Non-lethal sampling of DNA from individuals in wild populations will often be required for studies of the conservation genetics of social insects, since it avoids destroying members of scarce or declining species. We investigated the effectiveness and consequences of methods of non-lethal sampling of DNA from bumble bee workers. In an experiment with two captive and confined Bombus terrestris colonies, we found that, unlike sampling haemolymph, sampling the terminal portion of the tarsus of a mid-leg of a worker reliably yielded amplifiable microsatellite DNA and did not significantly reduce worker survivorship. In a further experiment with four B. terrestris colonies whose workers were allowed to forage freely at flowers in the external environment, tarsal sampling of either a mid-leg or a hind-leg had no significant effects on worker survivorship, the mean body mass of foraging workers, the frequency or duration of foraging trips, mass of pollen loads or mass of nectar loads. We therefore suggest that tarsal sampling of either a mid-leg or a hind-leg is an effective and acceptable means of non-lethally sampling DNA from workers in wild populations of bumble bees, because effects on individual and colony performance are likely to be absent or minimal.