Several highland regions of Africa recently have suffered malaria epidemics. Because malaria transmission is unstable and the population has little or no immunity, these highlands are prone to explosive outbreaks when densities of Anopheles exceed critical levels and conditions favor transmission. If an incipient epidemic can be detected early enough, control efforts may reduce morbidity, mortality, and transmission. Here we present three methods (direct, minimum sample size, and sequential sampling approaches) that could be used to determine whether the household indoor resting density of Anopheles gambiae s.l. has exceeded critical levels associated with epidemic transmission. Data on Anopheles density before, during, and after a malaria epidemic (December 1997–July 1998) in the highlands of southwestern Uganda were evaluated to demonstrate the application of these three approaches. During this epidemic, a density of 0.25 Anopheles mosquitoes per house was associated with epidemic transmission, whereas 0.05 mosquitoes per house was chosen as a normal level expected during nonepidemic months. The direct approach to calculating mean Anopheles density with an allowable error of 20–50% of the mean would require the sampling of 102–16 houses, respectively. In contrast, with only seven houses, the minimum sample size approach could be used to determine whether Anopheles density had exceeded the critical level. This method, however, would result in an overestimation of the risk of an epidemic at low Anopheles density. Finally, a sequential sampling plan could require as many as 50 houses to conclude that risk of an epidemic existed, but this disadvantage is offset by the ability to preset the probabilities of concluding that risk of an epidemic exists at both the critical and normal Anopheles densities. Our study illustrated that it is feasible, and probably expedient, to include monitoring of Anopheles density in highland malaria epidemic early warning systems.