In our daily lives, we constantly perceive, recognize, and understand other people’s actions. Both object- and scene-related information has been shown to contribute to these processes. However, the time course underlying the recognition of actions relative to the recognition of scenes and objects is not well understood. Revealing this information is crucial for developing cognitive models of action understanding. To address this question, we used a free-recall paradigm.
In the first stage, we presented human participants with grayscale images of everyday actions performed in naturalistic contexts at variable presentation times (33.3 – 500 ms), followed by a mask (500 ms). Participants (N = 30) were instructed to provide detailed descriptions of what they perceived in the image. In the second stage, a group of naïve raters (N = 10) evaluated whether these descriptions referred to the action, scene, objects, or to sensory information, and if so, whether these descriptions were correct. In order to reveal the temporal gradient in which these four components are recognized, thresholds for presentation times were calculated and compared, using probabilities for correct descriptions as the dependent variable.
We estimated lower recognition thresholds for actions than for scenes and objects, indicating that actions can be perceived faster than objects and scenes. Additionally, we found that objects were recognized faster than scenes. Considering prior work showing effects of scene information on action- and object recognition, our results suggest that scene, object, and action information might be processed in parallel.