NOTE-1009: Rohde, Di Luca & Ernst — Feeling of ownership and proprioceptive drift do not go hand in hand

Contribution

Before this study, proprioceptive drift was the standard objective measure of the rubber hand illusion, on the assumption that synchronous stroking produces both the feeling of ownership and the drift. The study adds the two conditions that test the assumption: seeing the rubber hand with no stroking, and asynchronous stroking broken up by frequent measurement. Drift appears in both, at the level of the synchronous condition, without ownership. What had looked like synchrony producing drift is, on this evidence, prolonged asynchrony suppressing it.

Key insight

The usual control condition is not neutral. Seeing a hand near where one’s own hand is felt pulls felt position toward it whatever the touch is doing; minutes of mismatched touch tell the system that the seen hand and the felt hand are not one thing, and the pull stops. Comparing synchronous with asynchronous stroking therefore shows a difference in drift and a difference in ownership, and makes them look like one effect when they come from different processes.

Assumptions

• A visual probe measures felt position. Experiment 1 used a two-alternative forced choice (dot left or right of the unseen index finger) with two interleaved staircases; Experiment 2 had participants adjust a dot to the felt finger position, three adjustments per measurement (pp. 2–3). Drift is the change from a pre-test in the dark. The authors grant that the result may not generalise “to other techniques to measure perceived finger location, like inter-manual reaches” (p. 6).
• The questionnaire measures ownership. Botvinick and Cohen’s nine items with a German translation, given once at the end of Experiment 1 (p. 2). In Experiment 2 “No questionnaire ratings were collected”, to avoid biasing the repeated blocks (p. 3).
• Robot stroking gives the illusion. Two PHANToM force-feedback arms with paintbrush tips stroked the hands along random, morphed paths (pp. 2–3). That this produces a “striking and intense ownership illusion” rests on piloting with “Experienced subjects” (p. 2); the Experiment 1 questionnaire is the check.
• Debriefing reports stand in for ownership in Experiment 2. “Anecdotally, the vast majority of participants reported a strong feeling of ownership in the synchronous condition, but not in the asynchronous condition or vision only condition” (p. 6). That vision alone gives no ownership is also supported by citation (Longo et al. 2008, on visual enhancement of touch, ref. 13).

Key results

• Set-up (pp. 2–3). A left rubber hand 17 cm to the right of the hidden real left hand, seen through a semi-silvered mirror; switching the light turned the mirror opaque and showed a probe dot instead, so that during measurement “The dot was the only thing visible in the dark room”.
• Experiment 1 (n = 20, two groups of 10, between subjects; p. 4, Fig. 2). Stroking for 7 min, measurement every 10 s. Questionnaire: only Q1–Q3 were positive, and only in the synchronous group (sign tests; Q1 and Q3 p = 0.002, p = 0.018 after Bonferroni; Q2 p = 0.022, 0.195 corrected). Drift from the last 2 min, fitted with a cumulative Gaussian: 6.50 ± 2.3 cm synchronous (t(9) = 2.7, p = 0.013, one-tailed) and 3.93 ± 1.88 cm asynchronous (t(9) = 2.0, p = 0.039); the difference “was not significant (two-sample t-test, t(18) = 0.8, p = 0.425)”.
• Experiment 2 (n = 30, three groups of 10, synchronous and asynchronous within subjects, counterbalanced; pp. 3–5, Fig. 3). 2 min of stroking, felt position measured once at the end (1×120), three times (3×40) or twelve times (12×10). Synchronous drift reached about 5–6 cm in every group and “gradually built up over time”. Asynchronous drift was 0.98 ± 0.69 cm in 1×120 (against 5.9 ± 1.59 synchronous), 2.75 ± 1.31 cm in 3×40 (against 5.84 ± 1.38) and 5.45 ± 1.48 cm in 12×10 (against 4.81 ± 1.44). Group × stroking interaction F(2,27) = 3.9, p = 0.031; paired tests significant in 1×120 and 3×40, not in 12×10 (t(9) = 0.6, p = 0.592). (The text lists the pairs in different orders; Fig. 4 shows that in 12×10 the asynchronous mean is the larger.)
• Vision only (p. 5, Fig. 4). Third block, no stroking, in the 1×120 and 12×10 groups: 6.47 ± 1.10 cm and 3.09 ± 1.46 cm. In 1×120 it differed from asynchronous (t(9) = 4.8, p < 0.001) and not from synchronous (t(9) = 0.3, p = 0.76). Synchronous and vision-only drifts correlated within participants, r = 0.49, p = 0.030, pooled over 20. Mixed ANOVA group × mode, interaction F(2,36) = 7.8, p = 0.002; no main effect of mode (p = 0.109).
• Not response bias (p. 6). Dots started at random positions centred on the real finger, so any bias from the procedure would pull toward the real hand, not the rubber one.
• The account (pp. 5–6). Asynchrony is evidence against the “unity assumption”, “the assumption that the proprioceptive and the visual sensations belong together”, which visuo-proprioceptive integration needs; frequent interruption resets the accumulation of that evidence. The gradual rise in drift may come from proprioception weakening while the hand is still. “The sensation of owning a rubber hand does not cause any automatic update of the perceived position of the body in space” (p. 6). They leave open that “visuotactile integration may be a pre-requisite of the feeling of ownership” and that drift may be a prerequisite for ownership (pp. 6–7).
• Against drift as a proxy (p. 7). The correlation between drift and ownership seen elsewhere could have “A more remote common cause”, for instance participants who are more “visual”. “If it was admissible to assess the feeling of ownership by proxy of the proprioceptive drift, we could conclude from our results that body ownership is modulated by the frequency of measurement in asynchronous stroking”, which the questionnaire shows is false.
• On other measures (p. 7). The same caution covers skin conductance, temperature and neural correlates: “a correlation is only meaningful in so far as it can be argued to be causally linked to the phenomenon in question”. Indirect measures are attractive because they would bypass “suggestion, variability due to beliefs and top-down influences”, and “The results presented in this paper demonstrate the dangers of such an endeavor.”

Claims

id	claim	strength	support
C1	Seeing a rubber hand, with no stroking, shifts felt finger position toward it as much as synchronous stroking does	moderate	Exp 2 vision only, 6.47 cm against about 6 cm, t(9) = 0.3; within-subject r = 0.49; groups of 10; vision only always third
C2	Prolonged asynchronous stroking suppresses drift; interrupted asynchronous stroking does not	moderate	Exp 2, 0.98 cm (1×120) against 5.45 cm (12×10); interaction p = 0.031; Exp 1 consistent
C3	Ownership requires synchronous stroking, under the same conditions in which drift does not	moderate in Exp 1, weak in Exp 2	Exp 1 questionnaire, groups of 10; Exp 2 anecdotal debriefing and a citation for vision only
C4	Drift and ownership arise from different multisensory processes: visuo-proprioceptive for drift, visuotactile for ownership	moderate as an inference	from C1–C3; the authors leave open that one may be a prerequisite of the other
C5	Asynchrony acts by breaking the “unity assumption”	weak	an interpretation fitted to C2; no independent test
C6	Drift alone is not a valid measure of ownership, and correlates of ownership need a causal argument to serve as measures	strong as a methodological conclusion, given C1–C3	the authors’ conclusion; it does not depend on C5

Concepts

• Proprioceptive drift — the change in felt position of the hidden hand toward the rubber hand, here measured by a visual probe rather than by reaching.
• Visual capture of proprioception — seeing a hand displaced from one’s own biases felt hand position toward it, with or without ownership (refs. 4–7, Holmes et al.).
• Unity assumption — Welch and Warren’s (1980) premise that signals in two senses are taken to come from one object, which integration between them requires.
• Behavioural proxy — Tsakiris’s (2010) phrase for using drift to measure ownership: “proprioceptive drifts can be used as a behavioural proxy” (quoted p. 2).

Connections

The study answers Botvinick and Cohen’s drift experiment and the models built on it: Makin, Holmes and Ehrsson’s peripersonal-space model, in which synchrony adds positive feedback to visual weighting, and Tsakiris’s comparator model. Both, the authors say, can be saved only by giving asynchrony a negative, suppressing role (p. 6). It draws on Holmes and colleagues’ earlier warning that “reaching or proprioceptive biases are not reliable objective measures of the rubber hand illusion itself” (p. 7). Abdulkarim and Ehrsson 2016 (LIT-1268, NOTE-992) takes up its open possibility, that drift is a prerequisite of ownership, and tests it.

Bearing on the record

• NOTE-1006 and LIT-1277 (Botvinick & Cohen 1998). That note asks “Is the drift a measure of ownership or only of visuo-proprioceptive recalibration?” This study’s answer, for drift measured by a visual probe, is recalibration: drift of the same size occurs without stroking and without ownership. Botvinick and Cohen’s correlation between reported duration and drift is not contradicted, but it no longer shows that ownership moves felt position; a common cause, such as how strongly a person weights vision, would give it too. The 1998 asynchronous group’s drift away from the rubber hand (13 mm) fits this account, in which prolonged asynchrony is the active ingredient; that reading is mine.
• THEORY-765 (Proposed). Its second refutation condition is that “drift toward the rubber hand appears as strongly under asynchrony, so that drift tracks visual exposure, not the illusion”. In Experiment 2’s 12×10 group and in Experiment 1 it did, and the vision-only condition shows drift tracking visual exposure. So the study meets that refutation condition for the drift clause of the THEORY, with three caveats: groups of ten, a visual probe rather than reaching, and the classic difference is reproduced when asynchrony is continuous (1×120), so the THEORY’s sentence “a small asynchrony removes these effects” stays literally true of drift while its causal reading, that synchronous brushing shifts felt position, does not. It does not touch the first refutation condition or the suggestion control: the questionnaire was Botvinick and Cohen’s, and expectation was neither measured nor matched. It partly meets the first half of promote_when (sample sizes, named tests and effect sizes for both measures; effect sizes only as means and s.e.m.). My proposal, not made here: narrow the THEORY to referred touch and ownership, and restate the drift as visuo-proprioceptive recalibration that prolonged asynchrony suppresses, citing this reading and Abdulkarim & Ehrsson.
• THEORY-769 (Proposed). Its promote_when asks for a common-cause model that predicts “ownership and drift in held-out” conditions. On this evidence drift is not a readout of ownership, so a model that fits drift would not thereby fit ownership. The result is not against a causal-inference account as such: the authors’ “unity assumption” is itself a common-cause judgement, but they place it between vision and proprioception, separately from the visuotactile judgement that gives ownership. A single graded belief that drives both measures is what the dissociation bears against. It does not test gradedness or the loss of ownership by the real hand. Proposal: the promotion test should treat drift as a separate output to be modelled, not as a second measure of ownership.
• THEORY-767 (Proposed). Its promote_when asks that a heartbeat-locked cue increase ownership “on questionnaire and on drift”. Drift cannot carry that weight: an increase in drift could come from changed visual weighting without any change in ownership. Proposal: drop drift from that test or require that it come with the questionnaire. Nothing here bears on interoception itself.
• THEORY-768 (Proposed). Not tested: nothing whole-body was measured. By analogy, which is mine and not the paper’s, it raises a question for the full-body evidence the THEORY leans on. Lenggenhager et al.’s walking-back drift (NOTE-999) is treated as felt self-location; if hand drift can be visual capture without ownership, whole-body drift may be too, and in that study the object gave drift without self-identification. The THEORY’s first promotion test also groups “hand ownership and drift” as if they moved together; this study says they need not.
• THEORY-458 (Proposed). No bearing: the hand is passive and agency is not at issue.
• QUESTION-048. A clean case of what a behavioural correlate of report can and cannot establish. Drift correlated with ownership reports across studies, and it was adopted as an objective measure that would sidestep suggestion. When conditions were added that pulled the two apart, the correlate proved to measure something else. The authors’ rule, that a correlate counts as a measure only so far as the same mechanism can be argued to produce both, is a usable answer to part of the question.
• NOTE-372, claim C6. The manipulation of ownership by multisensory conflict stands on the questionnaire, which replicated; it gets no support from drift.
• No instruction for machine-learning practice.

Limitations

• Small groups. Ten per group; standard errors of 1–2 cm on drifts of 1–6 cm.
• Ownership in Experiment 2 is anecdotal. The vision-only and interrupted-asynchrony conditions, which carry the dissociation, have no questionnaire; the claim that they gave no ownership rests on debriefing and on a citation. Experiment 1 gives the questionnaire dissociation only for interrupted asynchrony against synchrony.
• The vision-only block was always third, so order is confounded with it, as the authors note (p. 6).
• A visual probe, not reaching. Felt position was judged against a projected dot; whether the result holds for intermanual reaching, the 1998 measure, is left open by the authors.
• Robot stroking with random paths, with a 500 ms pause between strokes in Experiment 2, differs from manual brushing. The questionnaire in Experiment 1 replicates the illusion, but the stimulus is not the classic one.
• No suggestion control. As in the 1998 study, the questionnaire’s predicted items are guessable from the set-up.
• A slip. The text says “The main effect of stroking was due to the 12×10 Group and the 3×40 Group, but not the 12×10 Group” (p. 5); from the t-tests that follow, the first should be the 1×120 group.

Open questions

• Does the dissociation hold for intermanual reaching, and with a questionnaire in every condition?
• If drift and ownership are separate, what explains their correlation across people? The authors’ “more visual” participants is a testable common cause.
• Does continuous asynchrony suppress drift by breaking the unity assumption, or by some other route, for instance by drawing attention to the real hand?
• Is the walking-back drift in full-body illusions also visual capture without identification?