Tuning for temporal interval in human apparent motion detection

Motion detection in apparent motion of random-dot patterns (RDPs) requires correlation across space and time. It has been difficult to study the temporal requirements for the initial correlation step because temporal measurements jointly depend on temporal filtering, delay-tuning, and successive temporal integration. Moreover, it has been difficult to construct a stimulus containing a single delay only. To measure delay tuning independently of temporal integration, we constructed a motion stimulus containing a single delay value only, and with constant motion energy, irrespective of delay. The stimulus consists of a sparse RDP with a two-frame, single-step dot lifetime. It is constructed by generating a dynamic RDP on each stimulus frame, and showing this pattern once again after a delay of n frames, superimposed on the newly generated RDP. Each frame thus consists of 50% of new random dots and 50% displaced random dots. The delay between corresponding dot patterns c an be chosen freely, without affecting the number of steps per second, steps in total, and temporal-frequency content. We measured left - right coherence thresholds for direction discrimination by varying coherence levels in a Quest staircase procedure, as a function of both step size and delay. Highest sensitivity was found at a temporal delay of 12 - 30 ms. Sensitivity decreased for lower and higher temporal delays. The fall-off at higher delay values was much sharper than previously described. The data allow us to describe to what extent delay tuning in coherence detection is independent of step size.