The increase in industrial activity has increased human exposure to metals and metalloids exponentially. In this context, arsenium is a good example, being increasingly found in several natural environments. It is an element with no essential biological function, and possessed of elevated toxicity. The discharge of toxic elements in aquatic bodies can promote significant alterations in physical, chemical and biological characteristics, of both the receiving body and the metal itself, regarding its original state. In Brazil, the legislative resolution CONAMA nº 357 of March 17th, 2005 (CONAMA, 2005) is used, which fixes the maximum limits of contaminants in water. The determination of trace-elements in sea-water is of great importance, due to the possibility of bioaccumulation and bioamplification in the food chain. Thus, it is important to establish an easy and efficient method to determine trace-elements in marine waters. The present study aims to compare two arsenium determination techniques for ultra-trace levels: atomic fluorescence spectrometry coupled to hydride generation and ICP-MS with the use of a collision cell. The former has as its major advantage its own hydride generation, that separates the analyte from the matrix, minimizing possible interferences while the latter, due to the use of the collision cell, proposes the determination of As even in the presence of high chloride concentrations, while avoiding isobaric interferences. The samples used in this study were collected from the Brazilian coast. Using HG AFS, the samples were diluted 1+1, mixed in a solution of HCl and KCl and ascorbic acid, in order to obtain final concentrations of 3,5 mol L-1, 1 and 0,2% (m/v), respectively. The reducing solution responsible for the hydride generation was consisted of NaBH4 1,5% (m/v), dissolved in NaOH 0,4% m/v. The analysis were conducted making use of external calibrations with aqueous standards, obtaining a LOD of 36 ng L-1 (3sd), sampling frequency of 45 h-1 and RSD of 2%, obtained from ten measurements of a spike of 500 ng L-1 in one of the samples. When using the ICP-MS methodology, the samples were diluted 10 times (1 + 9). In this case, the calibration conducted was a standard addition, due to interference of the saline matrix in the analysis. A collision cell with He gas was used, in order to minimize saline matrix interferences. The HG AFS technique was demonstrated as the most adequate choice for the direct determination of total As in sea water, due to its adequate limits of detection and absence of interferences. A pre-concentration method was not necessary, and external aqueous calibration was possible. The method was demonstrated as precise and quick, although there was need of a prereduction step. On the other hand, the ICP-MS with direct sampling aspiration, even with the use of a collision cell, did not lead to the expected results while analyzing certified reference material of sea water, while the HG AFS technique did. An observation was made that, in conditions where isobaric interferences were avoided, the limit of detection became improper for the analysis of the samples in question.