BACKGROUND: The incidence of cholera in Bangladesh shows clear seasonality, suggesting that weather factors could play a role in its epidemiology. We estimated the effects of rainfall on the incidence of cholera in Dhaka, Bangladesh. METHODS: We examined time-series patterns of the weekly number of hospital visits due to cholera in relation to weekly rainfall from 1996 to 2002. We used Poisson regression models, adjusted for seasonal variation, between-year variation, public holidays, and temperature. The role of river level on the rainfall-cholera relationship was also examined by incorporating river-level terms into the models. RESULTS: The weekly number of cholera cases increased by 14% (95% confidence interval = 10.1%-18.9%) for each 10-mm increase above the threshold of 45 mm for the average rainfall, over lags 0 to 8 weeks. Conversely, the number of cholera cases increased by 24% (10.7%-38.6%) for a 10-mm decrease below the same threshold of average rainfall, over lags 0 to 16 weeks. River level partly explained the association between high rainfall and the number of cholera cases. CONCLUSIONS: The number of cholera cases increased with both high and low rainfall in the weeks preceding hospital visits. These results suggest that factors associated with river level are on the causal pathway between high rainfall and incidence of cholera.