Capture-mark-recapture was used to study small mammal populations in tropical montane forests in southern India. Eleven plots in six montane forest patches were sampled from February-October, 1994. Six species were captured, including four rodents and two shrews. PROGRAM CAPTURE was used to derive estimates of density of the most abundant species in the study area, Rattus rattus Linnaeus. The coefficient of variation of the density estimate was used as an index of precision. The coefficient of variation decreased exponentially with increasing capture probability and with an increase in trapping duration. The coefficient of variation and the capture probability were not correlated with estimates of density. The density estimate increased with trapping duration, as did trap mortality. The latter may have been due to the trend of increased mortality with recaptures of the same individual, which in turn may have been due to weight loss over consecutive captures. Estimates of density derived using four estimators were different for 2, 3, 4 and 5 days of trapping. The coefficient of variation was highest for the generalized removal estimate and lowest for Darroch's estimate. The models and estimators could not be applied to more than one species, and for this species, only in select habitats in a few seasons. Therefore, models of density estimation developed for temperate areas may not be suitable for tropical habitats due to low densities of small mammals in these habitats.