A series of steady velocity fields at the core surface have been calculated using main field and secular variation models from the period 1900–1980 in the frozen-flux approximation. Secular variation model uncertainties have been estimated from the main field models of Bloxham and Gubbins (1986). For the period 1940–1980 it was possible to derive four consecutive flows, each assumed steady over 10 years, which fitted the spherical harmonic models to within their assumed uncertainties. For the period 1900–1940 it was found that the flows fitted the models if the SV model uncertainties were increased by a factor of one and a third for the period 1920–1940 and one and two thirds for the period 1900–1920. With the original assumed uncertainties, four consecutive flows, assumed steady over 5 years, were calculated for the period 1920–1940. For the period 1900–1920 it was only possible to derive steady flows over an interval of 2 years which fitted the SV models to within their assumed uncertainties over the periods 1900–1909 and 1917–1920. The computed flows are highly suggestive of a time-dependent flow with a dominant toroidal component which was increasing in the period 1900–1920, decelerating between 1910 and 1920 and reaching a minimum around 1930, and again increasing to a maximum in the mid-1950s. The flows computed before and after the 1969.5 secular acceleration impulse imply that the flow slowed down through the jerk event.