Francesca Peveri¹, Andrea Canessa¹, Silvio Paolo Sabatini¹; ¹University of Genoa

Traditional psychophysics treats participants as relatively passive observers: stimuli are generated by the experimenter, and responses are measured afterward. Recent continuous psychophysics approaches accelerate data collection and enable dynamic response analyses, yet they still consider screen-fixed coordinates, and position the observer in a largely reactive relationship to the stimulus. By analogy with fluid mechanics, current methods follow an Eulerian approach by defining visual input at fixed spatial locations and recording how observers respond to changes at those locations. Here, we propose a methodological shift: instead of limiting participant to generating responses, we engage them as active agents who explore and manipulate 3D visual stimuli in a closed perception-action loop. In this framework, sensory properties change contingently with the observer’s actions, allowing us to characterize how exploration dynamics modulates sensory encoding and perceptual sensitivity. The key innovation is to follow a specific visual “parcel” as it evolves over time and to evaluate perceptual transformations along its trajectory. This treats the moving object, or feature cluster, as the reference frame of the experiment, corresponding to a genuinely Lagrangian description of visual input. We developed a combined experimental and computational framework to instantiate this paradigm. A virtual object was implemented in Unity3D and placed in peripersonal space. Its time-varying 3D pose was yoked to that of a real handheld display operated by the observer. Thus, the display serves simultaneously as the object’s physical manipulator and as the aperture through which the object’s appearance is sampled according to the contingent vantage point of the observer. We show that identical stimulus properties can produce different sensory statistics depending on the observer’s actions, emphasizing the need for perceptual measures that surpass existing continuous paradigms. Incorporating sensorimotor contingencies, our method offers a quantitative framework for perception in true naturalistic action loops.