A number of analytical tools were applied to estimate fish species richness in a small tropical stream (Brazil) that had been sampled by conventional methods (double stick net, electrofishing) and rotenone. Three sites were compared in two seasons each and a total of 34 species were captured. The sampling efficiency of electrofishing was much better than that of the double stick net and little worse than of rotenone. No estimator was thoroughly consistent and accurate in predicting species richness on the basis of electrofishing, but the Homogenous Model, which assumes that all species have the same detection probabilities, was best. The rarefaction technique should be used with caution as it may produce misleading results in species richness predictions.
