The mercury contents of 109 reference samples of 12 organizations have been determined by flameless atomic absorption spectrometry. The homogeneity of mercury in the six USGS samples issued in 1964 was tested twice, a year apart, using unopened bottles of samples. The analysis of variance of these data in the form of a design with a single variable of classification showed that the mercury contents of the 12 sets can be considered homogeneous at F0.95and that 10 of these sets can be considered homogeneous at F0.90, a more stringent test.Data for other reference samples were generally obtained using Youden Squares. This experimental design was used so that we could determine the significance of the variation attributable to some cause on the days on which mercury was determined, to the order in which the determinations were made on the several days, and to the mercury contents of the samples in each design. For the latter, one may assume that the mercury contents of samples will differ and that a significant F ratio may be obtained. For the principal variable of classification, the variation attributable to some cause on the days on which mercury was determined was significant for only 1 of 9 analyses of variance of sets of data but we have not been able to explain the cause of the significance.The mercury contents of the several types of samples, except for micas, dunites, and perhaps the granodiorites, depend on the origin of the samples. The mercury contents of ande‐sites and basalts of continental origin are higher than those for similar samples originating in island arcs or the circum‐Pacific belt. We conclude that the high and extremely variable mercury content of W‐1 is probably due to mercury resulting from the use of mercury fulminate detonators in the quarry prior to the collection of the s