The atomic absorption spectrophotometric (AAS) determination of (\mathrm{Fe}, \mathrm{Ca}), and (\mathrm{Mg}) in the widely used fluoboric-boric acid matrix (\left(\mathrm{HBF}{4}-\mathrm{H}{3} \mathrm{BO}{3}\right)) is seriously interfered with by (\mathrm{Si}) and (\mathrm{Al}). These interferences can be overcome by a matrix buffer (a mixture of (\mathrm{KCl}), tartaric acid, (\mathrm{HCl}), and (\mathrm{H}{2} \mathrm{SO}{4}) ). Based on this, an AAS method involving (\mathrm{HF}-\mathrm{HClO}{4}(2: 1) mixture) decomposition has been proposed and results are given for determination of (\mathrm{Al}, \mathrm{Fe}, \mathrm{Ca}), and (\mathrm{Mg}) in 12 Japanese rocks along with three well-characterized USGS geochemical samples covering the range of compositions most often encountered in routine analysis. Alkali contents (\left(% \mathrm{Na}{2} \mathrm{O}\right.) and (\left.% \mathrm{~K}{2} \mathrm{O}\right)) have been estimated using a reverse alkali buffer (i.e., (\mathrm{KCl}) for (\mathrm{Na}{2} \mathrm{O}) and (\mathrm{NaCl}) for (\mathrm{K}{2} \mathrm{O}). A mixture of (\mathrm{KCl}), tartaric acid, and (\mathrm{SrCl}_{2}) has been found to be an appropriate buffer for (\mathrm{Mn}). Statistical data for all estimations are presented, and precision values are comparable with those of conventional methods used for silicate analysis. For most of the samples the agreement between the observed value and the published data is extremely good. Deviations have been noted in some samples where values are of "proposed" status, indicating that more data are still required to upgrade the status of some elements from the proposed/conditional to the certified level. 1994 Academic Press. Inc.