A method is described to correct for the gradient in intensity of the irradiating neutron flux, which is a potential source of error in neutron activation analysis. For rock samples containing >3% total FeO, the Fe content can be determined by X-ray spectrometry with a coefficient of variation of close to 0.25%. This error rises to ∼0.5% for samples containing 1% total FeO. In instrumental neutron activation analysis, the two strong Fe photopeaks at 1099 and 1292 keV can be measured with a comparable precision. Hence the Fe content determined by X-ray spectrometry can be used to correct for the effects of flux gradient in activation analysis. Replicate analyses of a sample of chilled margin from a Tasmanian dolerite show that the determination of 20 elements by activation analysis of duplicate samples gives errors having coefficients of variation ranging from 0.5% to 5.4%, with 9 elements having a precision of better than 2%. The small range of Sc in 36 samples of chilled margin from the Tasmanian dolerites, from 40.4 to 43.1 ppm, is shown to be geologically significant, and there is a good correlation between Sc abundance and Mg-number. This shows that small but significant amounts of fractionation occurred in the dolerite magma prior to intrusion.