Efficient mixed representation of active and passive motion in the mouse visual thalamus during natural behaviour

Abstract

Most retinal motion is generated by the observer's movements. These are often self-generated (active) but can also arise from external forces such as gravity (passive). Active movements modulate spontaneous and visually evoked activity throughout the visual system, providing neural circuits with information about behavioural state. By contrast, much less is known about how visual circuits represent passive movements. Critically, it remains unclear whether the retina-recipient thalamus tracks passive motion and, if so, whether this representation is separable from active motion. To address these questions, we developed a paradigm that dissociates active and passive motion in freely moving mice and used it to examine their effects on neural activity. Mice actively explored an arena while intermittent floor tilts induced passive movement. By combining 3D reconstructions of animal pose and arena motion, we quantified active and passive linear and angular motion independently, even when they occurred simultaneously. We found that neurons in the dorsal lateral geniculate nucleus (dLGN) and surrounding regions tracked combinations of active and passive motion signals both in the dark and during visual stimulation. Individual neurons differentially weighted active and passive motion signals. This diversity enabled active and passive motion signals to be linearly decoded and partially separated from population activity, while recurrent dynamics retained passive motion information over timescales of several seconds. These findings identify a coding strategy through which early visual circuits integrate active and passive motion into an estimate of self-motion while preserving information about whether visual inputs arise from self-generated or externally imposed movements.