Several procedures for the determination of Ca, Mg and Sr in soils have been compared on the basis of the accuracy of analysis of two NIST reference materials (Montana Soils SRM 2710 and SRM 2711). Samples were dissolved in a mixture of hydrofluoric and nitric acids in sealed vessels in a microwave oven and in teflon beakers on a hot plate. The digests obtained from both dissolution methods were evaporated to dryness in an attempt to remove silicon. Boric acid was added to prevent the precipitation of the lanthanum releasing agent (as lanthanum fluoride) and potassium was added as an ionization buffer. Determinations were made by flame atomic absorption spectrometry with both the nitrous oxide-acetylene flame and the air-acetylene flame, with calibration either by standard additions or against external standards matrix matched with respect to nitric acid, boric acid, lanthanum and potassium. The silicon remaining in the solution was also determined by external calibration. A single-line flow injection manifold was used to overcome any problems due to the presence of high dissolved solids. A volume of 300 mul was injected into a water carrier stream flowing at 8 ml min(-1). To determine Ca in the air-acetylene flame, it was necessary to remove silicon. Magnesium was determined in either flame without complete removal of the silicon, however, for the determination of Sr, it was necessary to remove the silicon and use the nitrous oxide-acetylene flame. The indicative value for Sr in SRM 2710 was too low: the value determined was 360+/-30 mug g(-1).