Extending body space in immersive virtual reality: a very long arm illusion

Abstract

Recent studies have shown that a fake body part can be incorporated into human body representation through synchronous multisensory stimulation on the fake and corresponding real body part - the most famous example being the Rubber Hand Illusion. However, the extent to which gross asymmetries in the fake body can be assimilated remains unknown. Participants experienced, through a head-tracked stereo head-mounted display a virtual body coincident with their real body. There were 5 conditions in a between-groups experiment, with 10 participants per condition. In all conditions there was visuo-motor congruence between the real and virtual dominant arm. In an Incongruent condition (I), where the virtual arm length was equal to the real length, there was visuo-tactile incongruence. In four Congruent conditions there was visuo-tactile congruence, but the virtual arm lengths were either equal to (C1), double (C2), triple (C3) or quadruple (C4) the real ones. Questionnaire scores and defensive withdrawal movements in response to a threat showed that the overall level of ownership was high in both C1 and I, and there was no significant difference between these conditions. Additionally, participants experienced ownership over the virtual arm up to three times the length of the real one, and less strongly at four times the length. The illusion did decline, however, with the length of the virtual arm. In the C2-C4 conditions although a measure of proprioceptive drift positively correlated with virtual arm length, there was no correlation between the drift and ownership of the virtual arm, suggesting different underlying mechanisms between ownership and drift. Overall, these findings extend and enrich previous results that multisensory and sensorimotor information can reconstruct our perception of the body shape, size and symmetry even when this is not consistent with normal body proportions.

Figure 1. The Head-Mounted Display and Tracking.

(A) Participants experienced the virtual environment through a stereo wide field-of-view Head Mounted Display. (B) Upper limbs were tracked by 12 Optitrack markers grouped in 4 trackable objects. The right and left forearms were tracked for all participants. For right handed people, the right hand and the left index finger were also tracked. For left handed people, the positions of the markers were swapped and thus the right finger and the left hand were tracked.

Figure 2. Spatial configuration of the physical and virtual scene.

(A) There were two physical boxes, the Stimulus Box shown on the right and Angle Box shown on the left. For left handed people the positions of the boxes were swapped. (B) A plastic ring was attached on top of the Angle Box. The participant was asked to put his or her dominant hand on the Stimulus Box and the other one on the Angle Box with the elbow in the plastic ring. (C) There were virtual replicas of the physical boxes. (D) In all Congruent conditions, the virtual dominant hand of the participant was seen to touch the virtual Stimulus Box corresponding to the real hand touching the real Stimulus Box. In these conditions, when the participant moved the hand over the surface of the Stimulus Box feeling its material, the same movement was made and the same tactile feedback was seen. (E) In the Incongruent condition, the virtual Stimulus Box was placed 4 meters frontwards. (F) Therefore, although the virtual movement was the same as the physical, the virtual hand was never seen to touch the virtual replica of the Stimulus Box.

Figure 3. Estimation of the position of dominant hand using the non dominant arm.

(A) The two limbs were aligned as shown pointing forward as shown, in the case shown here the dominant hand was the right hand. (B) The participant was instructed to rotate the non dominant arm to point with the index finger towards where they felt the other hand to be. The position of the elbow was restricted by the plastic ring. The angle was recorded. Each participant repeated this 10 times before elongation (to give the mean AngleBefore) and once after the elongation (to give the AngleAfter) for conditions C2, C3 and C4.

Figure 4. The elongation of the virtual arm and the threat event to the virtual hand.

(A) At the start of the experiment, both virtual arms were of the same size as the participant’s arms. In C1 the virtual arm did not change length during the experiment. (B) The arm elongated to double the true length (C2) (C) The arm elongated to triple the true length (C3) (D) The arm elongated to four times the true length (C4). When the elongation was complete for the condition and after the last angular estimation was made, a virtual saw fell to cut the virtual arm. The participants had been instructed to stay motionless just before this. (E) The position of the virtual threat was also close to the physical body and the real hand in the no elongation condition C1 (F) The threat was far from the real body and real hand in condition C4.

Figure 5. Estimated probabilities of the scores on the illusion of ownership (Q3).

The probabilities are estimated from the fitted values of the ordered logistic regression of the Q3 scores on elongation.

Figure 6. Means and Standard Errors of the Angular drifts for the elongation conditions.

AngleBefore is the mean of 10 estimations of hand position at the start of the experiment. AngleAfter is the single estimation of hand position after the arm elongation period. AngleAfter is significantly greater than AngleBefore for C2 (P = 0.04) and C3 (P = 0.01) but not for C4 (P = 0.17), Wilcoxon matched-pairs signed-rank tests.

Figure 7. Scatter diagram of the Saw Time dispersion on the Control Time Dispersion by Elongation Condition on a log-log scale.

Figure 8. Path analysis diagram.

The boxes represent the variables where elongation is 1 to 4, ownership is the response on Q3, harmed is the response on Q7, and and are the dispersions in Control Time and Saw Time. The variable longerarm is the response on Q6, and anglediff is AngleBefore-AngleAfter. The paths represent the regression lines, where, for example, where The circles represent the random error terms and the corresponding numbers are the variances of the errors.