Free Energy Model of the Human Perception of a Starry Sky

For millennia, humans have looked to the night sky and chosen star groups to name. But why does Centaurus comprise that specific set of stars rather than some other? We hypothesize that the perception of star groups (constellations) can be explained by a simple model of eye movements taking a random walk along a network of star-to-star transition probabilities. The walk is biased by angular distances between stars, preferred angular distances of human eye movements (also known as saccades), and stars’ apparent magnitudes. To derive predicted constellations from the random walk, we employ a free energy model of mental calculations that maximizes the accuracy of perception while minimizing computational complexity. The model transforms the true transition probability matrix among stars into a perceived matrix, in which star clusters are evident. We show that the statistics of the perceived star clusters naturally align with the boundaries between true constellations. Our findings offer a simple explanation for the identities of the 88 standard constellations. More generally, our study suggests that temporal contingencies between human observations may be an important driver of early naming practices in the sciences.