Motor prediction is a key concept in sensorimotor control, where the central nervous system (CNS) predicts the consequences of motor commands to guide behavior. This concept was first proposed by Helmholtz, who suggested that the brain uses an "effection copy" of motor commands to predict eye position rather than directly sensing it. Experimental work by Von Holst and Sperry in the 1950s confirmed this idea. Motor prediction involves estimating future states of systems directly influenced by motor commands, such as arm movement or car control. The CNS uses forward models to simulate body and environmental dynamics, allowing for accurate predictions. These models are not fixed but are learned and updated through experience, using prediction errors to adjust synaptic weights. State estimation is crucial for motor control, as it involves predicting body states based on motor commands. This helps overcome sensory delays and noise, and can be combined with sensory feedback for better accuracy. Prediction also allows filtering of sensory information, helping to cancel out movement-induced sensory effects. This is important for distinguishing self-generated from externally generated movements, a failure of which may underlie delusions of control in schizophrenia. Context estimation involves using multiple forward models to predict sensory feedback based on movement context, selecting the most appropriate controller. This is exemplified by the MOSAIC model, which uses multiple forward models to predict outcomes for different contexts. Prediction is also fundamental for higher cognitive functions such as mental practice, imitation, and social cognition. Forward models help predict sensory outcomes of actions, aiding in action observation and understanding. These models may have evolved from sensorimotor prediction to support other cognitive functions. This review highlights the importance of motor prediction in sensorimotor control and its broader implications for cognitive functions.Motor prediction is a key concept in sensorimotor control, where the central nervous system (CNS) predicts the consequences of motor commands to guide behavior. This concept was first proposed by Helmholtz, who suggested that the brain uses an "effection copy" of motor commands to predict eye position rather than directly sensing it. Experimental work by Von Holst and Sperry in the 1950s confirmed this idea. Motor prediction involves estimating future states of systems directly influenced by motor commands, such as arm movement or car control. The CNS uses forward models to simulate body and environmental dynamics, allowing for accurate predictions. These models are not fixed but are learned and updated through experience, using prediction errors to adjust synaptic weights. State estimation is crucial for motor control, as it involves predicting body states based on motor commands. This helps overcome sensory delays and noise, and can be combined with sensory feedback for better accuracy. Prediction also allows filtering of sensory information, helping to cancel out movement-induced sensory effects. This is important for distinguishing self-generated from externally generated movements, a failure of which may underlie delusions of control in schizophrenia. Context estimation involves using multiple forward models to predict sensory feedback based on movement context, selecting the most appropriate controller. This is exemplified by the MOSAIC model, which uses multiple forward models to predict outcomes for different contexts. Prediction is also fundamental for higher cognitive functions such as mental practice, imitation, and social cognition. Forward models help predict sensory outcomes of actions, aiding in action observation and understanding. These models may have evolved from sensorimotor prediction to support other cognitive functions. This review highlights the importance of motor prediction in sensorimotor control and its broader implications for cognitive functions.