Colony population size of Coptotermes acinaciformis (Froggatt), a species of mound-building, wood-eating, subterranean termite from tropical Australia, was estimated using a market-recapture protocol and by direct counts of individuals collected from mounds. The mark-recapture estimates varied widely within and between colonies (0.4-19.1 million),depending on definitions of recaptures (intensity of color), and could be 10 times larger than the direct counts (1.2-1.6 million). Assumptions of the mark-recapture protocol were shown to be violated in field and laboratory tests. Nile blue A, a fat-stain marker, did not persist and was transferred by cannibalism. Marked individuals did not mix uniformly with unmarked individuals; instead, foragers displayed feeding-site fidelity. The likelihood of recapture differed between castes and instars, there was a higher recapture rate of large workers and soldiers relative to small workers. These violations would increase population estimates from mark-recapture protocols. These results agree with those found for temperate species, suggesting that the difficulties with mark-recapture protocols are not limited to 1 habitat type or species, and that mark-recapture protocols do not estimate population size accurately.