Have you ever wondered how a seasoned batter faces a spin ball and also hits it for a boundary? Even as the ball swings, the batter uses their brain’s predictive powers to correctly anticipate the ball’s behaviour. We respond to events around us based on our prior knowledge, patterns, and predictability. If, however, events don’t occur according to the predictions, the brain detects an ‘error’ and readjusts the response.

Researchers from the National Centre for Biological Sciences (NCBS), Bengaluru studied the predictive power of the brain using a non-mammalian model—a zebrafish. In their study in Science Advances, researchers have demonstrated that larval zebrafish can predict outcomes based on their experiences, and if the predictions don’t match what they expect, the zebrafish can also update their behavior.

Larval zebrafish have a small size and transparent body that allow researchers to visualize the brain in an intact animal while it exhibits a range of natural behaviours. This study is focused on the cerebellum, a highly conserved region across all vertebrates. The cerebellum plays a pivotal role in motor coordination, motor learning, and cognitive functions. The present study used 7-days post-fertilization larval zebrafish, in which the cerebellum was functional. Researchers asked how the brain learns predictable patterns in the world and makes inferences to guide actions in future when they encounter familiar situations.

In their experiments, researchers stimulated the larval zebrafish to swim by showing them repeated visual patterns that created an illusion of the fish being in motion. Typically, zebrafish are able to sense the optic flow and swim in the same direction. This is known as the optomotor response, which is an innate response that helps the zebrafish stabilize themselves when there are currents in the water.

In a stream of flowing water, the zebrafish orient themselves against the direction of the current, and they swim upstream. This keeps them in place so they cancel out the flow of the water. The visual sense of things stuck to the bottom or the walls of the stream drives this behaviour.

In this study, while the pattern on the screen moved coherently to produce the illusion, the zebrafishes’ motion was physically controlled by restricting their head using agarose. “The restrained fish could use their tail movements as a joystick to control their position in a virtual world on a screen, just like playing a video game. This kept them engaged for hours while their brains were imaged,” says Sriram Narayanan, a former PhD student at NCBS and the first author of this study.

When the pattern on the screen was consistent, the fish started predicting it and quickly showed an optomotor response. However, when a different stimulus was presented, there was a delay in the optomotor response.To check if the cerebellum played a role in generating the optomotor response, researchers lesioned the cerebellum. They observed that zebrafish with lesioned cerebellums were not affected by a different, unexpected stimulus. They behaved as if there was no change in the stimulus. “So now this feature of waiting and thinking before responding was gone when I lesioned the cerebellum. But the baseline optomotor response was similar,” said Narayanan.

The team also measured the neural response in two cell types in the cerebellum—Purkinje cells (PCs) and granule cells (GCs). The PCs are neurons present in the cerebellar cortex that play a crucial role in sensory and motor coordination. Granule cells feed excitatory inputs into the PCs. The PC activity was studied using calcium imaging in the head-restrained larval zebrafish.

The imaging showed that when stimulated, the PCs encode predictions of expected sensory input—here, optic flow. However, when optic flow was presented in a different direction, the PCs encode an ‘error signal’ to update future predictions. With repeated trials, the response time improves, and the fish predict better to minimise future errors.

Researchers observed that upon consistent repetition, the zebrafish would expect the same stimuli, and would behave based on the expectation of the subsequent stimuli. However upon receiving a deviant stimulus, the PC activity showed a stronger error response. The results show how the zebrafishes’ past experiences shape their expectations of the real world. The error signals can be used to make more accurate predictions in future.

Further, the study, has sparked new questions on the brain’s ability to compute errors and make correct predictions. “Where and how in the fish brain are these predictions generated and how these circuits, both in their organization and function, are conserved across species and cognitive tasks are some exciting questions that are now within reach,” Narayanan says.

This study has implications for understanding zebrafish behaviour in natural settings, e.g., in improving predictions while evading predators or finding food. The findings advance our understanding of motor learning, cognition, and adaptation to changing environments.