Testing a model of the cognitive structure of perceivable human actions

Several models have been proposed to explain the cognitive structure underpinning our perception of human actions. These studies have conceptualized this structure as a domain specific ‘action space’ defined by a limited number of important dimensions that describe the meaningful information on which we evaluate actions. However, these different models have disagreed on both the number of dimensions, and what action qualities each dimension represents. To help resolve these differences we retested Vinton et al.’s (2023) 4-D action space model whilst considering additional qualities identified by other models. We combined Vinton’s original dataset where participants rated 240 different controlled actions against 23 action characteristics, with ratings of the same actions against 8 new characteristics derived from other action space model dimensions. We used an EFA-to-CFA approach to determine the optimal model of action space and assess which action qualities define that space. Action space was best modelled by 5-factors. Two substantial dimensions originally identified by Vinton (Friendliness, Formidableness) each explained 18-19% of the variance. Whilst 3 smaller dimensions each explained 10-8% of the variance; these included Abduction (as per Vinton) and Locomotion and Environmental Interaction. We interpret this representation of action space as having 2+3 dimensions. We show that friendliness and formidableness retain their central role as the two most important action qualities when alternative theoretically motivated characteristics are incorporated, consistent with their interpretation as action-specific signals of warmth and competence. Differences between earlier studies of action space can be attributed to non-representative samples of actions or the use of static images of actions.