A stagnant film model of the effects of thermosolutal convection on free dendritic growth of alloys is dc eloped, and its predictions are compared to available earth-based experimental data for succinonitrile-acetone alloys. It is found that the convection model gives excellent agreement with the measured dendrite tip velocities and radii for low solute concentrations. However, at higher solute concentrations the present predictions show some deviations from the measured data. and the measured (thermal) Peclet numbers, tend to Call even below the predictions from diffusion theory. Furthermore. the measured selection parameter sigma is significantly above the expected value of 0.02 and exhibits strong scatter. It is shown that convection is not responsible Cor these discrepancies. Some of the deviations between the predicted and measured data at higher supercoolings could be caused by measurement difficulties. The systematic disagreement in the selection parameter for higher solute concentrations and all supercoolings examined, indicate that the theory for the selection of the dendrite tip operating state in alloys may need to be reexamined. (C) 2002 Elsevier Science B.V. All rights reserved.