I examined patterns of species distributions in ground ants on Barro Colorado Island (BCI), Republic of Panama, using baited transect samples and Berlese extraction of litter arthropods. I sampled during wet and dry seasons in 1976 and 1977. All sites sampled were under closed—canopy rain forest at least 75 yr old. More species and more individuals were collected in wet than dry season using either method. Evenness of species abundance at baits also increased during the wet season. Ant activity was higher in 1977 than 1976 at all periods; this was correlated with increased rainfall during the dry season. Measures of species abundance and activity were positively correlated with protection from physical stresses. Over all seasons, fewer species and fewer individuals were collected at drier, sunnier sites. Patterns of seasonal and annual change were similar at bait transects and in Berlese samples. Species that were both active at baits and collected in Berlese samples varied in abundance the same way in both sampling methods. Several lines of evidence suggest that moisture availability is an important contributor to these patterns of among—site and among—season variation. First, ant activity, measured in several different ways, increased rapidly when moisture availability was increased, either through rainfall or through experimental watering during the dry season. Second, moisture availability may affect the distribution of suitable nest sites. This hypothesis was supported by differences in the number of colonies collected in Berlese samples from sites which differ in moisture availability. Third, the availability of food among sites may be correlated positively with soil moisture content, but there was not enough evidence to test this possibility. Fourth, there is some evidence that reduction of army ant raids in drier parts of the island during the dry season may concentrate the effects of these ant predators in moist, sheltered areas of BCI. Establishment and persistence of a given species in an area is a complex function of the availability of suitable food and nest sites and the strength of competitors and predators. Patchy species distributions were more the result of changes in the frequency of a species' occurrence among areas, than due to the existence of unique groups of species at separate sites, at least during the periods I sampled. Thus physical conditions were probably at least marginally acceptable to a large fraction of the fauna at all sites. However, differences in abundance and activity of ant species were correlated with changes in moisture availability. I have considered four hypotheses to account for these patterns, but they are by no means exhaustive. Moisture availability, acting through many intermediate steps, has major effects on the small—scale composition of ant faunas, even in a relatively sheltered environment like a tropical forest floor.