Interactions between active sensing and internal models in natural behavior

Animals actively actuate their sensory organs to explore their environment. The sensorimotor information is processed and integrated to construct internal models that generate predictions and guide future behavior. The electrosensory system of fish offers exceptional access to study the interactions between active sensing and the construction of internal models. Past studies of the initial stages of electrosensory processing showed that motor-related signals are used to construct an internal ‘forward model' that predicts and cancels out self-generated sensory information (termed ‘re-afference'). Paradoxically, these studies have been performed exclusively in immobilized fish, raising the question of whether and how these circuits function in natural behavior. To address this, I developed methods for long-term behavioral and neuronal recording in freely swimming fish. We found that the sensory consequences of the fish’s movements are strongly modulated by features of the external environment. This context dependence implies that the task of re-afference cancellation is far more complex than previously appreciated. Modeling and analysis suggest that a fast feedback circuit provides the internal model with contextual information required for reafference cancelation under natural conditions. Indeed, I found a class of re-afference invariant neurons in the fish’s brainstem responding exclusively to external objects. I will discuss future directions related to how the fish uses sensorimotor information to construct an internal model of the environment and locate itself within it, and how this internal model guides future behavior.