Turbulence measurements were made off the northwest coast of Spain in January and August 1998. In winter the water column was vertically mixed to about 100 to 150 m, due to the combined effects of the vertical convection of warm northward-moving water and wind stress. A highly dissipative surface boundary layer was present at all times to a depth (of about 20 m) that correlated well with the local wind and wave amplitude. Below this layer dissipation levels decreased from about 10-7 m2 s-3 at a rate that was commensurate with 'law of the wall' boundary theory. Near the coast local brackish surface stratification served to depress mixing below the pycnocline. In summer, when the water column was thermally stratified, average dissipation levels were typically an order of magnitude smaller than in winter, even though the wind stress in the ocean was of similar magnitude. Bursts of enhanced mixing were occasionally observed in an internal wave field on the shelf. Dissipation levels were higher on the northern side of an upwelling filament (up to 10-7 m2 s-3) than in other parts of the ocean. Although eddy viscosity levels on the shelf and in the ocean were almost identical (about 8 cm2 s-1), eddy diffusion on the shelf (0.37 cm2 s-1) was about three times larger than in the ocean. This may indicate a higher frequency of mixing events on the shelf. The summer data were used to determine a mixing length (of about 0.3 ± 0.05 m) using an algorithm that mimicked the way that turbulence closure models compute dissipation from vertical shear and buoyancy over grid scales of several meters. The correlation between dissipation and the gradient Richardson number was poor and it is suggested that at the scales of the observations, and of some models, buoyancy is just as likely to act as a source of mixing as it is to act as a sink.