The geoid signal and the flow patterns of two-dimensional steady state convection models with exponential temperature- and depth dependent viscosity are compared with results for an equivalent stratified viscosity structure. In analogy to Richards and Hager [1989>, the later are computed by a ''dynamic response'' approach. The flow fields obtained with this approach are quite different from the full solution; the geoid signals are similar but the amplitudes differ significantly. The differences are analysed in the horizontal wavenumber domain and in the spatial domain. They may lead to an overestimation of the viscosity constrast of the earth's mantle derived by modeling the earth's geoid with dynamic response functions. There are implications on the interpretation of observed geoid anomalies in terms of up- or downwelling flow structures. ¿American Geophysical Union 1992