Sampling errors in eddy correlation flux measurements arise from the small number of large eddies that dominate the flux during typical sampling periods. Several methods to estimate sampling, or random error in flux measurements, have been published. These methods are compared to a more statistically rigorous method which calculates the variance of a covariance when the two variables in the covariance are auto- and cross-correlated. Comparisons are offered between the various methods. Compared to previously published methods, error estimates from this technique were 20 to 25% higher because of the incorporation of additional terms in the estimate of the variance. This new approach is then applied to define the random error component of representative eddy correlation flux measurements of momentum, sensible and latent heat, carbon dioxide, and ozone from five field studies, three over agricultural crops (corn, soybean, and pasture), and two from towers over forests (deciduous and mixed). The mean normalized error for each type of flux measurement over the five studies ranged from 12% for sensible heat flux to 31% for ozone flux. There were not large or significant differences between random errors for fluxes measured over crops versus those measured over forests. The effects of stability, flux magnitude, and wind speed on measurement error are discussed. ¿ 2001 American Geophysical Union