The influence of sea surface temperature (SST) on wet season rainfall in North East Brazil is presented on time scales from individual seasons to a century. First, we analyse rainfall variability and predictability for 1912-1998 using dynamical and empirical techniques. Dynamical assessments are mainly based on the HadAM2b atmospheric model forced with the Met Office Global sea Ice and Sea Surface Temperature data set (GISST3.0). The results demonstrate a relatively high degree of seasonal predictability whose source lies largely in the effects of varying north-south gradients of SST across the tropical Atlantic and a remote influence of El Nino. A relatively high level of real-time skill is also shown by the performance of the seasonal forecasts issued by the Met Office between 1987 and 2002. In addition, there is some evidence of the coherent effects of SST on decadal time scales. Finally, we show evidence from a 1400 year control run of the Met Office HadAM3 coupled model that variations in the Atlantic thermohaline circulation may modulate North East Brazil rainfall on century time scales through variations of the north-south SST gradient in the tropical Atlantic.