The dynamics of the circulation near an eastern-ocean boundary are investigated using a 2 1/2 layer numerical model that includes entrainment of cool water into the upper layer. Solutions are found in a regional ocean basin and are forced by an upwelling-favorable, alongshore wind field without curl. Dynamically simpler versions of the model are also utilized in order to isolate the effects of various physical processes: in particular, a linearized instability model is used to study the unstable waves associated with various background coastal circulations. The model is quite successful in simulating many features of the observed flow and SST fields. Initially, the ''main run'' solution spins up as in a linear model, generating a surface jet, an undercurrent, and an upwelling front. Subsequently, small-scale, ''fingerlike'' disturbances appear along the coast and grow rapidly in amplitude and scale. The growth in scale has two causes: the slower development of larger-scale disturbances, and eddy coalescence via nonlinear interactions. Eventually, the solution adjusts to a realistic equilibrium state that contains upwelling filaments, squirts, eddies, dipole eddy pairs, and a realistic mean circulation. Entrainment is a crucial process in the dynamics of the mean flow. In particular, entrainment damps the westward propagation of Rossby waves, thereby ensuring that the currents remain coastally trapped. (In this aspect entrainment plays a dynamical role similar to vertical mixing in continuously stratified models.) Eddies influence the mean flow primarily by providing a strong heat source near the coast, but they have very little effect on the mean momentum field. The small-scale, finger-like disturbances are caused by a frontal instability that requires the existence of an upper-layer temperature gradient. Some dynamically simpler solutions indicate that the larger-scale disturbances are caused by baroclinic instability. On the other hand, unstable-wave solutions to the linearized instability model suggest that their cause is frontal instability, becoming pure baroclinic instability only when there is no temperature gradient. ¿ American Geophysical Union 1991