Forager population sizes of colonies of Coptotermes lacteus (Froggatt) in New South Wales were estimated using two methods: mark-recapture and constant removal, in two disturbed habitats: a pine plantation and cleared farmland. Mark-recapture population estimates were unrealistic and unreliable: they could be improbably large, over 200 million foragers, and they varied enormously between samples for each colony without any pattern. The constant removal population estimates could also be unrealistic: they could be negative or quite different when calculated using regression and maximum likelihood methods. However, the unrealistic results could be predicted reliably, and explained by the lack of re-contact with the sampling devices (bait stations) - a violation of an assumption of the method. This happened more frequently in the plantation than in the farmland, probably because of the greater abundance of alternative food sources in the plantation. Of the two methods, constant removal provided reasonable forager population estimates, relative to direct counts, at least some of the time, plus a mechanism by which reliability could be tested, whereas mark-recapture provided neither. Further refinement and testing of constant removal methods are urged to provide a more reliable population estimation technique for termites.