The Rubber Hand Illusion: Why Neuroscience Explains the "How" but Not the "Why"

Introduction: A Phenomenon That Became a Testing Ground for Theories of Consciousness

In 1998, Matthew Botvinick and Jonathan Cohen described a deceptively simple experiment: a participant's real hand is hidden from view, a rubber hand is placed in front of them, and both are stroked synchronously with a paintbrush. Within one to two minutes, most participants report experiencing the rubber hand as their own. Since then, this phenomenon has become one of the most cited findings in the science of consciousness and embodiment, precisely because it reveals, as if under a magnifying glass, that the body is not a given fact but a continuously constructed inference of the brain.

Over the past quarter-century, several influential explanatory frameworks have accumulated—from straightforward "multisensory integration" accounts to rigorous Bayesian models of causal inference. Each of these frameworks accurately captures correlations between stimuli and brain activity. Yet none, in my view, answers the central question: why does the brain need to make such a decision at all, and what determines which elements are drawn within the boundary of "self" and which remain "other"? This is precisely the space in which Pyotr Anokhin's theory of functional systems becomes relevant—a framework developed entirely outside the context of this illusion, yet unexpectedly precise in capturing its underlying logic.

What the Existing Models Explain

There are three principal explanatory layers, and it is worth distinguishing them clearly before addressing their limitations.

Multisensory integration. The earliest and still most widely cited account: the brain combines visual, tactile, and proprioceptive signals into a single body model, and under synchronous stimulation, the visual and tactile channels "fuse," pulling the felt location of the hand toward the rubber hand. Henrik Ehrsson demonstrated experimentally that this process is mediated by bimodal neurons in the ventral premotor and intraparietal cortices, which encode the space immediately surrounding the body and integrate signals across modalities into a unified map.

Bayesian causal inference. A more formal layer developed over the past decade: the brain estimates the probability that tactile and visual signals share a common source, and the greater the uncertainty of the proprioceptive signal, the stronger the illusion. This explains why the illusion intensifies in darkness or under divided attention—the uncertainty of the "native" signal increases, making it easier to override with the more confident visual one.

Predictive coding / the free-energy principle. The most abstract layer: the brain continuously generates predictions about the body and minimizes the discrepancy—the prediction error—between these predictions and incoming signals. A rubber hand stroked in synchrony produces minimal prediction error under the assumption that it is recognized as one's own, and the system statistically "settles" on this assumption because it minimizes overall surprise.

Each of these three layers is well supported empirically. Yet they share a common blind spot.

Where the Explanatory Power of These Models Runs Out

All three approaches describe a mechanism of signal reconciliation, but none explains why the outcome of this reconciliation is specifically a subjective feeling of ownership—the sense that "this is my hand"—rather than merely a more accurate estimate of an object's spatial location. Bayesian inference successfully predicts proprioceptive drift (the shift in the felt position of the hand), yet experimental evidence shows that drift and subjective ownership dissociate: one can obtain strong drift without any accompanying sense of ownership, and vice versa. Formal models of signal correlation do not predict this dissociation—they merely register it after the fact.

A second gap: these models are static and passive. They describe the brain as a Bayesian calculator weighing probabilities, but they do not explain why the organism needs a boundary between "mine" and "not mine" in the first place. The answer to "why" cannot be found in signal statistics alone—it lies in what the body is actually for, which is a question of action, not merely perception.

A third gap is empirical: studies using transcranial magnetic stimulation show that during the illusion, the excitability of motor circuits controlling the real hand decreases—indicating that the brain does not simply "confuse" sensory maps, but temporarily withdraws the real hand from the circuit of action-readiness. Neither multisensory integration nor Bayesian inference predicts this effect directly; it emerges as a side observation that fits awkwardly within purely perceptual models.

A Perspective Through Anokhin's Theory of Functional Systems

Between the 1930s and 1970s, Pyotr Anokhin proposed an alternative to the reflex arc: organismic behavior, he argued, is not built from discrete reflexes but from functional systems—dynamic, self-organizing assemblies of central and peripheral structures that form anew for each specific task and dissolve once that task is resolved. The key components of such a system are afferent synthesis (the integration of all relevant information about the current situation, including memory and motivation), decision-making, the acceptor of the results of action (a neural model predicting what the outcome of an action should be), and reverse afferentation (the actual feedback signal about what occurred, compared against that predictive model).

Applying this framework to the rubber hand illusion fundamentally reframes the picture. The brain is not abstractly "integrating signals"—it is constructing a functional system oriented toward the task of maintaining an effector apparatus ready for action. The hand is not merely an object of perception but a potential instrument of movement: grasping, defending, orienting. The acceptor of the results of action continuously maintains a model of "what should happen if I use this hand as an effector." When the rubber hand responds synchronously to touch, it begins to satisfy this predictive model just as well as—sometimes better than—the real one, precisely because it is visible, and a visible effector is more readily incorporated into a planned action circuit than a hidden one.

From this standpoint, ownership is not a byproduct of sensory map fusion but a decision made by the functional system about which object is sufficiently reliable to be incorporated into the action circuit at this particular moment. This is precisely why motor circuit excitability for the real hand decreases: the system is literally reassigning effector status from one object to another, rather than simply confusing spatial coordinates. Bayesian inference describes how the brain weighs the probability of competing signal sources, but functional systems theory explains why this choice is made at all: the body, as such, is not of interest to the brain in itself, but as an instrument for achieving an adaptively useful outcome, and the boundary of "mine" is drawn precisely where the zone of reliable, predictable control ends.

A Tempting but Misleading Parallel

At first glance, one might be tempted to draw a parallel with phantom limb pain following amputation—after all, in that condition too, the brain seems to "sustain" the image of a limb that no longer exists. On closer inspection, however, functional systems theory reveals these to be two fundamentally different situations, and conflating them would be a mistake.

In the rubber hand illusion, both candidates are physically present: the real hand, generating a proprioceptive signal, and the rubber hand, generating a congruent visuo-tactile signal. The functional system is not filling an absence—it is choosing between two genuinely existing candidates, determining which one holds greater effector status in the moment. This is a competition, not a compensation for a missing element.

Phantom limb pain following amputation is an entirely different case: there, no real candidate for effector status exists at all, yet the acceptor of the results of action continues, by inertia, to generate predictions about a limb that no longer exists, while the reverse afferentation that should confirm or correct this prediction never arrives. The system cannot complete its cycle—and it is precisely this incompletion, not confusion between two objects, that generates the pain. In other words, where the rubber hand illusion involves the system resolving the question of "which of the two," phantom limb pain involves the system becoming trapped in the question of "where is the one that no longer exists at all." The surface-level similarity—"the brain constructs the image of a hand that is not physically present within the field of perception"—obscures the fact that the underlying causal structures of these two processes are, in fact, opposite.

Conclusion

Existing explanations of the rubber hand illusion—from multisensory integration to predictive coding—accurately describe the statistics and neuroanatomy of the process, but remain at the level of "how the computation proceeds" without addressing why the body requires such a boundary at all. Anokhin's theory of functional systems restores what has been lost in purely correlational models: a teleological, action-oriented dimension. The body is not simply a model that the brain fits to the most probable sensory data; it is a dynamically assembled effector circuit, the boundaries of which are determined not by what appears plausible, but by what proves functionally viable for achieving a given outcome in the moment. This is precisely why the illusion collapses within seconds once synchrony is broken: what dissolves is not a "false hypothesis" but a functional system that has ceased to receive confirmation of its own utility.

You can learn more by reading our e-book 

Mykola Iabluchanskyi together with Andriy Yabluchanskiy