Material perception relies on context-dependent active sensing strategies

Abstract

Material perception is typically studied under passive and highly constrained viewing conditions, thus leaving it unclear whether humans rely on active sampling strategies to resolve perceptual ambiguities. In everyday vision, however, observers naturally move their heads and manipulate objects with their hands to obtain informative cues, thus raising the question of how such exploratory actions contribute to material recognition. We used immersive virtual reality to examine whether viewpoint changes and object manipulations support the discrimination of visually challenging materials, specifically metal versus glass, whose appearances depend strongly on the associated illumination and viewing geometry. On the basis of three experiments, we revealed that observers systematically increased their exploration tendencies when the identity of a material was ambiguous and that this increased movement was associated with higher discrimination accuracy. By independently manipulating head- and hand-based motions, we identified the context-dependent contributions of each modality, revealing that viewpoint changes dominated when multiple objects could be compared simultaneously, whereas object manipulation was more effective when only a single target was available. Moreover, the participants differed in terms of the efficiency of their sampling strategies, and those who employed more informative exploration patterns exhibited both higher accuracy and greater performance gains across different trials. These results demonstrate that material recognition depends on flexible, context-dependent active sensing rather than passive evaluations of static images. Our findings introduce a new framework, according to which material perception emerges from an adaptive perception-action loop, thus highlighting the functional role of exploratory behaviors in complex visual environments. Significance StatementThe task of material perception in natural environments depends not only on the optical properties of surfaces but also on the actions that observers take to reveal informative visual cues. However, most laboratory studies rely on passive viewing, and how humans actively regulate their own movements to reduce perceptual uncertainty remains unknown. On the basis of three virtual-reality experiments, we showed that observers flexibly perform head- and hand-based explorations to enhance the discriminability of visually ambiguous materials. These exploratory strategies were strongly context-dependent and exhibited systematic individual differences that predicted learning-related improvements. Our findings demonstrate that material perception emerges from an adaptive perception-action loop rather than static image analyses, providing a framework for understanding how humans actively acquire diagnostic sensory information in complex environments.