Information used in judging impending collision.

Many subhuman species and human infants, children, and adults can use two-dimensional information of relative rate of angular-size change to anticipate collisions between the self and approaching objects or surfaces. But extant studies have not determined what information is used when subjects view simulated approach events providing two-dimensional information and three-dimensional information (distance and distance change), as well as lower-order visual information contained in real approach events. Three experiments suggest that, given these several possibilities, adults' judgments of collision time are best predicted by two-dimensional spatiotemporal values which are invariant over object sizes, distances traversed, approach velocities, and several lower-order variables such as absolute angular size. However, collision time is substantially underestimated, with absolute amount of underestimation increasing as a function of actual time-to-collision. Large constant errors and loss of judgment linearity beyond about 10 s to contact time suggest that current models of human performances based on use of time-to-collision information require modified assumptions of operator efficiency.