The leeway of 20-ft containers in typical distress conditions is established through field experiments in a Norwegian fjord and in open-ocean conditions off the coast of France with a wind speed ranging from calm to 14 m s(-1). The experimental setup is described in detail, and certain recommendations were given for experiments on objects of this size. The results are compared with the leeway of a scaled-down container before the full set of measured leeway characteristics are compared with a semianalytical model of immersed containers. Our results are broadly consistent with the semianalytical model, but the model is found to be sensitive to choice of drag coefficient and makes no estimate of the crosswind leeway of containers. We extend the results from the semianalytical immersion model by extrapolating the observed leeway divergence and estimates of the experimental uncertainty to various realistic immersion levels. The sensitivity of these leeway estimates at different immersion levels are tested using a stochastic trajectory model. Search areas are found to be sensitive to the exact immersion levels, the choice of drag coefficient, and somewhat less sensitive to the inclusion of leeway divergence. We further compare the search areas, thus, found with a range of trajectories estimated using the semianalytical model with only perturbations to the immersion level. We find that the search areas calculated without estimates of crosswind leeway and its uncertainty will grossly underestimate the rate of expansion of the search areas. We recommend that stochastic trajectory models of container drift should account for these uncertainties by generating search areas for different immersion levels and with the uncertainties in crosswind and downwind leeway reported from our field experiments.