The free air gravity anomaly over Canada, which has been associated with glacial isostatic, disequilibrium, has traditionally been an important datum in the inference of deep mantel viscosity. In this study we investigate the sensitivity of the inference on the assumed space and time dependent geometry of the Laurentide ice sheet. We find, in particular, that the computed peak anomaly is extremely sensitive to the timing of the final ice sheet collapse at the centre of Laurentia; a fact which has not previously been fully appreciated or quantified. As an example, the deep mantle viscosity preferred on the basis of a deglaciation model which maintains an equilibrium plastic profile can differ by a factor of four from that obtained using a model with the same maximum thickness but a fixed ice sheet perimeter. Furthermore, discrepancies in the inference of deep mantle viscosity can reach orders of magnitude when certain commonly used models of the deglaciation process are adopted. ¿American Geophysical Union 1991