Abstract
We perceive causality even in simple kinematic displays—a moving disk that stops next to another disk appears to launch the second disk into motion. Visual adaptation alters perceived causality in retinotopic coordinates, suggesting that causality is computed early in the visual system. Here, we used adaptation to show that causality detection is tuned to a key low-level feature of the visual event—its motion direction. Observers saw brief test events in which a peripheral disk moved swiftly towards a stationary one and, as soon as they overlapped by some amount (zero to full overlap in seven steps), the first disk stopped and the second disk started to move along the same trajectory. Observers then reported whether they perceived the first disk to pass over the stationary one (reported commonly at full overlap), or rather, to launch it into motion (common at zero overlap). We presented test events in one of two horizontal directions both before and after adaptation to causal launches presented in a narrow range around one of these two directions. To induce adaptation, we exposed participants to a stream of 320 launch events before the first adaptation test trial, topped-up with an additional stream of 16 launch events before each subsequent trial. For each observer, adaptation decreased the amount of overlap that could elicit perception of a causal launch, replicating previous findings. This negative aftereffect was strongest when the test event’s motion direction matched the adapted one. The direction-specific tuning cannot be accounted for by adaptation to non-causal features of the adaptation stimulus, as no adaptation occurred after exposure to control events, designed to match the launch in as many physical properties as possible. Direction-specific adaptation to causal events demonstrates that the computation of phenomenological causality relies on low-level routines implemented at early stages in visual processing.