Inventories of land molluscs based on samples are liable to biased sampling error, because species that are present may be missed. Such errors may give a false impression of the degree of difference between faunas from similar sites, and they are confounded with genuine heterogeneity, which has considerable ecological and conservation significance. The statistical properties of such errors are examined briefly in the context of sample size and the frequency distribution of species. The effectiveness of different sampling strategies is assessed, taking account of the different life cycles and behaviour of various mollusc species. Taking random quadrats alone is not an efficient method for inventory; a combination of visual search and extraction from litter and sail is required. A set of practical guidelines is given, and the importance of good inventories, especially in oligotrophic habitats, is underlined.