Billions of animals undertake migratory journeys every year, with powerful consequences for ecosystem dynamics. Key behaviours that enable successful migration are often guided by the visual system. The amount and quality of information that animals can extract from visual scenes are directly related to structural eye size — larger eyes can house larger pupils, enhancing light-gathering capacity and vision by improving visual acuity and contrast sensitivity. Migration should exert strong demands on individual visual performance, for example via foraging, anti-predator benefits or navigational requirements. Yet, it remains elusive whether variations in eye morphology and corresponding visual capabilities are associated with migratory propensity. Here, we capitalize upon intra-population variation in migratory propensity (a.k.a. partial migration) in roach, a common freshwater fish, to directly test for migration-associated variation in image-forming eyes within a species. In a multi-year field study tracking the migration decisions of over 2000 individuals in two different lake systems, we found that relative pupil size was positively associated with individual migration propensity. Computational simulations of the visual ecology associated with the observed differences in pupil size show that migrants have an extended visual detection range, and that the performance gain is most pronounced for viewing small targets (e.g. planktonic prey) under low-light conditions. These results suggest that the larger pupils of migrants represent an adaptation for increased foraging efficiency to aid in the accumulation of critical pre-migratory energy reserves. Together, our anatomical and functional findings provide new perspectives on visual system design in relation to individual-level migratory decision-making. 

Data for this study consists of measurements of relative eye and pupil size extracted from digital photographs, as well as migration propensities of individual roach (Rutilus rutilus) migrating between lakes and connected streams. Data was collected for multiple years and from two different lakes (Loldrup and Søgård Sø) using passive biotelemetry with stationary and continuously operating antenna arrays.