaching methods is recommended for these students.