Motor control - Perception and Dynamic Control

Perception in Motor Control Introduction When you reach for a cup of coffee, catch a falling ball, or navigate around obstacles while walking, your motor system must coordinate with your perceptual systems to produce smooth, accurate movements. But how does perception actually guide action? This is the central question of perception in motor control. Different theoretical approaches offer competing answers: some emphasize the role of internal models that predict and plan movements, while others argue that perception directly specifies the actions needed. Understanding these approaches is essential for grasping how the brain solves the fundamental problem of turning sensory information into coordinated behavior. Model-Based Control Strategies Model-based control relies on the idea that your nervous system constructs and uses internal representations of the world to plan and execute actions. These internal models are built from sensory information and prior knowledge about how the environment works. The key insight here is that perceptual information is rarely perfect. It may be incomplete (you can't see behind a wall), ambiguous (is that distant object small and close or large and far?), or distorted (your own movements change what you perceive). Rather than treating perception as a direct window into reality, model-based approaches assume that your motor system must interpret and reconstruct the world from incomplete sensory signals. This framework explains many puzzling aspects of motor control. When you throw a ball, your nervous system doesn't simply react to current visual information—it predicts where the ball will be, calculates the arm movements needed, and accounts for factors like gravity and air resistance. All of this happens using internal models. Forward Models A forward model is an internal simulation that predicts what sensory consequences will result from a given motor command. Think of it as your brain asking: "If I move my arm this way, what will I see and feel?" Forward models work by taking your current motor program (the command to move) and available perceptual information (your body's position, what you see) and predicting the sensory feedback that should result. When your actual sensory feedback differs from your prediction, a prediction error is generated. This error signal is crucial—it drives learning and causes your internal models to update and improve. A compelling example illustrates why forward models matter: Why can't you tickle yourself? When you try to tickle your own foot, it doesn't feel ticklish because your forward model predicts exactly what sensory input will occur when your hand moves. Your brain receives the predicted, self-generated sensation and, crucially, subtracts it from what you actually perceive. The result is that the sensation seems weak or absent. In contrast, when someone else tickles you unexpectedly, your brain has no forward model to subtract from perception, so the tickle sensation is vivid and often involuntary laughter results. This demonstrates that our perception of our own actions is fundamentally different from perception of external events—forward models make our self-generated sensations seem less intense. Inverse Models While a forward model predicts sensory consequences from motor commands, an inverse model works in the opposite direction: it calculates which motor commands are needed to achieve a desired perceptual outcome. If forward models answer "what will happen if I do this?", inverse models answer "what do I need to do to achieve that?" Inverse models are particularly important for open-loop actions—movements that cannot rely on continuous visual feedback. A classic example is stabilizing your gaze while your head moves. When you turn your head to look left, your eyes must simultaneously move right to keep your gaze fixed on a target. This happens too quickly to rely on feedback ("Did I move my eyes enough?"), so your motor system must use an inverse model to calculate the exact eye movement needed before head movement begins. Together, forward and inverse models create a powerful learning system. The process works like this: You form an intention (what you want to achieve), use an inverse model to generate motor commands, execute those commands, observe the sensory consequences through your forward model, detect any discrepancy between predicted and actual outcomes, and use this prediction error to refine both models. This cycle of prediction → execution → error detection → refinement is the essence of motor learning. Each time you practice a skill, your models become more accurate and your movements more skilled. Information-Based Control and Direct Perception Not all theorists believe internal models are necessary for motor control. An alternative approach called information-based control proposes that actions can be organized directly from perceptual information about the environment, without requiring complex internal cognitive models. In this view, the organism and environment form a coupled system. Rather than your brain building an internal model of the world and then planning actions, behavior emerges directly from the interaction between your perceptual systems and the environment. This is fundamentally different from the model-based view—it's less like "think, then act" and more like "perception directly guides action." Direct perception takes this idea further. According to ecological perception theory, the visual information available in your environment directly specifies the physical properties of objects and what actions are possible. You don't need to consciously interpret visual patterns or construct mental models; the information is already structured in a way that specifies action possibilities. This leads to the concept of affordances: an affordance is an action possibility that exists in the relationship between an organism and its environment. A chair affords sitting-on, a doorway affords passing-through, a sloped ramp affords climbing. Importantly, affordances are not properties of objects alone (a chair is just shaped wood without an organism to sit on it) nor are they purely in the observer's mind (not every organism can climb every ramp). Affordances are relational—they specify possibilities for this organism in this environment right now. The perceptual system is tuned to detect these affordances directly, allowing the motor system to respond appropriately without conscious interpretation or internal modeling. <extrainfo> This contrast between model-based and information-based approaches represents a genuine theoretical divide in motor control. Model-based approaches emphasize internal computation and representation, treating the nervous system as solving an engineering problem. Information-based approaches emphasize the perception-action coupling and treat behavior as an emergent property of organism-environment interaction. Both approaches have evidence supporting them, and modern views often integrate elements of both. </extrainfo> Behavioral Dynamics: Perception and Action as System Properties A complementary framework called behavioral dynamics offers another way to understand how perception guides action. Rather than asking "does the motor system use models or direct perception?" it asks "how do organisms respond to informational variables?" The core assumption is that you can understand behavior by treating the organism as a dynamic system—a system that changes over time according to principles similar to those in physics. Your behavior responds functionally to informational variables in your environment, and your body characteristics (like limb length and mass) shape how you respond. In this view, actions unfold as the natural consequence of your interaction with environmental information that is expressed in body-relevant variables. You don't need to conscious compute anything; the dynamics of your body coupled to your perceptual system naturally produce coordinated behavior. Visual information in locomotion provides clear examples. When walking or running, you use several sources of visual information to control your movement: Optic flow: the pattern of motion in your visual field. As you move forward, the visual world expands around a central point ahead of you. The rate of this expansion tells you how fast you're moving. Time-to-contact: information about how long until you'll reach or collide with something. This is extracted from the rate at which objects expand in your visual field. Optical expansion: how quickly objects grow larger in your visual field, which directly indicates approach speed and time to collision. These variables are not abstract or symbolic—they're directly available in visual perception and can be measured physically. Your motor system is tuned to respond to them, adjusting your speed, direction, and gait accordingly. When you see something expanding rapidly in your visual field, you naturally slow down or stop without consciously calculating time-to-collision. <extrainfo> The behavioral dynamics approach is sometimes called the "dynamical systems" or "action-based" approach. It's particularly successful at explaining phenomena like how humans adjust their walking speed when obstacles appear or how infants learn to crawl. The key insight is that behavior can be understood as the trajectory of a dynamic system rather than the output of a computational process. </extrainfo> The motor system also responds to forces exchanged between your body and the environment. Neural control of limb stiffness—the resistance of your limbs to being moved—exemplifies this. When you hold a pen to write, your nervous system doesn't rigidly lock your arm in place. Instead, it sets an appropriate level of stiffness that allows fine control while maintaining stability. This stiffness adjusts based on the task demands and feedback from the environment. A pushing task might require high stiffness, while writing requires low stiffness. These adjustments happen naturally as part of the coupled organism-environment system. Summary Perception guides motor control through multiple mechanisms. Internal models (both forward and inverse) allow prediction and learning, organizing actions based on internal representations. Alternatively, information-based and ecological approaches emphasize how perceptual information directly specifies affordances and action possibilities. Behavioral dynamics explains how perception and action emerge from the natural properties of organism-environment coupled systems. Together, these frameworks show that motor control is not a single process but rather an elegant coordination between multiple systems that have evolved to couple perception and action.