Based on recent theoretical work by Robert May and Richard Levins, two hypotheses about time fluctuations in abundance of competing species were generated. Data for isopods and amphipods from four cave stream communities in Virginia and West Virginia were used to test the predictions. First, variance of total abundance should be less than the sum of the variances of individual species’ abundances. In three of four communities studied, the prediction was confirmed, but none were statistically significant. Positive correlations among carrying capacities of competing species may explain the poor agreement with predictions. Second, the signs of 19 correlations and partial correlations of species abundances were predicted on the basis of relative magnitudes of direct and indirect effects of competition, and of these predictions, 16 were confirmed by the data, including 5 statistically significant ones. Most interesting was the finding that competitors can be positively correlated.