Three numerical experiments have been carried out with a global coupled ice-ocean model to investigate its sensitivity to the treatment of vertical mixing in the upper ocean. In the first experiment, a widely used fixed profile of vertical diffusivity and viscosity is imposed, with large values in the upper 50 m to crudely represent wind-driven mixing. In the second experiment, the eddy coefficients are functions of the Richardson number, and, in the third case, a relatively sophisticated parameterization, based on the turbulence closure scheme of Mellor and Yamada version 2.5, is introduced. We monitor the way the different mixing schemes affect the simulated ocean ventilation, water mass properties, and sea ice distributions. CFC uptake is also diagnosed in the model experiments. The simulation of the mixed layer depth is improved in the experiment which includes the sophisticated turbulence closure scheme. This results in a good representation of the upper ocean thermohaline structure and in heat exchange with the atmosphere within the range of current estimates. However, the error in heat flux in the experiment with simple fixed vertical mixing coefficients can be as high as 50 W m-2 in zonal mean during summer. Using CFC tracers allows us to demonstrate that the ventilation of the deep ocean is not significantly influenced by the paramertization of vertical mixing in the upper ocean. The only exception is the Southern Ocean. There, the ventilation is too strong in all three experiments. However, modifications of the vertical diffusivity and, surprisingly, the vertical viscosity significantly affect the stability of the water column in this region through their influence on upper ocean salinity, resulting in a more realistic Southern Ocean circulation. The turbulence scheme also results in an improved simulation of Antarctic sea ice coverage. This is due to to a better simulation of the mixed layer depth and thus of heat exchanges between ice and ocean. The large-scale mean summer ice-ocean heat flux can vary by more than 15% between the three experiments. Because of this influence of vertical mixing on Southern Ocean ventilation, sea ice extent, and ocean-atmosphere heat fluxes, we recommend that global climate models adopt a sufficiently realistic representation of vertical mixing in the ocean. ¿ 1999 American Geophysical Union