Five experiments addressed the perception of curvilinear heading under various conditions of optical flow. Perception of heading was unaffected by optical noise (Experiment 1) and was successful and equally accurate for flows generated by circular and elliptical paths of locomotion (Experiment 2). In Experiment 3, random perturbations of vector magnitudes in general curvilinear fields did not reduce the accuracy of perceived heading. When vector directions were randomly perturbed in Experiments 4 and 5, curvilinear heading perception was impaired. Discussion focuses on the inappropriateness of the vector normal hypothesis to the general curvilinear case, the importance of the pattern of relative vector directions, and the practical and theoretical significance of investigating different forms of noncanonical optical flow.