Computational modeling uncovers a dynamic interaction between feature uncertainty and perception-action mapping...

Previous studies have shown that visual feature estimation is influenced by internal (sensory) and external (physical) noise, as well as the perception-action mapping process - the transformation of internal perceptual representations into motor responses. However, their interactions remain poorly understood. To investigate this, we conducted two experiments. In Experiment 1, participants estimated self-motion direction (heading) from 3D dot-cloud optic flow, with or without a preceding color-discrimination task to manipulate internal noise magnitudes. In Experiment 2, we introduced randomly moving noise dots (0% or 40% replacement) to vary internal and external noise concurrently. We employed a mixed experimental design that included both between- and within-subject comparisons across participant groups. Results showed systematic variations in estimation errors and response variability across conditions. We developed computational models with one {c}, two { $c$ , ${\sigma }_E$ }, or three free parameters { $c$ , ${\sigma }_E$ , $a$ }, where these parameters represented internal noise magnitude, external noise, and perception-action mapping scaling, respectively. Model comparisons revealed that the color-discrimination task increased internal noise { $c$ }, whereas noise dots elevated both { $c$ } and { ${\sigma }_E$ }. A negative correlation between noise magnitudes { $c$ , ${\sigma }_E$ } and scaling magnitude { $a$ } suggested that noise effects counteract perception-action mapping. In summary, these results provide evidence for a dynamic interaction mechanism in visual feature estimation, whereby increased sensory noise (either internal or external) is counterbalanced by adaptive scaling of perception-action transformation. Our findings highlight the interaction of sensory noise and sensorimotor mapping in perception, and our model serves as a tool for dissecting sensory uncertainty and action scaling across different perceptual domains.