A bulk analyser based on the gamma-ray resonance scattering process can be used to give a rapid and accurate measure of the percentage by weight of copper or nickel averaged over a large mass of sample, typi- cally 20 kg. It is shown that for a bulk sample containing about 3 wt.% Cu, crushing to -25 mm is the only sample preparation requited for analysis to within an error of about 0.06 wt.% Cu (errors quoted are one standard deviation). A comparison of bulk analyser and con- ventional assays on the same bulk samples can be used to estimate the sample preparation errors in the conventional methods. The squared deviations between bulk analyser and conventional assays are approximated by the sum of variances in sample preparation, conventional analysis, and bulk sample analy- sis. The sample preparation error was esti- mated as 0.26 wt.% Cu for samples crushed to -2 mm before splitting, 0.08 wt.% Cu for samples crushed to -0.35 mm before splitting, and 0.04 wt.% Cu for the average of 15 grab samples from bulk samples which had been crushed to -2 mm. The average copper grade for each group of samples was about 1.7 wt.%. Results are presented which show that