Divergent spatiotemporal integration of whole-field visual motion in medaka and zebrafish larvae

Cross-species comparisons offer leverage for identifying conserved and divergent neural computations underlying innate behavior. Visual motion integration is a fundamental operation that stabilizes position relative to the moving environment and supports object tracking, yet how its underlying algorithms vary across closely related vertebrate brains remains poorly understood. We investigated how zebrafish ( Danio rerio ) and medaka ( Oryzias latipes ) larvae implement visual motion integration using distinct spatiotemporal filters that trade speed for persistence through separable control modules. Using controlled whole-field motion stimuli, we found that medaka pool motion signals over visual fields nearly twice as large as those of zebrafish and exhibit enhanced weighting of peripheral inputs, whereas zebrafish rely more strongly on motion signals directly beneath the body. Temporally, zebrafish respond robustly to motion signals with lifetimes as short as 100 ms, whereas medaka require stimulus durations exceeding one second and maintain motion-driven activity for several seconds after stimulus offset. Decomposition of turning behavior revealed separable control modules for large and small corrective maneuvers, with species differences arising primarily from prolonged temporal integration in medaka small-turn control. Together, these differences reveal species-specific tuning of spatial kernels and temporal filters underlying visuomotor control. Our results demonstrate how alterations in basic computational motifs, spatiotemporal pooling, gain, and persistence, can generate divergent visuomotor strategies across closely related vertebrate brains.