Previous work (Harris and Giachritsis 2000, Vision Research40 601–611) has shown that, when global and local image expansion are placed in conflict, estimates of time-to-contact (TTC) are based almost exclusively upon global expansion. Here we extend this finding by demonstrating that global image expansion continues to predominate even under conditions that seem more favourable to a local analysis. We added a global rotation to the stimulus so that the global pattern of expansion was distorted while leaving the local expansion unaffected. Even under relatively high rotation rates (30° s−1), local expansion continued to have little systematic effect upon estimates of TTC.