Items located within an array were presented to alcoholic Korsakoff and nonalcoholic mixed-etiology amnesics and to alcoholic and normal controls. Recognition memory for the locations of items was tested after incidental and intentional encoding. When equated on item recognition, neither Korsakoff amnesics nor alcoholic controls benefited from intentional, relative to incidental, encoding instructions. Furthermore, Korsakoff amnesics showed neither disproportionately impaired incidental nor intentional location recognition memory relative to alcoholic controls. In contrast, mixed-etiology amnesics profited significantly from intentional location acquisition relative to incidental instructions, and were impaired somewhat in incidental, but not intentional, location memory relative to normal controls. We discuss these data in relation to Mayes' (1992) contextual memory deficit hypothesis and Hirst's (1982) automatic encoding deficit account, and propose an alternative framework in which the location memory deficit observed in mixed-etiology amnesics is interpreted as a disruption to the ability to bind item and location information.