Three experiments explored word/name homographs (GARLAND-musical star/ wreath of flowers) to determine the extent to which presenting them for study as surnames or as content words affected memory on immediate retest. Implicit and explicit measures were taken. Implicit memory occurred reliably on all variants of this task. Names were better remembered than words when the forename was presented as a cue and subjects were asked to generate a surname. This finding conforms with the suggestion (Bredart, Valentine, Calder, & Gassi, 1995) that there is a dedicated name-output stage in the memory system, and suggests that this can be activated under some implicit memory conditions. Further experiments showed that this name advantage was not upheld under other (word but not name) cued-recall conditions. Under these conditions, implicit memory is sensitive to the perceptual rather than the contextual conditions of presentation.