Every time you move, speak, or even blink, your sensory organs are flooded with information. When you walk, your skin feels the friction of clothes; when you speak, your ears hear a voice; when you blink, your vision goes dark. Yet, you rarely perceive these events as external disruptions. You don't think a stranger is talking when you open your mouth, nor do you think the lights have failed when you blink.
This seamless experience is the result of sensory attenuation, a predictive filtering mechanism that allows the brain to distinguish between "self" and "world." By anticipating the sensory consequences of its own actions, the nervous system can subtract predictable "noise" and prioritize novel information from the environment.
The Problem of Reafference
In sensory biology, the signals resulting from an animal's own movements are called reafference. As noted by Daniel C McNamee in Internal Models in Biological Control, a fundamental challenge for any biological agent is distinguishing these self-generated signals from exafference—sensory changes caused by external events.
Without a way to filter reafference, the brain would be overwhelmed. A rat whisking to navigate would be unable to tell the difference between the motion of its own muscles and the touch of a wall. As Chunxiu Yu and colleagues demonstrated in Parallel Thalamic Pathways for Whisking and Touch Signals in the Rat, the brain solves this by creating specialized pathways that separate pure motion signals ("I am moving") from contact signals ("I hit something").
The Efference Copy and Forward Models
The mechanism behind this filter is a "heads-up" signal known as an efference copy (or corollary discharge). When the motor cortex sends a command to a muscle, it simultaneously sends a duplicate of that command to sensory processing areas.
This copy is used by an internal forward model—a neural simulation that predicts what the resulting sensory feedback should feel like. Johannes Schröder-Schetelig, Poramate Manoonpong, and Florentin Wörgötter applied this concept to robotics in Using efference copy and a forward internal model for adaptive biped walking, showing that by subtracting predicted self-generated acceleration from raw sensor data, a robot can isolate external forces, such as a change in slope, even without "seeing" it.
Neural Cancellation and Prediction Error
In the human brain, this prediction manifests as a physical dampening of neural activity. In The Sensory Consequences of Speaking: Parametric Neural Cancellation during Speech in Auditory Cortex, Ingrid K. Christoffels describes this as neural cancellation. When your actual auditory feedback matches the internal prediction of your voice, the auditory cortex suppresses its response.
This suppression is not an "all-or-nothing" switch but a graded, parametric process. If you hear your voice through headphones and the pitch is subtly shifted, the mismatch between the prediction and the feedback generates a prediction error. This error signal alerts the brain that something is wrong, triggering rapid motor corrections to bring the output back in line with the goal.
The Sense of Agency
This predictive loop is the physiological foundation for the sense of agency—the subjective feeling that "I" am the author of an action. Ke Zhao and Xiaolan Fu argue in The spatiotemporal markers for the sense of agency in the human brain that the brain uses the alignment between intentions and outcomes to tag events as self-caused.
When this alignment is high, sensory attenuation is strong. Conversely, as Wen Wen explores in Categorical Perception of Control, our brains are hyper-sensitive to the boundaries where we lose control. If a delay is introduced between a movement and its visual effect, the sense of agency collapses, and the brain stops attenuating the sensory signal, treating it instead as an external event.
When Predictions Fail: Clinical Insights
Understanding these internal models provides a window into neurodegenerative and motor disorders. In Alzheimer’s Disease (AD), the brain's ability to predict its own actions appears to break down.
Research by Kamalini G. Ranasinghe and Kyunghee X. Kim in Impaired auditory feedback prediction in Alzheimer’s disease shows that AD patients fail to suppress their auditory cortex during speech. Because their internal forward models are noisy or inaccurate, they perceive their own voices as external errors. This leads to "behavioral over-correction," where the patient constantly adjusts their speech in response to perceived mismatches that aren't actually there, resulting in vocal instability.
Similarly, André Lee and Eckart Altenmüller found that focal hand dystonia—a condition where musicians lose voluntary control of their fingers—is linked to a failure in predicting the sensory consequences of touch (Alteration in forward model prediction of sensory outcome of motor action in focal hand dystonia).
Conclusion
Sensory attenuation is more than just a way to ignore the sound of our own breathing. It is a sophisticated, predictive architecture that allows embodied agents—whether biological or artificial—to maintain a stable representation of the self while navigating a noisy, changing world. By continuously comparing what we do with what we feel, the brain creates the coherent experience of being an active participant in the environment rather than a passive observer of sensory chaos.