The illusion of owning a third arm

Abstract

Could it be possible that, in the not-so-distant future, we will be able to reshape the human body so as to have extra limbs? A third arm helping us out with the weekly shopping in the local grocery store, or an extra artificial limb assisting a paralysed person? Here we report a perceptual illusion in which a rubber right hand, placed beside the real hand in full view of the participant, is perceived as a supernumerary limb belonging to the participant's own body. This effect was supported by questionnaire data in conjunction with physiological evidence obtained from skin conductance responses when physically threatening either the rubber hand or the real one. In four well-controlled experiments, we demonstrate the minimal required conditions for the elicitation of this "supernumerary hand illusion". In the fifth, and final experiment, we show that the illusion reported here is qualitatively different from the traditional rubber hand illusion as it is characterised by less disownership of the real hand and a stronger feeling of having two right hands. These results suggest that the artificial hand 'borrows' some of the multisensory processes that represent the real hand, leading to duplication of touch and ownership of two right arms. This work represents a major advance because it challenges the traditional view of the gross morphology of the human body as a fundamental constraint on what we can come to experience as our physical self, by showing that the body representation can easily be updated to incorporate an additional limb.

Figure 1. Illusion set-up.

Set-up used to elicit the supernumerary hand illusion (left panel) and an illustration of the threat procedure, where we moved a knife close to the rubber hand or the real hand (right panel).

Figure 2. Control conditions.

Set-ups used for the three different control conditions used in experiments 1, 3 and 4. The application of the brushstrokes on the real and artificial limbs is demonstrated in the left picture of each picture pair; to the right, the procedure of threatening the rubber limb with the knife is depicted. From left to right: the rotated rubber right hand, rubber left hand and rubber right foot conditions.

Figure 3. Supernumerary hand illusion vs. traditional rubber hand illusion.

Set-up used in experiment 5 to elicit the supernumerary hand illusion (left panel) and the traditional rubber hand illusion (right panel). Note that the only difference was the screen occluding vision of the real right hand in the latter condition.

Figure 4. Introspective evidence for the supernumerary hand illusion.

Questionnaire data from experiment 1 comparing synchronous brushing (the illusion condition) with asynchronous brushing and using a rotated rubber hand (the control conditions). The questionnaire consisted of ten statements, S1–S10, and the participants indicated their responses on a ten-point visual analogue scale ranging from 0 (“I do not agree at all”) to 9 (“I agree completely”). We observed significantly stronger rubber hand ownership (S1–S2) (p<0.001 and p<0.001, respectively), no significant difference in real hand disownership (S3–S4) (p = 0.137 and p = 0.063, respectively) and a significantly stronger feeling of owning two right hands (S5–S6) (p<0.001 and p = 0.002, respectively) during the synchronous condition compared to the asynchronous and rotated rubber hand control conditions, respectively. The error bars represent the standard error.

Figure 5. Physiological evidence for the supernumerary hand illusion.

Physiological data from experiment 2 showing the mean skin conductance response (SCR) for 44 participants when the real hand or the rubber hand was threatened during synchronous brushing (the illusion condition) or asynchronous brushing (the control condition). There was a significantly greater SCR when threatening the rubber hand during the illusion (p = 0.001), but no significant difference in the SCR when threatening the real hand during the illusion compared to the control condition (p = 0.534). Thus, participants experienced ownership of the rubber hand without disowning their real hand. The error bars represent the standard error.

Figure 6. Introspective evidence for what type of limb that can be owned as a supernumerary limb.

Questionnaire data from experiment 3 where synchronous brushing was applied on the real right hand and a rubber right hand (the illusion condition), a rubber left hand or a rubber right foot (the control conditions). During the illusion, we observed significantly stronger rubber hand ownership (S1–S2) compared to the rubber left hand (p = 0.007) and rubber right foot conditions (p<0.001), respectively; significantly stronger real hand disownership (S3–S4) than for both the rubber left hand (p<0.001) and the rubber right foot conditions (p<0.001); and a significantly stronger feeling of having two right hands (S5–S6) compared to the rubber right foot condition (p<0.001). This implies that participants only experience ownership of a supernumerary rubber hand when it resembles the real hand in respect to laterality (i.e. right-left matching) and limb type (i.e. both hands, but not a hand and a foot). The error bars represent the standard error.

Figure 7. Physiological evidence for what type of limb that can be owned as a supernumerary limb.

Physiological data from experiment 4 showing the mean SCR for 26 participants when the artificial limb was threatened during the illusion (rubber right hand) and three control conditions (rotated rubber right hand, rubber left hand and rubber right foot). Planned comparisons revealed significantly greater SCR when threatening the artificial limb during the illusion compared to each of the three control conditions involving a: Rotated rubber right hand (p = 0.001), rubber left hand (p = 0.048), rubber right foot (p<0.001), respectively. Thus, these results provide SCR evidence that the rubber limb needs to resemble the real limb in respect of anatomical alignment, laterality and limb type for the supernumerary limb illusion to arise. The error bars represent the standard error.

Figure 8. The unique qualities of the supernumerary hand illusion.

Questionnaire data from experiment 5 demonstrating the differences between the supernumerary hand illusion and the traditional rubber hand illusion. During the supernumerary hand illusion, participants experienced significantly weaker rubber hand ownership (S1–S2) (p<0.001), less real hand disownership (S3–S4) (p = 0.001) and had a stronger feeling of having two right hands (S5–S6) (p<0.001). The error bars represent the standard error.