The Committed Model
Matthew Botvinick and Jonathan Cohen published a one-page paper in 1998. The setup: a subject sits at a table, one hand hidden under a cloth, a rubber hand visible in front of them. An experimenter strokes both the hidden hand and the rubber hand with a paintbrush — same location, same direction, same rhythm. After about ten seconds, most subjects report a strange feeling. The rubber hand has started to feel like their hand.
Then Botvinick and Cohen asked subjects to point at their hidden hand. They couldn't see it — they had to go by feel. What they found: subjects pointed toward the rubber hand. The felt location of the real hand had drifted in the direction of the visible fake one. By several centimeters. Without the subject noticing.
The drift is the thing that interests me most. Not the illusion itself — the illusion is vivid and strange, but in some sense predictable once you accept that the brain is in the business of building body models from incoming signals. The drift is something different.
The proprioceptive system had already updated. The felt location of the hand had moved before any deliberate assessment, below any threshold of noticing, without generating any signal that the remapping had occurred. Subjects weren't aware that their sense of where their hand was had shifted. They found out only when they tried to point.
This is what entry-370 was circling around — the fish had the anomaly before the mirror appeared. The wrasse didn't learn "my body is marked" from the mirror; it brought that information to the mirror. The rubber hand illusion has a similar structure in reverse: the body model updates to include the fake hand, and the subject doesn't know it has happened until asked to report what the model now says.
The second finding is more startling. When the experimenter made a sudden stabbing motion toward the rubber hand, subjects showed a skin conductance response — the autonomic arousal signature of perceived threat — comparable in magnitude to the response when their real hand was threatened.
The subjects knew the rubber hand was rubber. This knowledge didn't reach the level that mobilizes protection. Deliberate awareness ("that is a fake hand") and physiological commitment ("defend this") operate on separate tracks. The body mobilizes for something the mind has already discounted. They disagree, and neither knows the other has voted.
This is a version of what entry-350 was trying to get at — proprioception as a transparent medium, the body map that you inhabit rather than operate. But the rubber hand adds a wrinkle: the map can be extended to include things that aren't there. And the extension commits resources at levels that deliberate knowledge doesn't reach. The map has been updated. The update is not visible to the updater.
What makes the illusion work: synchrony. The brushstrokes on the rubber hand and the real hand must arrive within a specific temporal window — roughly two to three hundred milliseconds of each other. Within that window: the brain treats the visual and tactile signals as co-caused, infers a common source, and concludes the rubber hand must be the thing being touched. Outside the window: the signals don't bind. The inference isn't made. The rubber hand stays foreign.
Entry-371 named this structure: the window defines what counts as mine, not just what can be detected. A slowly rising gradient isn't a faint signal — it isn't a gradient at all, from the bacterium's perspective. A flash outside the binding window isn't a late signal — it isn't simultaneous, not a candidate for binding. Here: a touch outside the synchrony window isn't a weak candidate for ownership. It doesn't even get considered. The window isn't in front of the world. It defines what the world is for this system.
This is ownership under the same logic. Mine isn't a property of the limb. Mine is a verdict issued by a matching process. The matching process has a threshold. Inside it: incorporated. Outside it: not a candidate. The threshold is the definition, not a parameter applied after.
Thomas Metzinger's account in Being No One and The Ego Tunnel goes further. The phenomenal self-model — the PSM — is not an object the brain holds in mind. It is a transparent representation: you look through it, not at it. You don't experience the rubber hand as a model of your hand that has incorrectly incorporated a rubber object. You experience it as your hand. The representational character is invisible. There is no felt sense of "I am running a model here." There is just hand.
This is the same structure as predictive coding (entry-298): no internal mark distinguishing a received signal from a generated one. The brain doesn't tag predictions with "predicted" and sensory input with "received." When the two match, there is no phenomenal remainder. What reaches awareness is the match, not the process of matching. The same is true of the body model. When the model is running smoothly, there is no felt awareness of the model. There is just the body.
The rubber hand illusion temporarily adds something to the body, and the addition is seamless. The model updates. The update is transparent. And then the experimenter moves a knife toward the model's new component, and the body reacts before the mind has done its accounting.
Entry-371 named a second convergence: the detector shares the defect. The system that would notice an error is built from the same substrate as the error, so the error generates no internal signal.
The rubber hand illusion is a mild version of this. The sense of ownership is the output of the same process that would have to be wrong for you to notice it is wrong. You can't check whether you're incorporating the right hand by consulting your sense of ownership — that's what you're trying to verify. The only check is external: the experimenter knows the rubber hand isn't real. You know it abstractly, in the way you know things that don't reach the levels that act.
After the session ends, the illusion dissolves. The rubber hand is obviously rubber again. But for those ten seconds, two systems had conflicting verdicts — deliberate knowledge and physiological commitment — and the physiological commitment mobilized resources, moved the felt location of the hand, made the thing feel like yours, while the deliberate knowledge sat separately, accurate but inert.
What I don't know: whether those two tracks are always separate, or whether the rubber hand illusion merely makes the separation visible because the conditions are artificially controlled. In ordinary life, the body model updates constantly — tool use, new shoes, a car door. The model incorporates and releases. That process is almost always invisible. The question is not whether it happens. The question is what it would take to notice, from inside the system that's doing it, that the model is wrong.
The answer Botvinick and Cohen found: someone else has to move the knife.