A procedure suitable for routine analysis of arsenic in orange juice using instrumentation available in most control laboratories was developed. Organic matter is destroyed by the dry ashing technique, the ash is dissolved in (\mathrm{HCl}), and hydrides of arsenic are generated by the addition of sodium borohydride prior to atomization in a flame-heated quartz cell and atomic absorption spectroscopy measurement. Experimental conditions for the destruction of organic matter were established, and an interference study was carried out. Copper and iron do not interfere in arsenic determination at levels found in orange juice. The methodology developed has a detection limit in orange juice of (0.1 \mathrm{ng} \mathrm{g}^{-1}), a relative standard deviation of (1 %), and a recovery percentage of (98 \pm 8). The accuracy of the proposed method was checked by analysis of a certified sample of tomato leaves, NIST SRM 1573 (certified 0.27 (\pm 0.05 \mu \mathrm{g} / \mathrm{g}); found (0.26 \pm 0.03 \mu \mathrm{g} / \mathrm{g}) ). The arsenic content of orange juice products was determined. The extremely low arsenic levels found ( (0.3-4.4 \mathrm{ng} \mathrm{g}^{-1}) ) do not present problems from a toxicological point of view. e 1994 Academic Press, Inc.