Rift Valley fever (RVF) epidemics have serious consequences for human and animal health and the livestock trade. Recent epidemics have occurred in previously unaffected regions, increasing concerns that the geographical range of RVF will continue to expand. We conducted an extensive, systematic review of the literature to obtain serological data for RVF in Africa, collected between 1970 and 2000 from human, livestock and wild ungulate populations. Aims were to calculate sub-national estimates of RVF infection prevalence and to define areas where no information was available. We presented the data (aggregated at the first administrative level of countries) using a geographical information system. Data from 71 publications were used to build a spatially explicit Bayesian logistic-regression model, with spatial and non-spatial random effects, allowing us to identify clusters of high and low RVF seroprevalence, and fixed effects that described the disparate nature of the survey subjects and methods. Significant high-prevalence clusters encompassed areas that had experienced epidemics during the late 20th century and significant low-prevalence clusters were located in contiguous areas of Western and Central Africa.