Abstract House mice (Mus domesticus L.) have been present on sub-Antarctic Marion Island since the early 1800s. Several authors have suggested that an increase in mice density as a result of a general warming trend in the sub-Antarctic climate from the 1960s has led to a decline in invertebrate biomass and abundance. These suggestions have been supported by the observation that the invertebrates of nearby mouse-free Prince Edward Island are apparently larger and more numerous than on Marion. Our experiment was designed to determine whether mice have a direct effect on invertebrate abundance, biomass and community structure, or an effect on the vegetation community and thus potentially an indirect effect on invertebrates. We constructed five wire-mesh mouse-free exclosures in one habitat type on Marion Island and recorded both the soil macro-invertebrate community and the vegetation inside and outside each of the exclosures before the start of the experiment in 1996 and twice thereafter (1998 and 2000). Mice had no significant effect on any of the eight prey groups' abundance or biomass, or on community structure (diversity and composition). Four of the prey groups changed significantly over time in either biomass or abundance, independent of the presence of mice. Our results, which may have been affected by generally low statistical power, suggest that factors other than mice had a larger impact on invertebrates than mice alone.