This paper shows the influence of different sample storage protocols, on the chemiluminescence signal of some metal ions. The storage protocols studied were: acid addition (HCl or HNO(3)) and no reagent addition to filtered and refrigerated (T=4 degrees C) samples. Light emission was produced for the chemiluminescence reaction between luminol and hydrogen peroxide in buffer carbonate conditions (pH 10.8) catalysed by Cr(III), Co(II) and Cu(II). Batch and/or flow modes in different conditions were tested. Fe(II), Fe(III), Ni(II) and Mn(II) did not give chemiluminescence in the studied conditions. A parallel study of sensitivity and selectivity was performed. Then the presence or absence of the masking agent EDTA, added to samples or used in the carrier stream, is assayed. If the samples are acidified with HNO(3), a previous neutralisation is needed using batch mode. The determination of Cr(III) is independent of storage protocol by flow injection (FI) method; however, the determination of Co(II) or Cu(II) or total determination of three metals requires the conditioning of standards. Detection limits achieved are ranged between 0.5 and 2 mug l(-1). For batch mode, detection limits are better for unacidified samples and worse for carbonate-neutralised samples. The influence of storage protocols was validated using standard metal mixtures and calibration solutions. The use of standard reference material (SRM(c) 1640) (Trace elements in natural water) corroborates the previous statements and validates the accuracy of the different approaches underlined. This paper demonstrates that it is possible to determine Cr(III) selectively in natural waters.