Time-to-collision (Tc) estimates were obtained from twenty-four subjects who viewed film clips for varying lengths of time. The film clips showed the view from a moving car travelling towards a stationary target car, but ended 100 m before reaching the target. Viewing time varied from 2 to 6 s, approach velocity from 40 to 100 km h-1, and Tc from 3.6 to 9.0s. It was hypothesised that, if time were needed to calculate Tc, the accuracy of Tc estimates would increase with viewing time up to some maximum. However, the results showed no effect of viewing time, and this was taken to indicate that estimates were based upon information directly available from the changing optic array at the eye of the observer. A significant velocity effect was found, accuracy increasing with velocity. Since velocity was inversely correlated with Tc, this probably implies that accuracy decreases with increasing Tc. Sex differences were found, with males giving higher and more accurate estimates than females. The relevance of these findings to the nature of Tc information is discussed.