The Mellor-Yamada level-2 and 2 1/2, Niiler, and Garwood one-dimensional mixed-layer models were compared for some simple forcing experiments and were tested by simulating changes in the mixed layer at ocean stations November and Papa for the year 1961. The ocean station simulations show that the models can be tuned to give fairly good results. However, the need to readjust model constants when changing locations suggests that there is room for improving some of the mixing parameterizations. The sensitivity of the model simulations to certain ''external'' parameterizations, including surface heat flux, seawater turbidity, and ambient diffusivity below the mixed layer, was investigated. The year-long simulations were found to be very sensitive to the seawater turbidity. Increasing the turbidity from Jerlov optical type I to type III causes a maximum increase in the monthly mean SST at November of 3¿C or more. The simulations are most sensitive to seawater turbidity during summer when the mixed-layer remains shallow. For the ambient diffusivity a decrease from 0.4 to 0.01 cm2/s results in a maximum increase in the monthly mean SST at November and Papa of 0.5 to 1.5¿C, respectively. The effects of using a constant ambient diffusivity are most noticeable in late summer when the seasonal thermocline is strongest. The effects are larger at Papa than at November because of the observed optical water types (I and II, respectively) and relatively small values for ambient diffusivity (less than 0.2 cm2/s).