Anaesthetists' ability to identify correctly a marked lumbar interspace was assessed in 100 patients undergoing spinal magnetic resonance imaging scans. Using ink, one anaesthetist marked an interspace on the lower spine and attempted to identify its level with the patient in the sitting position. A second anaesthetist attempted to identify the level with the patient in the flexed lateral position. A marker capsule was taped over the ink mark and a routine scan performed. The actual level of markers ranged from one space below to four spaces above the level at which the anaesthetist believed it to be. The marker was one space higher than assumed in 51% of cases and was identified correctly in only 29%. Accuracy was unaffected by patient position (sitting or lateral), although it was impaired by obesity (p = 0.001) and positioning of the markers high on the lower back (p < 0.001). The spinal cord terminated below L(1) in 19% of patients. This, together with the risk of accidentally selecting a higher interspace than intended for intrathecal injection, implies that spinal cord trauma is more likely when higher interspaces are selected.