Theoretical predictions from analytical, self-gravitating, layered earth models with Maxwell, Burgers' body, and standard linear solid rheologies are employed to explore the extent to which transient creep may have influenced the inference of long-term mantle viscosity from analysis of the various signatures derived from the process of glacial isostatic adjustment. The compatibility of the various models is checked by their ability to fit simultaneously the relative sea level, free air gravity, and the rotational data. We conclude that transient creep cannot operate throughout the entire mantle. On account of the extreme sensitivity of the uplift and gravity data to any significant departures of the steady upper mantle viscosity from 1021 Pa s, transient rheology would most likely be unimportant in the upper mantle. Restricting transient rheology to the lower mantle, we find that models with an adiabatic density jump at 670 km depth are not satisfactory because of their inability to produce sufficiently large gravity anomalies and polar wander speeds. Our calculations show that the preferred models with transient creep in the lower mantle are chemically stratified with a nonadiabatic density change of around 10%. The long-term viscosities in the lower mantle, associated with compatible transient rheological models, lie between 5 × 1022 and 10 23 Pa s. These values are consistent with viscosity estimates derived from correlation of long-wavelength geoid anomalies and the density perturbations in the deep mantle.