A wealth of information on climate and weather in Africa during the nineteenth century is available from documentary sources.  Unfortunately long-term time series from a single observer at a single location are uncommon.  This paper deals with a methodology that allows such sources to be utilized and combined with gauge data and natural proxies, such as information on lake levels.  This approach reduces the problems of calibration and observer bias and allows for fragmentary information from many locations to be optimally utilized.  The resultant data set has seven classes that represent anomalous conditions of rainfall, ranging from extreme drought to extremely wet.  Resolution is annual and information is available for some 90 regions on the continent, beginning in the 1820s.  The data set can be evaluated with standard climatological methods, including principal component analysis, linear correlation, and spectral analysis.  Time series can be extended to the present time by converting modern data to the seven-class system.  In this talk, the data set and the results of various statistical analysis are presented.  A comparison is made with results for the twentieth century.  The relationship to factors governing interannual variability, such as sea-surface temperatures, is also explored.