Studies of the impact of hypoxic or anoxic environments on both climate and ecosystems rely on a detailed characterization of the oxygen (O-2) distribution along the water column. The former trivial separation between oxic and anoxic conditions is now often redefined as a blurry concentration range in which both aerobic and anaerobic processes might coexist, both in situ and during experimental incubations. The O-2 concentrations during such incubations have often been assumed to be equal to in situ levels, but the concentration was rarely measured. In order to evaluate the actual oxygen concentration in samples collected from low-oxygen environments, a series of measurements were performed on samples collected in the Pacific oxygen minimum zones. Our results show a significant deviation from in situ anoxic conditions in samples collected by Niskin bottles where leakage from the bottle material resulted in O-2 concentrations of up to 1 mu M. Subsequent sampling further increased the O-2 contamination. Sampling and analysis by Winkler method resulted in variable apparent concentrations of 2-4 mu M. Two common procedures to avoid atmospheric contamination were also tested: allowing gentle overflow and keeping the sampling bottle submersed in a portion of the sampled water. Both procedures resulted in similar O-2 contamination with values of 0.5-1.5 mu M when bottles were immediately closed and measurements performed with optical sensors, and 3-4 mu M apparent concentration when analyzed by the Winkler method. Winkler titration is thus not suited for analysis of low-O-2 samples. It can be concluded that incubation under anoxic conditions requires deoxygenation after conventional sampling.