Matching biological motion across viewpoints

Abstract

There has been much debate as to how objects can be recognized across viewpoint changes. Here we ask whether viewpoint changes affect performance when participants make judgements about human actions depicted as point-light stimuli. Previous research has suggested that bodies may be “special” objects and may thus be immune to such viewpoint costs. We used a concurrent matching task in which three dynamic point- light figures performed familiar actions taken from a standard biological motion database. On each trial the action performed by the central “target” figure was also performed by one of the two flanking figures. The task was to make a speeded left/right response to indicate which flanker was copying the target. Separate, random depth orientations were assigned to the two flanking figures and the target could either have the same orientation or appear with an offset of 45◦ or 90◦ relative to the matching flanker. The starting animation frame was randomly chosen for each of the three figures. We found that viewpoint differences between the target and matching flanker affected both speed and accuracy. This indicates that the recognition of human bodies depicted as biological motion stimuli is viewpoint-dependent, as with many other types of object. We also suggest that concurrent matching is a flexible tool for exploring biological motion as decisions can be made on a variety of actions without the need for explicit action-naming or training.