Abstract As part of a EUROMET project to produce a range of γ-ray emitting radioactive spiked reference materials, a technique, based on analysis of variance, was developed to evaluate the uncertainties in characterising such materials. A particular concern was to ensure that users of these materials are properly informed regarding the consequences of any remaining inhomogeneities in them. A minimum of five random samples of each reference material were counted six times on a suitable γ-ray spectrometer. Once all of the data were collected and normalised, between-sample ( MS BS) and within-sample ( MS WS) variances were calculated. The quantity MS WS includes factors associated with sample measurement such as counting statistics, background stability, detector response stability, source-to-detector geometry and electronic stability. The quantity MS BS includes not only all of the terms in MS WS, but also includes terms for sample inhomogeneity, mass determination and density determination. These quantities enable the following calculation to be performed; (1) σ mat 2 = σ tot 2 − σ meas 2 where σ tot is the overall uncertainty, σ meas is the measurement uncertainty and σ mat is the uncertainty arising from sample inhomogeneity; it is assumed here that the effects arising from mass and density determinations are negligible. The value of σ mat is only valid at the sample weight which it was measured, and data are presented to show the variation of the uncertainty arising from sample inhomogeneity with changing sample weight.