Vocal communication as cooperative sensing for navigation

Affiliative vocalizations are used by humans and other animals to facilitate social cohesion and help separated individuals reunite. In both the laboratory and field studies, changes in acoustic features of affiliative vocalizations exhibit a consistent relationship with physical distance from conspecifics: longer distances induce louder calls. Thus, vocalizations are likely to be driven by low-dimensional latent states associated with distance. Here, we propose a dynamical model for the use of affiliative vocalizations as a means of active sensing to guide navigation. In this model, vocalization and locomotion are inversely driven by the uncertainty of the locations of communicating conspecifics, and the dynamics of actions are regulated by energy constraints. Using data from captive animals, we show that the context-dependent vocalizations are a natural consequence of such an active sensing system. Optimal tuning for the probability of vocalization exists as measured by the rate of spatial convergence. This study predicts the potential mechanisms for the coordination between vocal production and the navigation system.