Fig. 3

Preferences for distinct motor strategies are deployed across sensory contexts. (A) Left, probability density of visiting microstates along . Right, example trajectory color coded by (Top) and median dwell time in each strategy (Bottom) (error bars represent 95% CI bootstrapped across fish, with individual fish scattered in the background). (A1) In large square arena in the light, fish mostly fast cruise ( , , 10 fish). (A2) In a smaller arena, fish in the light use wandering at the short ends when forced to reorient ( , arena, 12 fish), resulting in a reduction of time spent cruising compared to (A1). (A3) In the dark, fish show a shift toward fast wandering ( , 37 fish, data combined from the ?Dark? condition and the first 30 min of the ?Dark Transitions? condition; see SI Appendix, Table S1). (A4) In a prey capture assay with 50 paramecia per arena, fish mostly engage in slow cruising and wandering when the eyes are converged, and introduce faster wandering when the eyes diverge ( , arena, 65 fish). Each bout is given an eye convergence index based on the angle between the eyes (SI Appendix). Note the near absence of fast cruising when compared to (A1). (B) Probability of hunting (B1) or detecting (B2) resources in each of the motor strategies, SI Appendix, Fig. S3D1. Using data from all fish, we infer transition matrices using only the microstates belonging to each motor strategy and simulate trajectories (SI Appendix, Fig. S5). (B1) We assess the fish?s ability to hunt prey that is uniformly distributed within cone of aperture ahead of its head (eyes converged) (40) and radius. We simulate fish trajectories for the average duration of hunting sequences ( ), SI Appendix, Fig. S5A, and count successful hunts when the head trajectory is within from the prey ( 5% of the body length); see SI Appendix. In this case, slow cruising is the most successful strategy. (B2) We assess the fish?s ability to detect resources uniformly scattered within a distance from the initial position (41) We now simulate 1,000 bout-long trajectories, and assume that the fish can detect resources within a distance around its head position (SI Appendix). In this case, the wandering states are most effective up to mesoscale search ( ) while fast cruising is effective at large scale dispersal. The black line represents the behavior of an average fish. Errorbars and shaded areas correspond to bootstrapped 95% CI obtained from 10,000 simulations from random initial conditions.