Opposing influences of sensory and response history in larval zebrafish
Variability is a prominent feature of animal behavior across species. Despite highly controlled experimental conditions, the same individual often responds differently to repeated identical stimulation. Part of this variability can be explained by the sequence of previous sensory stimuli and decision-making events - the trial history. Most studies about trial history are limited to animals that have a cortex, such as rodents or primates, observed as they perform learned cognitive tasks. It is currently unknown whether trial history can shape the behavior of animals lacking cortical structures during untrained, naturalistic behaviors. Here, we address this question in larval zebrafish, performing the optomotor response, an innate sensorimotor behavior in which animals turn in the direction of whole-field visual motion. We observed that a substantial proportion of variability can be explained by recent stimulation and decision-making events, with influences lasting for tens of minutes. Intriguingly, sensory and response histories bias the current response in opposite directions - repulsive and attractive, respectively - analogous to what has been previously reported in humans and rodents. An integrator model operating across multiple timescales explains a considerable fraction of response variability based on trial history alone. Our findings demonstrate that history dependency in animal behavior is not an exclusive feature of higher-order cortical computation after learning, but that it reflects a fundamental and evolutionarily shared property of vertebrate sensorimotor systems. SIGNIFICANCE STATEMENTBehavioral variability is often attributed to noise in sensory and neural processing. We show instead that much of this variability can arise predictably from an animals recent sensory and response history. In freely swimming larval zebrafish performing an innate visual behavior, previous stimuli and responses exert opposing, long-lasting influences: sensory history biases behavior away from the past, whereas response history promotes repetition. Such trial-to-trial effects are increasingly recognized in cognitive neuroscience but, to our knowledge, have not previously been demonstrated in a nonmammalian vertebrate. Their presence in zebrafish, a model system offering brain-wide neural access and powerful molecular tools, creates new opportunities to identify potentially conserved circuit mechanisms underlying persistent sensorimotor computations. HIGHLIGHTSO_LIVisual motion and response histories have opposing influences on behavior C_LIO_LIStimulus history negatively biases response direction over tens of seconds C_LIO_LIAnimals tend to repeat responses across trials, a positive bias that slowly increases over minutes C_LIO_LIA deterministic model applied to experimental sessions explains considerable fractions of response variability C_LI