Meander migration models include an as yet poorly investigated source of numerical errors related to the computation of the channel curvature, which are amplified by the procedure of adding and deleting grid points as the river planform evolves. The methods adopted to reduce these errors may influence size, form, and migration rate of the developing meanders, which creates uncertainties in the analysis of the results, limits the model applicability, and makes it necessary to treat the bank erodibility coefficients as calibration parameters. This becomes evident from a series of computational tests performed in order to compare two different methods of error reduction in the computed local channel curvature: cubic spline interpolations versus different levels of curvature smoothing. Since the problems discussed are common to most meander migration models, the tests performed were carried out for three models of different complexity. These were derived by applying different degrees of simplification to the basic equations for water flow and sediment motion of shallow curved channels.