Non-spectral interferences caused by a sea-water matrix diluted 5-fold on the analyte signals of 24 trace elements in ICP-MS were studied in relation to instrumental parameters. Both quadrupole (ICP-QMS) and double focusing-sector (ICP-SMS) ICP-MS were studied. The parameters were torch injector diameter, sampling depth (i.e., distance from the load coil to the sampler orifice) and rf power. A distinction was made between absolute matrix effects, expressed as integrated signal recovery over a range of nebulizer flow rates (NFR), and matrix effects monitored at a fixed NFR. For the elements studied, it was found that the degree of non-spectral interference depends on both ionization potential (IP) and atomic mass. Generally, the greater the IP and atomic mass of the analyte, the lower the recovery in the presence of the matrix. It appears that, by using optimum plasma settings in ICP-SMS, analyte signal changes may be kept to a moderate level during prolonged introduction of a sea-water matrix. The accuracy of the analytical results can be further improved by using a systematically selected set of internal standard elements.