In the present study the empirical orthogonal function or ‘eigenvector’ approach is used to determine the dominant rainfall patterns from normal seasonal rainfall records over Rajasthan. Two contrasting years (1917 and 1918) in which rainfall was in excess and deficient are also examined separately to see what anomalies, if any, exist in the associated patterns. Empirical orthogonal functions or eigenvectors are derived from the sets of 12-monthly rainfall values of 40 stations in Rajasthan. In the years of normal rainfall the first eigenvector is found to account for 99%; of the variance in the original 12 × 40 matrix of rainfall data, thus indicating that the entire area is homogeneous as far as the normal seasonal variation of rainfall is concerned. However, in years of excessive or deficient rainfall, 3 or 4 vectors are needed to account for 99%; of the variance. The first eigenvector in practically all cases largely resembles the seasonal variation of rainfall over the area, while the higher-order eigenvectors arise mainly as adjustment vectors to account for the balance of the variance. The eigenfunctions are used to estimate the mean monthly rainfall for places having no rainfall records. It is found that a reasonably good estimate of the normal seasonal distribution of rainfall over Rajasthan is given by just one vector.