The spatial patterns of precipitation variability over the United Kingdom are examined for the period 1861–1970 using principal components analysis. Most of the variance (around 50 per cent) is explained by the first component, which is characterized by uniform variability over the whole region. Higher-order components isolate differential variability between different regions. The first four components are essentially the same for all months, all seasons and the annual total, highlighting the intra-annual constancy of inter-regional variability and pointing to strong geographical and topographical control. A subdivision of the analysis period into two halves produces no change in the component patterns, but changes in the variance explained by the first component imply that summer precipitation has become less spatially variable since 1915. On the basis of the principal components analysis the England and Wales region is divided into five coherent subregions. A regression technique is developed to produce a homogeneous area-average precipitation series for England and Wales using the longest site precipitation records available and maintaining even spatial coverage. The method used allows changes in the reliability of the area-average due to changes in the density of station coverage to be quantified. An empirical relationship between the standard error of estimate of area-averaged precipitation and the number of recording sites is derived. Significant inhomogeneities in the Nicholas and Glasspoole England and Wales precipitation series at 1915 and 1957 are detected and quantified. The Nicholas and Glasspoole series as a whole is found to underestimate precipitation amounts noticeably in the late eighteenth and early nineteenth centuries.