Efficient solution of the Tower of London (T of L) test (Shallice, 1982) has been hypothesized to depend on frontal lobe mediation. Performance on the test by 20 patients with severe, diffuse, traumatic head injury was compared with that of control subjects, matched on age and years of education, and, within the patient group, according to broad location of damage (frontal or nonfrontal). Scores were also compared with those obtained on two commonly used tests of frontal lobe function, Verbal Fluency, and the Wisconsin Card Sorting test (WCST). Results indicated that whereas the Verbal Fluency test and WCST discriminated well between patients and control subjects, the T of L test did not, nor did it discriminate well between patients with and without documented frontal lobe damage. However, there were large individual differences in planning and solution times across all subjects. Interpretation of results was complicated by interaction of other variables such as premorbid IQ and duration of posttraumatic amnesia. It is suggested that closer attention to the relation between planning and total times separately for correct and incorrect solutions might be informative. A need for systematic study of the relationship between test performance and everyday behavior is also identified. (JINS, 1995, 7, 537–544.)