A person sits at a table with one hand hidden and a lifelike rubber hand in view. An experimenter strokes both hands with brushes. After a short time, the visible hand may begin to feel strangely connected to the hidden one. A touch seen on rubber seems to coincide with a touch felt on skin. When asked where their own hand is, the person may point closer to the rubber hand than before.
Botvinick and Cohen’s 1998 report made this arrangement a landmark in the study of body ownership. It offered a simple way to ask how vision, touch and information about limb position combine in the experience of “my hand.” The result did not show that the brain is easily fooled in every circumstance, or that ownership is stored in a single location. It showed that a carefully timed conflict among senses can change reported experience and position judgments under particular conditions.
The paper and its experimental question
The original Nature correspondence was brief, just one journal page. Participants placed a real hand out of sight while a rubber hand appeared in a plausible position. The experimenters brushed the real and fake hands in synchrony. Participants described the sensation as if the touch were associated with the rubber hand, and their judgments of the hidden hand’s position shifted toward it. The paper’s strength was a striking, reproducible task question: can correlated visual and tactile signals alter the felt location and ownership of a limb?
The setup matters. A rubber hand shaped like a human hand was positioned where a hand could plausibly be. The real hand was hidden so that direct visual evidence did not settle the conflict. Strokes were timed to match. Without those details, the same report of “it felt like my hand” would be difficult to interpret. It might reflect demand, ordinary imagination or a different visual trick. The design focused attention on the way coherent signals can be bound to a body-like object.
The original report remains a starting point, not a final model of the bodily self. Its brevity left many conditions and measures for later studies to examine. The useful question after 1998 was not merely whether the demonstration could be repeated. It was which combination of timing, position, hand shape, prior expectation and instruction makes the effect more or less likely, and which outcome counts as the effect.
Two outcomes should not be collapsed
One outcome is subjective ownership: how strongly a person reports that the rubber hand feels like their own. Another is proprioceptive drift: a change in the estimated location of the hidden real hand. A participant may show both, one or neither, depending on task and measurement. Calling drift an objective ownership meter skips an important step. It is a location judgment, while ownership is a felt relation to a body part.
Rohde, Di Luca and Ernst directly examined whether the two outcomes move together. Their experiments found that a change in judged hand position could occur without a corresponding feeling of ownership in some conditions. The title states the lesson plainly: ownership and proprioceptive drift do not go hand in hand. This does not make either measure useless. It asks researchers to report them separately and to avoid using one as a proxy for the other without evidence.
Abdulkarim and Ehrsson later tested causal interpretations of drift and ownership. Their work challenged the idea that a change in hand-position sense is necessary to create ownership. A relationship between two measures in the classic setup is not enough to establish that one causes the other. If an intervention changes drift while ownership remains, or changes ownership without the expected drift, a single-path explanation becomes difficult to maintain.
Questionnaires also require care. Participants may infer what a demonstration is meant to evoke; wording can invite agreement; a scale may combine touch referral, ownership and location impressions. Control statements and conditions help, but they do not make subjective reports irrelevant. The experience is part of what the experiment studies. Good measurement respects it while keeping its different aspects visible.
Why timing and plausibility matter
Synchronous stroking supplies evidence that the seen and felt touches come from the same event. When the brushes move at different times, that evidence weakens. Yet synchrony alone is not the whole story. The artificial hand’s form, its orientation and its position relative to the person can affect whether the scene makes bodily sense. A wooden block in an impossible orientation may be a much harder candidate for ownership than a hand aligned with an arm.
Tsakiris and Haggard’s follow-up experiments explored how visual-tactile matching and self-attribution interact. Their findings helped establish that top-down constraints about what a body can be and bottom-up sensory correspondence both contribute. A person does not merely add matching signals regardless of what object receives them. Nor does an existing body model make new sensory information irrelevant. The task exposes an interaction between expectations and current evidence.
This interaction is a reason to avoid a simple slogan such as “vision overrides touch.” The person still feels the real brush. Vision helps determine where that touch seems to belong. Proprioception supplies a different estimate of hand location. The outcome reflects how multiple signals are weighted under uncertainty. The exact weighting cannot be read off from a compelling anecdote; it needs systematic comparisons.
A later methodological review by Riemer and colleagues surveyed how studies differ in hand placement, stimulation, induction time, outcome measures and analysis. Results that appear inconsistent may sometimes arise from different protocols. That possibility does not explain away every discrepancy. It makes replication more informative when the procedure and outcome are specified closely enough to compare.
What brain imaging adds, and what it does not
Ehrsson, Spence and Passingham combined a related rubber-hand manipulation with functional magnetic resonance imaging. They reported premotor activity associated with the feeling that the artificial hand was one’s own. The study supplied a neural correlate for one version of the illusion and brought bodily self-representation into closer dialogue with multisensory neuroscience.
A correlate is not a complete causal account. Activity in a brain area during an illusion does not prove that the area alone generates ownership, or that one can diagnose an individual’s bodily experience from a scan. Imaging comparisons depend on the control condition, reported experience and the timing of the task. Later work has broadened the network and the possible mechanisms. The 2004 result is useful because it tied a defined subjective change to a measured neural pattern, not because it closed the question.
The relation between a laboratory trick and ordinary body ownership is also selective. Most of the time, vision, touch, proprioception and movement are richly coordinated across many events. The rubber-hand setup removes some ordinary evidence and engineers a conflict. It is a probe of how the system can update, not a complete simulation of living with a body.
Ownership, agency and position are different questions
Someone can feel that a hand belongs to them without deciding to move it, and can move a tool without feeling that the tool is literally part of their body. Ownership concerns the relation “this is mine”; agency concerns initiating or controlling an action; position sense concerns where a body part seems to be. The rubber-hand illusion is especially useful because it can alter one of these while the others remain more stable.
Experiments that add movement, virtual reality or a robotic limb raise new questions, but they should not be treated as direct replications of the 1998 brushing task. A moving artificial hand supplies action-related information. A virtual display changes realism and timing. A prosthetic device is embedded in daily goals and adaptation. It may be appropriate to connect these literatures, yet each has a different outcome and time scale.
For clinical or design work, the distinction prevents overpromising. A brief ownership illusion does not establish that a device will be accepted, useful or safe over months. Nor can failure to report the illusion diagnose a disorder. People vary in attention, sensory thresholds, interpretation and response to the setup. The original paper opens a research question about integration; it is not a treatment manual or a test of human worth.
How a strong replication would read
A careful replication would describe the artificial hand’s position, the hidden hand’s posture, the stroke timing and the duration. It would compare synchronous and asynchronous conditions and report who was included. It would specify whether ownership questionnaire items, referred touch or position judgments are primary outcomes. If several measurements are taken, it would not silently select only the most dramatic one afterward.
The analyst would look at individual variation and uncertainty around estimates, not just a group average. A participant who reports no ownership may still show some drift, and vice versa. If the study asks whether hand shape matters, it should manipulate shape while keeping other features as constant as possible. If it asks about the real-world relevance of an illusion, it should identify what everyday outcome would count as transfer and test that outcome separately.
One more comparison concerns expectation. If a participant can see what the researcher wants, they might answer in that direction without actually experiencing ownership. Conversely, a participant who has a strong experience may hesitate to describe it because the claim sounds odd. Neither possibility is settled by dismissing questionnaires. Researchers can use carefully worded control items, more than one outcome and transparent instructions. They can also ask participants to describe the experience in their own words, while acknowledging that open descriptions bring their own interpretation challenges.
Time should be reported as well. Does a feeling emerge immediately or only after sustained brushing? Does it vanish when stimulation stops? A brief position shift measured after the task may not have the same course as the ownership report. Comparing these trajectories can test whether the two measures share a process or merely coincide at one selected moment. Such detail turns a dramatic effect into a series of answerable questions.
Riemer and colleagues’ methods review makes this discipline especially important. Two reports can both call their task “the rubber-hand illusion” while differing in enough details to create different opportunities for the effect. Openness about the procedure helps readers decide whether a new result narrows, extends or fails to reproduce an earlier finding.
Institute synthesis: the body is an active inference problem
The Institute uses this paper as an example of a broader capacity question: how do people maintain a useful sense of their own body while information changes? In ordinary life, the body must be stable enough for action and flexible enough to use a tool, adjust posture or respond to injury. The rubber-hand task dramatizes flexibility under constrained laboratory conditions. It does not establish a universal recipe for changing body experience.
Calling the body an “active inference problem” is a conceptual synthesis here, not a claim that the 1998 paper tested one current computational theory. The relevant practical habit is more modest: when an experience of body position or ownership changes, ask which signals, expectations and tasks changed too. Resist turning a memorable demonstration into a single story about brain control, suggestibility or identity.
For designers of technology, the paper suggests questions rather than promises. Does a haptic cue arrive at the right time? Does the visual object align with expected movement and position? Is the user’s subjective comfort assessed separately from task performance? A convincing effect in a brief demonstration is a reason to test durability and safety, not to assume them.
There is a social dimension to this caution. A person describing an unusual bodily experience should not be told that a laboratory illusion explains it. The experiment was created to vary a stimulus in healthy research participants, not to interpret someone’s history, distress or medical condition. The respectful use of the study is to show that bodily experience is constructed from multiple sources and to remain curious about what evidence is missing in a different setting.
What the evidence does not show
The 1998 correspondence does not prove that people accept any object as part of their body. It studied a plausible rubber hand under deliberately aligned visual and tactile events. Later studies show that timing, form, position, task and measurement matter. The illusion is not a general statement that vision always dominates the other senses.
Proprioceptive drift is not a transparent measure of felt ownership. Rohde and colleagues and Abdulkarim and Ehrsson demonstrate why their relationship needs testing. Brain imaging does not reduce ownership to a single region. Neither the original experiment nor its extensions show how to diagnose an individual or prescribe a rehabilitation program from one illusion score.
There is also no direct evidence here that a stronger laboratory illusion produces better everyday embodiment, mental health or device use. Those claims would require relevant populations, comparison conditions and longer outcomes. The appropriate conclusion is narrower and more interesting: bodily experience can be updated by coordinated sensory evidence, but its components and limits must be measured separately.
Return to the table
The person looks at the rubber hand and feels the brush on the hidden skin. For a moment the visible hand is an unusually persuasive candidate for “mine.” Then the brushes stop, and the person can look at the real hand again. The experiment leaves a question that extends beyond the trick: how does a body remain coherent while it continually incorporates new information?
Botvinick and Cohen’s paper gave researchers a remarkably economical way to ask that question. Its enduring value is not the spectacle of a fake hand. It is the invitation to distinguish touch, location, ownership and action, and to test how their relations change when a situation is carefully rearranged.
Research references
- Botvinick, M., & Cohen, J. (1998). Rubber hands ‘feel’ touch that eyes see. Nature, 391, 756. View source
- Tsakiris, M., & Haggard, P. (2005). The rubber hand illusion revisited: Visuotactile integration and self-attribution. Journal of Experimental Psychology: Human Perception and Performance, 31(1), 80–91. View source
- Rohde, M., Di Luca, M., & Ernst, M. O. (2011). The rubber hand illusion: Feeling of ownership and proprioceptive drift do not go hand in hand. PLOS ONE, 6(6), e21659. View source
- Abdulkarim, Z., & Ehrsson, H. H. (2016). No causal link between changes in hand position sense and feeling of limb ownership in the rubber hand illusion. Attention, Perception, & Psychophysics, 78, 707–720. View source
- Ehrsson, H. H., Spence, C., & Passingham, R. E. (2004). That’s my hand! Activity in premotor cortex reflects feeling of ownership of a limb. Science, 305(5685), 875–877. View source
- Riemer, M., Trojan, J., Beauchamp, M., & Fuchs, X. (2019). The rubber hand universe: On the impact of methodological differences in the rubber hand illusion. Neuroscience & Biobehavioral Reviews, 104, 268–280. View source