Abstract
In a point-light walker display we can recognize both event and actor. Recognition is reduced when the display is upside down. This may be due to difficulty recognizing a person in an unfamiliar orientation, or it may reflect difficulty recognizing the event because the dynamic relations are unfamiliar (in such a display objects would fall up). To disentangle these two explanations displays were created with walkers that were upside down without reversing gravity — they walked on their hands.
A signal detection task was employed to evaluate subjects' ability to detect humans in four types of point light displays: Walking on feet right-side up (actor and gravity in a familiar orientation), walking on feet upside down (actor and gravity in an unfamiliar orientation), walking on hands right-side up (actor unfamiliar and gravity familiar), walking on hands upside down (actor familiar and gravity unfamiliar).
To make the displays, an actor walked on feet and then on hands on a walkway with a low ceiling so that when walking on feet they could move their hands along the ceiling, and when walking on hands they could move their feet along the ceiling. The displays were based on the x-y coordinates of all visible joints. Noise was added to displays in the form of masking dots that moved in the same manner as the walker points but displaced in random direction and phase. Target-absent trials were matched for density by adding additional random-phase elements. Prior to testing, subjects were shown each mode of locomotion without a mask.
Accuracy for walking on feet replicated previous studies: D′ for upside down walkers was lower than for right-side up walkers. In the walking-on-hands conditions, turning the walker upside down reduced accuracy. Familiarity of dynamic relations was more important than familiarity of actor orientation. Perception of biological motion may require recognition of events, which in turn may allow recognition of the actor.
Research supported by National Eye Institute Grant RO1 EY13518