Gaze mechanisms enabling the detection of faint stars in the night sky

For millennia, people have used “averted vision” to improve their detection of faint celestial objects, a technique first documented around 325 BCE. Yet, no studies have assessed gaze location during averted vision to determine what pattern best facilitates perception. Here, we characterized averted vision while recording eye‐positions of dark‐adapted human participants, for the first time. We simulated stars of apparent magnitudes 3.3 and 3.5, matching their brightness to Megrez (the dimmest star in the Big Dipper) and Tau Ceti. Participants indicated whether each star was visible from a series of fixation locations, providing a comprehensive map of detection performance in all directions. Contrary to prior predictions, maximum detection was first achieved at ~8° from the star, much closer to the fovea than expected from rod‐cone distributions alone. These findings challenge the assumption of optimal detection at the rod density peak and provide the first systematic assessment of an age‐old facet of human vision. For over 2,000 years, stargazers have used “averted vision” to improve their detection of celestial objects. We tracked eye position during averted vision to determine the gaze patterns that produce optimal star detection, for the first time. Detectability was highest at ~8–14° from the star, contrary to prior predictions that peak star detection should overlap with the peak of retinal rod density.