The eddy correlation technique is rapidly becoming an established method for resolving dissolved oxygen fluxes in natural aquatic systems. This direct and noninvasive determination of oxygen fluxes close to the sediment by simultaneously measuring the velocity and the dissolved oxygen fluctuations has considerable advantages compared to traditional methods. This paper describes the measurement principle and analyzes the spatial and temporal scales of those fluctuations as a function of turbulence levels. The magnitudes and spectral structure of the expected fluctuations provide the required sensor specifications and define practical boundary conditions for the eddy correlation instrumentation and its deployment. In addition, data analysis and spectral corrections are proposed for the usual nonideal conditions, such as the time shift between the sensor pair and the limited frequency response of the oxygen sensor. The consistency of the eddy correlation measurements in a riverine reservoir has been confirmed-observing a night-day transition from oxygen respiration to net oxygen production, ranging from -20 to +5 mmol m(-2) day(-1)-by comparing two physically independent, eddy correlation instruments deployed side by side. The natural variability of the fluctuations calls for at least similar to 1 h of flux data record to achieve a relative accuracy of better than similar to 20%. Although various aspects still need improvement, eddy correlation is seen as a promising and soon-to-be widely applied method in natural waters.