Using two-photon calcium imaging, we show that visual information alone is sufficient to drive the brain's internal compass—the head-direction network—in head-fixed mice, independent of physical head movement. We characterized the feedforward projections from the anterodorsal thalamus and visual cortex into retrosplenial cortex (RSC), and identified a population of RSC neurons that combine both signals, providing the first direct neural evidence for the integration of visual and heading information. From this we proposed a simple circuit model for how an animal's internal compass locks onto external visual landmarks.

Adapting random dot kinematograms—a stimulus commonly used in primates—for mice, we mapped how coherent motion is represented across the visual cortex. We found heterogeneity in motion responsiveness both across and within higher visual areas, but with far less distinction between areas than in primates, supporting the view that the mouse visual cortex is less modular and more interconnected. We also discovered a visual-elevation-dependent gradient, in which neurons representing the inferior visual field respond more strongly to coherent motion—potentially an optimization driven by natural scene statistics.