Memory for visual motion.

Observers briefly viewed random dots moving in a given direction and subsequently recalled that direction. When required to remember a single direction, observers performed accurately for memory intervals of up to 8 s; this high-fidelity memory for motion was maintained when observers executed a vigilance task during the memory interval. When observers tried to remember multiple directions of motion, performance deteriorated with increasing number of directions. Still, memory for multiple directions was unchanged over delays of up to 30 s. In a forced-choice experiment, observers viewed 2 successive animation sequences separated by a memory interval; for both sequences, dots moved in any direction within a limited bandwidth. Observers accurately judged which animation sequence was more coherent, even with memory intervals of 30 s. The findings are considered within the context of cognitive bias and memory for other aspects of perception.