The difficulty of obtaining an adequate quantitative estimate of the composition and van-ability of a granitic complex is discussed. Sampling problems are considered in relation to the sampled and target populations. Trend-surface analysis is an efficient method of studying selected areas of the Donegal “older granite.” Different levels of sampling are compared. The trend components for several mineralogical variables and the results of regression analysis suggest that one specimen per quarter square mile gives a reasonable approximation of the trend components of the granite; when averaged these data yield comparable trend components more economically than the modes for individual data points. Use of only one specimen per square mile gives spurious results, although with this level of sampling some variates give better results than others (e.g., color index gives better results than total feldspar percentage). The most efficient sampling plan for effective quantitative estimates to be made varies according to the objective of the study. Although averages yield good trend surfaces, they mask important features in the deviation maps. Closely spaced individual data points appear to allow better correlation between systematic deviations and the underlying geologic causes of such local departures (deviations) from the regional trend.