Predation by introduced house mice Mus musculus on islands is one cause of decline in native birds and has deleterious impacts on other ecological aspects. Eradication of rats (Rattus spp.) from islands of up to >10 000 ha has been achieved, but for mice scale is still an issue with the largest island cleared being only 710 ha. The feasibility of eradicating mice from larger islands is being considered, and to support these assessments, we undertook a field study on Gough Island (6400 ha) to determine whether all mice would be likely to accept toxic bait. We replicated a toxic bait operation as closely as possible, in timing, probable bait density and distribution, using a bait formulation used commonly in rodent eradication operations. Baits lacked toxin but were coated with the fluorescent dye rhodamine B. Mice trapped in and around the baited areas were inspected under ultraviolet light for fluorescent marking indicative of bait consumption. Of 434 mice, 97% tested positive, including mice trapped on assessment lines up to 90 m from the closest bait. There was no difference in the proportions of unstained mice from assessment lines outside baited sites compared with mice trapped in the core baited sites, suggesting large-scale foraging movements over relatively large distances into the baited sites from surrounding, non-baited habitat. Despite the high bait densities (15.7 kg ha^-1 at initial application and 7.9 kg ha^-1 at second application), bait consumption rates of ~4 kg ha^-1 day^-1 occurred after both applications. This was much higher than expected (probably the result of large-scale movements) and meant that all baits were consumed before trapping began. Thus the 13 unstained mice trapped in the core of the baited area may have moved there after bait was consumed. Further trials are required to assess whether all unmarked mice were false negatives (not exposed to bait) or if any were true negatives (rejected bait). A separate experiment found that all 11 mice trapped in a cave had eaten bait applied aboveground around the cave’s entrances, suggesting that caves do not serve as refugia for mice and are thus unlikely to compromise an eradication attempt.