NOTE-1004: Longo et al. — What is embodiment? A psychometric approach

Contribution

The paper is the first to measure the experience of the rubber hand illusion with many items in a large sample and to ask what latent structure the reports have. It gives the field a 27-item questionnaire, a decomposition of “embodiment” of a hand into ownership, location and agency, and a link between the ownership and location components and proprioceptive drift. It also finds that the asynchronous control is not a neutral baseline: it has an experience of its own, deafference.

Key insight

Owning a seen hand is not one feeling. In report, it is a cluster of three, the hand being mine, being where my hand is, and being under my control, which move together across people but can be pulled apart. The illusion moves the first two and not the third. On the authors’ reading, “‘me-ness’ is not reducible to ‘here-ness’” (p. 992), and the new hand replaces the old one instead of being added to it.

Assumptions

• Item covariance across people reveals the structure of an experience. “To the extent that participants do not agree, no structure should emerge. Conversely, to the extent that structure does emerge … it demonstrates agreement (however implicit) across participants regarding the structure of their experience” (p. 994). This is the inference the whole paper rests on.
• The items cover the experience. The 27 statements came from free reports of five pilot participants and from hypothesised constructs such as ownership and agency (p. 982). The authors grant that “there are no doubt additional components of embodiment which were not covered by our questionnaire” (p. 995). No item asks about the whole body, self-location in the room or perspective.
• PCA with varimax rotation, run separately per condition. The number of components came from “the scree plot and eigenvalues” (p. 982). Varimax makes the components orthogonal by construction, so component scores are uncorrelated within a condition.
• The secondary split. Items loading above .50 on embodiment (items 1–10) were re-analysed alone (p. 986). Three subcomponents were kept, though by Table 2 the second and third have eigenvalues of 0.75 and 0.66 (synchronous) and 0.84 and 0.71 (asynchronous), against 6.49 and 6.07 for the first. Before rotation, then, one component carries about 65% and 61% of the variance of these ten items (my arithmetic from Table 2’s eigenvalues). The paper does not state the criterion for keeping components below 1.
• Design. Within subjects: each participant did both conditions, in alternating order (66 synchronous first, 65 asynchronous first), on the right hand (N = 67) or left (N = 64). Induction was 60 s at about 1 Hz; asynchrony was 180° out of phase (pp. 981–982). The first two items were always “enjoyable” and “interesting”; the rest were randomised.
• Sample. “Current and prospective students” recruited at university open days, 75 of 131 female (p. 981).

Key results

• Top-level structure (Table 1, pp. 983–986). Synchronous: four components, 55.3% of variance; embodiment of rubber hand 26.3% (eigenvalue 9.52), loss of own hand 12.2%, movement 9.0%, affect 7.8%. Asynchronous: five components, 59.4%; embodiment 24.6% (9.74), then deafference 9.6%, movement 9.2%, loss of own hand 8.4%, affect 7.6%. Deafference (items 23–25: pins and needles, numbness, “the experience of my hands was less vivid than normal”) loads only in the asynchronous analysis (.580, .780, .711).
• Item 27 loads on nothing. “It seemed like I was feeling the touch of the paintbrush in the location where I saw the rubber hand being touched”, close to the item the original report found most strongly affirmed, has no loading above .50 in either condition (communalities .199 and .277; Table 1, p. 985). Item 8, that the touch “was caused by the paintbrush touching the rubber hand”, loads on embodiment and then on location. The paper does not comment on item 27.
• Ownership, location, agency (Table 2, p. 987). Synchronous rotated variance 35.4%, 24.8%, 18.8% (79.0% in all); asynchronous 34.3%, 23.7%, 18.2% (76.2%). Ownership: items 1–5 (looking at my own hand, rubber hand resembling mine, belonging to me, being my hand, being part of my body; loadings .696–.811 synchronous). Location: items 6–8 (.776, .831, .670). Agency: items 9 (“I could have moved the rubber hand if I had wanted”, .843) and 10 (“I was in control of the rubber hand”, .745). In the asynchronous analysis items 1 and 6 load on both ownership and location.
• Presence by condition, top level (pp. 988–989, Fig. 1). Condition × component interaction F(3, 390) = 69.24, p < .0001 (partial η² ≈ .35, my computation). Embodiment was “significantly positive following synchronous stroking, but significantly negative following asynchronous stroking”: about +0.7 against −1.2 on the −3 to +3 scale, read off Fig. 1. Loss of own hand was higher after synchrony, about +0.1 against −0.3, so near neutral in both. Movement was more negative after synchrony (about −2.4 against −1.2); affect positive in both (about +1.6 against +1.35). Deafference after asynchrony, about +0.3.
• Presence by condition, subcomponents (p. 989, Fig. 2). Condition F(1, 130) = 110.69, p < .0001; component × condition F(2, 260) = 3.60, p < .05 (partial η² ≈ .04, my computation). “Component scores for ownership and location were positive following synchronous stroking, but negative following asynchronous stroking. Agency scores were negative in both conditions.” Read off Fig. 2: ownership about +0.35 against −0.55, location about +0.7 against −0.65, agency about −0.55 against −1.1. So synchrony raised agency scores too, by roughly as much as the others, without bringing them above zero.
• Sex and laterality (p. 990). Women agreed more with items overall (−.055 against −.324; F(1, 127) = 4.06, p < .05), which the authors relate to women being “slightly more suggestible than males”. The effect did not interact with component or synchrony. Laterality had no effect. Suggestibility was not measured.
• Proprioceptive drift (p. 990). n = 120 (11 missing or unusable). Pre-test bias toward the midline 0.58 cm, t(119) = 2.43, p < .05. Further displacement toward the rubber hand after synchronous stroking 1.34 cm, t(119) = 4.56, p < .0001; after asynchronous 0.30 cm, t(119) = 1.20, n.s.; difference t(119) = 3.22, p < .005 (dz ≈ 0.29, my computation). Drift was measured before the questionnaire.
• Report predicts drift (pp. 990–991). Regressing the synchronous minus asynchronous difference in drift on the differences in component scores: embodiment β = 0.609, t(115) = 3.40, p < .001; movement β = 0.290, p = .096; loss of own hand and affect n.s. Among the subcomponents: location β = .485, t(116) = 3.02, p < .005; ownership β = .425, t(116) = 2.85, p < .01; agency β = .110, n.s. “Proprioceptive displacement therefore appears to be a genuine by-product of embodiment, but also related specifically to body ownership, but not agency” (p. 991).
• Body constancy (pp. 991–992). From loss of own hand: “the novel rubber hand is incorporated by functionally suppressing the existing hand, rather than by adding an additional ‘supernumerary’ hand”. Item 19, “I had three hands”, loaded on movement, not on embodiment.
• Touch goes with location, not ownership (p. 992). Because item 8 loaded on location, “somatic sensory experiences are tied to location, not to ownership”, which the authors compare with somatoparaphrenic patients who feel touch on a hand they disown.
• Two causes (p. 993). The asynchronous condition counts as conflict only if “some rudimentary form of ownership of the rubber hand is elicited merely from the visual perception of the rubber hand”, so “there are at least two types of cause of the rubber hand illusion: purely visual information from the perception of the hand in plausible configuration, and multisensory synchrony.”
• Ownership and agency, both ways. “We provided the first direct empirical evidence confirming the intuition that body ownership and agency reflect dissociable components of embodiment” (p. 991). And: “ownership and location along with the sense of agency grouped together in the omnibus PCA … This experiential link between the senses of ownership and agency belies the traditional view that these are distinct” (p. 995).

Claims

id	claim	strength	support
C1	Synchronous brushing raises reported embodiment of a rubber hand above neutral and asynchronous brushing leaves it below	strong	n = 131, within subjects, F(3, 390) = 69.24 interaction, Fig. 1
C2	Drift toward the rubber hand follows synchronous but not asynchronous brushing	moderate	n = 120, 1.34 cm against 0.30 cm n.s., difference t(119) = 3.22; dz ≈ 0.29
C3	Reports of the illusion have a stable four-component structure across conditions	moderate	separate PCAs per condition, Table 1; same sample in both, so not an independent replication
C4	Embodiment of a hand comprises dissociable ownership, location and agency	weak to moderate	secondary varimax PCA on ten items, two of three components with eigenvalues below 1; the split is a rotation within a dominant single component
C5	Ownership and location each predict drift independently, and agency does not	moderate	multiple regression on difference scores, β = .425 and .485; correlational
C6	Synchrony raises ownership above neutral without raising agency above it	moderate	Fig. 2; agency rose with synchrony too, staying negative; the hand never moved
C7	The rubber hand displaces the real hand rather than adding to it (“body constancy”)	weak	loss of own hand near neutral in both conditions, higher after synchrony; not related to drift; one item about three hands
C8	Asynchronous brushing produces a distinct experience of deafference	moderate as a finding, weak as an interpretation	component in the asynchronous PCA only, Fig. 1 about +0.3; the gating account is speculation
C9	Covariance structure in introspective reports measures the structure of an experience	weak	argued (p. 994); nothing separates shared experience from shared reading of the items or shared expectation

Concepts

• Embodiment — “the sense of one’s own body” (p. 979), here measured as the experience of a rubber hand as part of one’s body. The authors grant the word has other senses in cognitive science (p. 995).
• Ownership — items 1–5: the rubber hand belongs to me, is my hand, is part of my body, looks like mine, and I seem to be looking at my own hand.
• Location — items 6–8: my hand is where the rubber hand is, or the reverse, and the felt touch is caused by the brush on the rubber hand. It is the location of a hand, not of the self.
• Agency — items 9–10: felt capacity to move and control the rubber hand. No movement occurs, so these are judgements of possible control, not the experience of controlling an act.
• Loss of own hand — items 12–16: unable to move my hand, my hand has disappeared, out of my control, couldn’t tell where it was, and (negatively) could have moved it.
• Deafference — items 23–25: pins and needles, numbness, hands less vivid than normal.
• Body constancy — the authors’ proposed principle that an incorporated object replaces an existing part.

Connections

The study is by Haggard’s group, then at UCL, and builds on Botvinick & Cohen (1998), which it cites as the source of the illusion, and on Tsakiris & Haggard (2005) and Tsakiris, Prabhu & Haggard (2006) on ownership and agency. It takes Gallagher (2000) as the source of the ownership–agency distinction and Jeannerod (2007) as a view on which location is constitutive of selfhood, which it argues against. It sets itself against accounts that tie bodily self to a single system, naming Damasio’s visceral interoception (p. 993). None of the cited works was read here.

Bearing on the record

• THEORY-765 (Proposed), the synchrony effect. This is the first replication in the record that gives what the first half of the promotion test asks for: a sample size (131, 120 for drift), named tests (repeated-measures ANOVA, t-tests, regressions) and drift measured in both conditions. Effect sizes are not reported as such but can be computed from the statistics given. It partly meets the test and does not meet its core: there is no measure or matching of expectations, and no trait suggestibility measure. The sex effect, which the authors relate to suggestibility, did not interact with synchrony, but that is not a suggestibility control. The design is within subjects, so each participant saw both conditions and could infer which was meant to work. It bears against the drift refutation clause: drift was 1.34 cm after synchrony and 0.30 cm, not significant, after asynchrony, so drift did not appear as strongly under asynchrony. It also differs from the 1998 report in ways the THEORY should note: 60 s of brushing rather than ten minutes, and item 27, close to the original’s most strongly affirmed item about referred touch, did not load on any component.
• THEORY-769 (Proposed), graded common-cause ownership. The loss-of-own-hand component, higher after synchrony, is a sign on a measure other than temperature or histamine that the real hand loses something as the rubber hand gains ownership, which is what the promotion test’s second result asks for. It partly meets that requirement and no more. The component sits near neutral in both conditions, its items are about paralysis, disappearance and control, not about the real hand ceasing to be mine, and it did not predict drift. Because varimax makes components uncorrelated, the paper cannot show the real hand’s loss tracking the rubber hand’s gain across people. Nothing is varied parametrically and no model is fitted, so the first, model-based half of the test is not touched. The authors’ “two causes”, visual plausibility and synchrony (p. 993), fit the THEORY’s use of a prior, but are inferred from other studies, not shown here.
• THEORY-767 (Proposed), interoception. No bearing. There are no interoceptive items or measures. The authors reject linking bodily self to “a single somatic sensory system such as visceral interoception” (p. 993), but that is an argument against singularity, not evidence against an interoceptive contribution.
• THEORY-768 (Proposed), minimal selfhood as whole-body identification. It does not meet the first promotion condition and does not refute it: no item measures whole-body self-location or self-identification, so it is another rubber-hand study that did not look. What it adds is a part-level parallel, which is the record’s pairing and not the authors’: ownership (identification with a hand) and location (where the hand is felt) are separable within the hand, much as Blanke & Metzinger separate global self-identification from self-location (LIT-1274, NOTE-1002). That fits the THEORY’s dimensions at a smaller scale; it is no evidence that the hand reaches the whole. It says nothing on the second, developmental condition.
• THEORY-458 (Proposed), ownership and agency. The condition means give ownership above neutral and agency below it after synchrony, which is consistent with the distinction. The structure also shows the two cohering: the agency items load with ownership on the top-level component (.651 and .740), and synchrony raised agency scores along with ownership. The authors say both that the two are dissociable and that their grouping “belies the traditional view that these are distinct”. The agency measured is felt capacity to move a hand that never moved, not the sense of initiating an act, so the comparator half of THEORY-458 is not tested. Haggard’s own later review (LIT-1009, NOTE-827) takes agency in the narrow motor sense these items do not reach, and THEORY-322 is about that sense; neither is touched.
• QUESTION-048. The paper’s central argument (p. 994) is that structure in the covariance of introspective reports across people “demonstrates agreement … regarding the structure of their experience”, and that its relation to drift “attests to its validity”. This is report evidence, and it does not separate experience from report. Items with similar wording can cluster because people read them alike, and a shared expectation of what the synchronous condition should feel like would produce shared structure too. The drift link is convergence with a behavioural measure that is itself not a measure of experience. The works by Lush and colleagues being filed alongside address the expectation side.
• NOTE-372, claim C6. A second, larger primary demonstration that ownership of a body part moves with multisensory conflict. Like Botvinick & Cohen it concerns a part, not the global ownership C6 calls “the most simple form of self-awareness”.
• No instruction for machine-learning practice.

Limitations

• No control for expectation or demand. Within-subjects design, predictable intended answers, no expectation or suggestibility measure. The authors invoke suggestibility only to explain the sex difference.
• The three-way split is thin. It comes from a second rotation of ten items whose first unrotated component carries most of their variance, with two components below an eigenvalue of 1 retained. Different rotation or retention choices could give a different split. The paper reports no reliability coefficients and no confirmatory analysis.
• Same sample, two analyses. The “convergent evidence” that the same components appear in both conditions (p. 986) comes from the same 131 people rating the same items twice, so it is not an independent replication of the structure.
• Small item-to-person ratio for PCA. 27 items and 131 people, about five per item.
• Short induction. 60 s, with drift measured before the questionnaire, so the questionnaire reflects the experience after a delay.
• Figure values are approximate. The component means are given only as bar charts; the values above are read off them. Item means are in a supplement not read here.
• A hand, not a body. The authors say the structure “may or may not generalise” to other senses of embodiment (p. 995). It says nothing about self-location or perspective.

Open questions

• Does the four-component structure, and the ownership–location–agency split, survive a confirmatory factor analysis in an independent sample?
• Does the structure survive when expectations of the two conditions are measured or matched? If participants expecting the illusion produce the same pattern without synchrony, the structure is of expectation, not of the experience.
• Would whole-body items (where am I, which body am I) move with the hand-level ownership and location components, or stay put? That is the measurement THEORY-768’s promotion test asks for.
• Is loss of own hand traded against rubber-hand ownership person by person? An oblique rotation or a direct correlation of item scores would show it; the orthogonal analysis here cannot.
• Why did the classic referred-touch item load on nothing?