Ozone measurements taken with an aircraft over the western North Atlantic in the summer of 1992 revealed numerous profiles that consisted of two principal layers with different, nearly constant mixing ratios in each layer. The lower layer was characterized by relatively low mixing ratios (<25--30 parts per billion by volume), while the upper layer had values a factor of 2 or more higher. The depths of the lower layer frequently corresponded to the depths of the mixed layer over the ocean as estimated from potential temperature observations. The upper ozone layer extended at least to the maximum sampling height of the aircraft on these flights, approximately 2.5 km. A mesoscale model is used to analyze the meteorology over the area for a representative case study day. Trajectory analysis shows that the air in the aircraft sampling region originated near Hudson Bay, followed a path that avoided major anthropogenic sources of ozone precursors over land, and eventually moved out over the ocean. Because of inhomogeneities in the sea surface temperatures in the observation area, the spatial and temporal evolution of the boundary layer over the ocean differed significantly over distances of only a few hundred kilometers. These differences, coupled with vertical mixing in the boundary layer and large-scale vertical motions arising from synoptic and terrain influences, are identified as mechanisms responsible for the development of the characteristic shapes of the ozone and potential temperature profiles. This conclusion is supported by additional results obtained from using the mesoscale model to drive a Lagrangian particle dispersion model to simulate the transport and diffusion of a passive tracer; the major features of the ozone profiles are reproduced in the particle profiles. These findings indicate the critical role played by meteorological processes, irrespective of the details of the photochemical reactions. The results also imply that a failure to incorporate detailed descriptions of meteorology in photochemical models may lead to erroneous interpretations of the data. ¿ American Geophysical Union 1996