The authors have declared that no competing interests exist.
Conceived and designed the experiments: IS AM NB CP. Performed the experiments: IS. Analyzed the data: IS CP. Contributed reagents/ materials/analysis tools: AM NB CP. Wrote the paper: IS AM NB CP.
Our body is made of flesh and bones. We know it, and in our daily lives all the senses constantly provide converging information about this simple, factual truth. But is this always the case? Here we report a surprising bodily illusion demonstrating that humans rapidly update their assumptions about the material qualities of their body, based on their recent multisensory perceptual experience. To induce a misperception of the material properties of the hand, we repeatedly gently hit participants' hand with a small hammer, while progressively replacing the natural sound of the hammer against the skin with the sound of a hammer hitting a piece of marble. After five minutes, the hand started feeling stiffer, heavier, harder, less sensitive, unnatural, and showed enhanced Galvanic skin response (GSR) to threatening stimuli. Notably, such a change in skin conductivity positively correlated with changes in perceived hand stiffness. Conversely, when hammer hits and impact sounds were temporally uncorrelated, participants did not spontaneously report any changes in the perceived properties of the hand, nor did they show any modulation in GSR. In two further experiments, we ruled out that mere audio-tactile synchrony is the causal factor triggering the illusion, further demonstrating the key role of material information conveyed by impact sounds in modulating the perceived material properties of the hand. This novel bodily illusion, the ‘Marble-Hand Illusion', demonstrates that the perceived material of our body, surely the most stable attribute of our bodily self, can be quickly updated through multisensory integration.
An accurate knowledge of the material properties of our body is essential for adaptive and successful behavior, yet little is known on how the brain achieves such knowledge. Whenever grasping an object, for example, the velocity of reaching and the grip force should be carefully adjusted according to the material of the object and of our body. In order to perceive our body and the world around us, the brain constantly combines multiple sources of incoming sensory information with prior knowledge retrieved from memory
In the present study we set out to investigate whether the brain can update its knowledge about the material properties of the body by inducing an illusory perception of the material of the hand. We gently hit participants' right hand with a small hammer, and manipulated the auditory feedback so that each time the hammer hit the hand, participants heard the sound of a hammer against a stone. Impact sounds provide reliable cues about an object's material properties, and it is well-known that humans are readily able to infer the material of an object just by hearing it bounce
Twenty-three naïve participants seated with their forearms resting on a table, and with their right hand hidden by an opaque screen (
A. Schematic representation of the experimental apparatus. The opaque screen occluded the vision of the hand, but not of the approaching hammer. B. Temporal structure of hammer hits and impact sounds in the MHI and in the control condition. C. Questionnaire results. Scores represent the changes in the response between the post-stimulation and pre-stimulation presentations of the questionnaire. Error bars represent the standard error of the mean, and the asterisks indicate a significant difference (one asterisk p<0.05; two asterisks p<0.01) between the MHI and the control condition. The items of the questionnaire were presented in a random order, and participants provided their responses on a 7-points scale. D. Galvanic skin response results. After the stimulation, the response to threatening stimuli increased for the MHI (p<0.01), but not for the control. Error bars represent the standard error of the mean. E. Correlation between stiffness rating and Galvanic skin response in the MHI group (Pearson's ρ = 0.6; p = 0.02). The data is fitted with a Deming regression line; larger dots represent two points falling in close proximity.
To quantitatively measure the perceptual correlates of the MHI, participants filled in a questionnaire before and after the stimulation. The questionnaire items assessed participants' hand perception in terms of stiffness, heaviness, hardness, temperature, naturalness, sensitivity, and size. Variations between pre- and post- stimulation responses were taken as evidence for the MHI. After the stimulation, several participants spontaneously described their right hand as feeling heavier, harder, and stiffer. They also reported feelings of numbness, pins and needles, and a lack of sensitivity, sometimes extending over the entire forearm (
Spontaneously reported sensations | # occurrences | |
MHI | Controls | |
My hand feels numb (e.g., insensitive, with pins and needles) | 10 | 0 |
My hand feels stiffer | 9 | 2 |
My hand feels heavier | 6 | 0 |
My hand feels colder | 6 | 1 |
The feeling of stiffness and numbness seems to extend to the whole foreharm | 5 | 0 |
My hand feels “different” | 3 | 0 |
My hand feels lighter | 0 | 3 |
My hand feels softer | 0 | 4 |
Questionnaire item |
MHI (post-pre) | Controls (post-pre) | Z | p | r |
Natural as usual - Unnatural | 2.4±0.3 ( |
1.1±0.3 ( |
2.78 | 0.58 | |
Extremely flexible - Extremely stiff | 1.7±0.3( |
0±0.3 (p = 0.8) | 3.26 | 0.68 | |
Extremely light - Extremely heavy | 1±0.3 ( |
−0.3±0.3 (p = 1) | 3.03 | 0.63 | |
Extremely soft - Extremely hard | 0.9±0.3 ( |
0±0.3 (p = 1) | 1.7 | 0.3 | 0.35 |
Extremely sensitive - Extremely insensitive | 0.8±0.4 (p = 0.3) | −0.6±0.3 (p = 0.2) | 2.71 | 0.57 | |
Extremely small - Extremely large | 0.1±0.3 (p = 0.6) | 0.2±0.3 (p = 1) | 0.31 | 1 | 0.06 |
Extremely cold - Extremely hot | −0.4±0.4 (p = 0.6) | 0±0.2 (p = 1) | −0.71 | 1 | −0.15 |
The values in brackets relate to the difference between before and after the stimulation as assessed by the Wilcoxon test. Comparison between MHI and control was done using the Mann-Whitney's U test. The last column reports the effect size r. Asterisks indicate a significant difference between the MHI and the control condition (one asterisk p<0.05, two asterisks p<0.01). All p-values are corrected with Bonferroni-Holm correction.
A second group of 23 naïve participants was tested in a control condition, where hammer hits and impact sounds were temporally uncorrelated
To substantiate the robustness of the effect over and above subjective reports, we measured how the Galvanic skin response to threatening stimuli varied during the MHI (see
Overall, such a GSR modulation would attest to a genuine alteration of body representation induced by the synchronous audio-tactile stimulation, more than any generic report bias due to the introspective nature of the questionnaire. Indeed, the modulation of skin conductance witnesses a modulation of the physiological response to threatening stimuli occurring at the level of the autonomic nervous system. Remarkably, such an increase in skin conductance positively correlated with the feeling of stiffness, hence demonstrating a relation between the phenomenology of the illusion and physiological response.
In the previous experiment, we measured the effects of the MHI by comparing a synchronous audio-tactile condition with an asynchronous audio-tactile condition. As a consequence, it may be argued that the results may be due to the mere synchrony of the auditory feedback in the MHI condition rather than to the material information conveyed by the impact sounds. Therefore we run two further control experiments using different auditory feedback, while maintaining audio-tactile synchrony.
In Experiment 2, we adopted the same procedure of the MHI group of Experiment 1, with the exception that now the auditory stimulus was a pure tone (440 Hz, 25 ms) played in synchrony with the tactile stimulation. Pure tones do not provide any reliable cue to an object's material, therefore they should not affect subjective reports about hand perception nor physiological responses. Eleven participants took part in this experiment. Overall, they did not spontaneously report any altered feelings about the hand following the stimulation, nor they showed any change in their ratings at the questionnaire (all ps>0.5,
A. Questionnaire results. Scores represent the changes in the response between the post-stimulation and pre-stimulation presentations of the questionnaire. Error bars represent the standard error of the mean. The items of the questionnaire were presented in a random order, and participants provided their responses on a 7-points scale. B. Galvanic skin response (GSR) results. GSR was not affected by the stimulation (no significant difference in the GSR before and after the stimulation).
To further exclude that the changes in ratings and GSR reported so far are due to the hammer hitting the hand irrespective of any manipulation of the auditory feedback, we run a third experiment in which participants simply heard the natural sound produced by the actual contact of the hammer against their hand.
Eleven participants were tested in Experiment 3. The GSR was not affected by the audio-tactile stimulation (pre-stimulation mean = 0.45 μs, SD = 0.3; post-stimulation mean = 0.44 μs, SD = 0.45; t = 0.12, p = 0.9,
A. Questionnaire results. Scores represent the changes in the response between the post-stimulation and pre-stimulation presentations of the questionnaire. Error bars represent the standard error of the mean. The items of the questionnaire were presented in a random order, and participants provided their responses on a 7-points scale. B. Galvanic skin response (GSR) results. GSR was not affected by the stimulation (no significant difference in the GSR before and after the stimulation).
When exposed to multisensory signals that correlate in time and space, but provide incongruent cues to body material, the brain can either keep those signals segregated, or else integrate them and resolve the incongruence by updating the perception of body material. The MHI demonstrates that the brain integrates correlated signals
Most features of our body continuously change over time. For instance, the size of our body varies due to growth and posture, the seen and felt position of the hands can be offset by muscle fatigue and distorting lenses, and the texture of the skin varies with moisture. Because of this inherent variability, the estimate of such variables should be updated continuously, based on internal models and incoming sensory information
Previous studies have already shown that sounds can alter body perception
In Experiment 1, to trigger the auditory stimuli, we attached a thin strip of metal foil tape to the right hand of the observers. The tape and the hammer were connected through a wire to a computer, and operated as a switch, so that each time the hammer touched the tape, a custom software based on the Psychtoolbox 3
To create spatialized sounds (that is, to simulate the source of the auditory stimuli in the 3D external space), we recorded the sound of a hammer hitting a marble stone using in-ear recording technology (see
Before and after the stimulation, which lasted approximately 5 minutes, participants filled in a questionnaire written in Italian, the native language of all the observers, consisting of 7 items (see
Overall, 68 right-handed healthy individuals participated in the three main experiments. In Experiment 1 we tested twenty-three participants per condition, for a total of forty-six participants (MHI group: 20 females, mean age = 25, SD = 4.1, range 19–31; Control group: 15 females, mean age = 25.4, SD = 3.7, range 19–35). Eleven participants also took part in Experiment 2 (6 females, mean age = 27.5, SD = 2.9, range 24–33), and Experiment 3 (9 females, mean age = 27.9, SD = 4, range 24–35).
Before and after the experimental stimulation (namely, after filling in the first questionnaire, and before filling in the second questionnaire), galvanic skin response (GSR) to threatening stimuli was recorded in a subset of participants in the MHI group (N = 11) and in the control group (N = 11) in Experiment 1. GSR was recorded in the same number of participants in Experiment 2 (N = 11) and 3 (N = 11). Two Ag–AgCl electrodes (1081FG Skin Conductance Electrode) with constant voltage (0.5 Volt) were attached to subjects' proximal phalanges of the index and middle fingers of their left hand (the non-stimulated hand). At the beginning of each trial, the experimenter held the needle under the table, then lift it above the table-top and moved it toward participant's right index finger, and eventually threatened the participant twice, by approaching the needle to the same finger, but without touching it (see
GSR was recorded with a SC 2071 device (Bioderm, UFI, Moro Bay, California) following standard guidelines
For each threatening stimulus, we calculated the GSR
To measure how the stimulation changed the subjective reports about hand perception, as assessed with the questionnaire, for each item we calculated the difference between the second and the first presentation of the questionnaire (i.e., after and before the audio-tactile stimulation, respectively). The statistical significance of the difference in the responses between the second and first presentation of the questionnaire was calculated using the Wilcoxon Signed rank test in Experiment 1, 2 and 3. Moreover, in Experiment 1 a Mann-Whitney U test was used to test for difference between the MHI and the control groups and we then calculated the effect size using the
To test for effects of the MHI on skin conductivity, in Experiment 1 the GSR data was submitted to an analysis of variance (ANOVA) with order of presentation (before and after the stimulation) as within-participant factor, and type of stimulation (synchronous vs asynchronous) as between-participants factor. None of the main effects reached statistical significance (type of stimulation: F1,20 = 1.47, p = 0.24; order of presentation: F1,20 = 4, p = 0.06), however there was a significant order by type of stimulation interaction (F1,20 = 4.58; p<0.04, total η2 = 0.23), indicating that the GSR changed as a function of both order of presentation and type of stimulation. Post-hoc comparisons using the Newman-Keuls test revealed a significant difference between the first and the second GSR measure in the MHI group (p<0.01) but not in the control group (all other comparisons did not reach statistical significance, ps>0.21).
Pearson's correlation analyses were performed to measure the association between the GSR and the magnitude of the illusion, i.e., the change in GSR before and after the stimulation and the changes in the responses to the questionnaires before and after the stimulation. GSR positively correlated with the subjective report of stiffness (r = 0.6; p<0.02), indicating that the more participants felt their hand to be stiff, the more their GSR to threatening stimuli increased. Notably, taking into account only the subset of participants who spontaneously reported altered sensations from the stimulated hand (see
In Experiment 2 and 3, paired t-tests were used to compare the GSR data before and after the stimulation. None of the paired t-tests reached significance (Experiment 2: t = 1.54; p = 0.15; Experiment 3: t = 0.12, p = 0.9), demonstrating that in Experiment 2 and 3 GSR was not affected by the audio-tactile stimulation.
Additionally, we run further analyses to directly compare results between the MHI condition of Experiment 1 and Experiments 2 and 3. Such analyses are reported in the
This study was conducted in accordance to the declaration of Helsinki, and had ethical approval from the Department of Experimental Psychology at the University of Milano-Bicocca. All participants provided written informed consent and received course credits in return.
(DOCX)
We are grateful to M. Ernst and the CNS team in Bielefeld, and to C. Spence for their comments on the manuscript.