This paper describes an experiment which examined how levels of processing (LOP) affected word fragment completion in a group of Wernicke-Korsakoff patients, a group of patients with closed head injury, and matched controls. The data showed that both the memory-impaired groups and the controls showed a LOP effect but that the effect was larger in controls. Data from other studies are reviewed and, in conjunction with the present findings, it is concluded that LOP effects obtained when memory-impaired individuals are tested using implicit memory tasks arise mainly from the contribution of lexical processing of targets and from contamination by explicit recollection.