Attenuated Input to the Primary Somatosensory Cortex is Associated with the Occurrence of Rubber Hand Illusion

The neural representation of the body is easily altered by the integration of multiple sensory signals in the brain. The “rubber hand illusion” (RHI) is one of the most popular experimental paradigms to investigate this phenomenon. During this illusion, ownership of the rubber hand is created. Some studies have shown that somatosensory processing in the brain is attenuated when RHI occurs. However, it is unknown where attenuation of somatosensory inputs occurs. Here, we show that somatosensory input from the hand is attenuated at the primary somatosensory cortex. We found that the early response of somatosensory evoked potential, which is thought to originate from the primary somatosensory cortex, was attenuated during RHI. Furthermore, this attenuation was observed before the occurrence of the illusion. Our results suggest that attenuation of somatosensory inputs from the hand to the brain is one of the factors in�uencing the occurrence of the RHI.


Introduction
Neural representation of the body is easily altered when multiple sensory signals are integrated in the brain.One of the most popular experimental paradigms for investigating this phenomenon is the 'rubber hand illusion' (RHI) 1 .In the paradigm, watching a fake rubber hand being stroked by a paintbrush in synchrony and in the same direction with one's own concealed hand creates the feeling that the rubber hand is one's own.Therefore, when visual and tactile signals are in con ict, the visual sense overrides the tactile sense, and the brain can incorporate the feel of a non-corporal object into the body 2,3 .
During the RHI, participants erroneously perceive the fake hand as their own, that is, they fail to perceive their real hand as their own.In this case, the question is how somatosensory inputs from their real hand induced by tactile stimulation are transmitted to the brain.Previous studies have investigated the matter [4][5][6][7] .Zeller et al. 7 recorded somatosensory evoked potentials (SEPs) elicited by a brush stroke during RHI, and illusion touch-evoked SEPs for a wide time window (-200 to 300 ms after stroking) were attenuated in the contralateral pre-and postcentral gyri and the superior and inferior parietal lobes.They speculated that the relative attenuation of somatosensory input may be accompanied by the occurrence of the RHI 8 .
We consider two critical problems that remain open.First, whether the attenuation of somatosensory inputs from the participants' hand to the brain occurs in areas where somatosensory inputs rst reach, the primary somatosensory cortex.Second, if the occurrence of the RHI is due to the attenuation of somatosensory inputs from the participants' hand to the brain, or the occurrence of the RHI causes the attenuation of somatosensory inputs to the brain.To address the rst issue, we recorded SEPs elicited by electrical stimulation of the peripheral nerve instead of the brush-evoked potentials used in multiple previous studies [4][5][6][7] .This is because the brush-evoked potentials do not include recognisable early components that are thought to be the origin of the primary somatosensory cortex [9][10][11][12] .To elucidate the second matter, we asked the participants to report the timing of the occurrence of the RHI.This allowed us to record SEPs before and after the occurrence of the illusion.This study was designed to elucidate the above two questions and to understand the neural mechanisms underlying the occurrence of RHI.

Experiment 1
Figure 1 shows the grand averaged waveforms of the SEP in all the conditions.The N1-P1 component was consistently recorded in all participants.It can be seen that the component showed an attenuation in amplitude in the congruent stroking condition.The number of average SEP in the control, congruent stroking, incongruent stroking, and tactile stimulation conditions were 37.2 ± 4.6, 34.8 ± 5.1, 33.1 ± 4.2, and 37.1 ± 4.4, respectively.
The amplitudes of the N1-P1 components in all the conditions are shown in Fig. 2. One-way repeated measures ANOVA revealed a signi cant main effect for the conditions (F(1.9,28.7)= 18.25, p = 0.0001, η 2 = 0.55).Post-hoc comparisons demonstrated that the N1-P1 component in the congruent stroking condition was signi cantly smaller than that in the control condition (p = 0.00004), incongruent stroking (p = 0.008), and tactile stimulation condition (p = 0.0002).The N1-P1 amplitude in the tactile stimulation condition was substantially smaller than that in the control (p = 0.055), although there was no signi cant difference between them.
Figure 3 demonstrates the frequency of electrical stimulation across all conditions.The interstimulus interval in all the conditions was approximately 10 s.One-way repeated measures ANOVA revealed no signi cant main effect for the conditions (F(3,45) = 2.1, p = 0.11, η 2 = 0.12).
Therefore, these questionnaire items were considered to be related to changes in the sense of body ownership.To calculate the degree of decrease in N1-P1 amplitude during RHI, the amplitude in the congruent stroking condition was normalised with respect to that obtained in the control condition.We compared the degree of decrease in the N1-P1 amplitude in the congruent stroking condition with ratings of questionnaire items 1, 2, and 3 by utilising the Spearman rank correlation coe cient (Fig. 5).There were no signi cant correlations between them (questionnaire 1: ρ=-0.093,p = 0.73; questionnaire 2: ρ = 0, p = 1; questionnaire 3: ρ=-0.041,p = 0.88).

Experiment 2
Figure 6 shows the grand averaged waveforms of SEP in the control condition, and during the pre-and post-RHI periods.It can be seen that the N1-P1 component showed an attenuation in amplitude during the pre-and post-RHI periods.The averages for SEP in the control condition, and during the pre-and post-RHI periods were 28.7 ± 8.3, 26.7 ± 4.0, 27.1 ± 7.3, respectively.
The amplitudes of the N1-P1 components in all conditions are shown in Fig. 7. One-way repeated measures ANOVA revealed a signi cant main effect for the conditions (F(2,26) = 8.39, p = 0.002, η 2 = 0.39).Post-hoc comparisons demonstrated that the N1-P1 component during both the pre-and post-RHI periods was signi cantly smaller than that in the control condition (pre RHI: p = 0.037, post RHI: p = 0.012).There was no signi cant difference in the N1-P1 amplitudes between the pre-and post-RHI periods (p = 0.74).

Discussion
In this study, we investigated the modulation of SEPs during the RHI.Compared with previous studies that showed attenuation of SEP components with a wide time window after stroking during the RHI 7,8 , our data indicates that the early component of SEPs, N1-P1 response, was attenuated when RHI was elicited.In addition, this attenuation began before participants felt the rubber hand as their own.This suggests that attenuation of somatosensory inputs from the participants' hand to the brain is one of the factors associated with the occurrence of the RHI.

SEPs have been typically recorded by continuous electrical stimulation with short interstimulus intervals
(1-2 Hz) [13][14][15][16][17][18] .Continuous electrical stimulation would direct participants' attention to their real hand.This situation makes it di cult to create an RHI.Therefore, in this study, electrical stimulation was provided manually by the experimenter with a long interstimulus interval (around 10 s, Fig. 3).This allowed the participants' attention to be distracted from their hand.However, the long interstimulus interval does not ensure a large number of waveforms triggered by electrical stimulation as the experiment time becomes longer and the physical or psychological load of the participants' increases.
Thus, we adopted a special arrangement of recording electrodes introduced by Brooke et al. 19,20 .A merit of this arrangement is that it records a clear waveform of the early components of SEP with a small sample size.
We considered that modulations of the early component of SEPs re ect the degree of ease in transmission of somatosensory information from the hand to the cerebral cortex.Previous studies used brush-stroke stimulation to evoke SEPs during RHI [4][5][6][7][8] .The obtained SEPs were derived from tactile stimulation that is necessary to produce the RHI.However, the demerits of brush-stroke-evoked SEPs does not include the recognisable early component, in particular components originating from the primary somatosensory cortex [9][10][11][12] .To address this issue, we used electrical stimulation.One might think that in ow of somatosensory inputs induced by electrical stimulation to the cerebral cortex does not originally occur during procedures in the RHI.Although somatosensory in ow produced by electrical stimulation is unnatural, we believe that modulation of the N1-P1 amplitude is the optimal index for evaluating the in ow of somatosensory inputs to the cerebral cortex.
The N1-P1 component originates from the primary somatosensory cortex.Desmedt 21 demonstrated that this component may represent the primary response of somatosensory cortical cells.SEPs recorded at the cortical surface during neurosurgery indicated that early components, 20-30 ms after electrical stimulation, are generated from the primary somatosensory cortex 22 .In addition, the modulation of the SEP amplitude depends on the frequency of the electrical stimulation 23 .In this study, the frequency did not differ among the four conditions (Fig. 3).Therefore, this is not the cause of the modulations of the N1-P1 amplitude in this study.These suggest that modulations of N1-P1 amplitude re ect alterations in somatosensory inputs at the primary somatosensory cortex according to the process of occurrence of the illusion.
The transmission of somatosensory signals to the primary somatosensory cortex is diminished during active or passive movements, and tactile stimulation of the hand [24][25][26][27][28][29][30] .This mechanism is called "gating".The gain in the SEP amplitude is modulated by centrifugal and centripetal gating mechanisms 27 .The former is when the efferent signals induced by the motor command from the motorrelated areas suppress the ascending somatosensory signals.The latter corresponds to when interfering effects between the given sensory afferent signals induced by electrical stimulation of the nerve and the afferent feedback from the skin caused by tactile stimulation of the hand.In this study, the participants remained resting throughout the experiment.Therefore, the effect of the centrifugal gating mechanism on the modulations of the N1-P1 amplitude need not be considered.
However, we believe that the decrease in the N1-P1 amplitude during RHI is not explained only by traditional centripetal gating mechanisms.In Experiment 1, the N1-P1 amplitude in the tactile stimulation condition was substantially smaller than that in the control condition.This nding would be affected by centripetal gating 24 .Similar tactile stimulations were provided to the participants in both the congruent and incongruent stroking conditions.Despite the same manner of the tactile stimulation, N1-P1 amplitude in the congruent stroking condition was signi cantly attenuated compared with those in the incongruent and tactile stroking conditions (Fig. 2).Thus, attenuation of the N1-P1 amplitude during RHI would not be caused by only centripetal gating.We do not know the mechanisms for this modulation, but certain mechanisms related to the occurrence of the illusion might affect the modulation of the transmission of somatosensory signals.
The subjective ratings for the questionnaire are typically used as an index for evaluating changes in the sense of body ownership 1 .In this study, ratings of questionnaire items 1, 2, and 3 in the congruent stroking condition were high, which is consistent with previous studies 1,3,31,32 .This indicates that our procedure properly induced the RHI.The degree of decrease in the N1-P1 amplitude in the congruent stroking condition was not signi cantly correlated with subjective ratings of the questionnaire (Fig. 5).Therefore, the attenuation of the N1-P1 amplitude relates to the occurrence of the illusion, but does not re ect the subjective strength of the illusion itself.
In Experiment 1, the electrical stimulation was continuously provided at intervals of approximately 10 s.In this case, the SEP waveform might be obtained from the electrical stimulation given before and after the occurrence of the illusion.Therefore, the ndings of the experiment were not able to explain whether the occurrence of the RHI was due to the attenuation of somatosensory inputs from the participants' hand to the brain, or the occurrence of the RHI causes the attenuation of somatosensory inputs to the brain.To address this, we performed Experiment 2 and found that the N1-P1 amplitude was attenuated before the occurrence of the illusion.We consider that the decrease in the transmission of somatosensory signals at the primary somatosensory cortex is one of the factors in uencing the occurrence of the RHI.
According to a model of body ownership during RHI 33 , the posterior parietal cortex integrates visual and somatosensory information of touch before the occurrence of the RHI.This suggests that the posterior parietal cortex is involved in the resolution of the con ict between the incoming visual and tactile information, and the resulting recalibration of the visual and tactile coordinate systems.When multiple sensory information integrate in the parietal cortex before the illusion occurs, it might be necessary to be accompanied by attenuated somatosensory signals from the hand.
A limitation of this study is that we cannot distinguish whether the N1-P1 response originates from cutaneous afferent or muscle afferents.Modulation of cutaneous afferent inputs is meaningful in the RHI paradigm.The median nerve is a mixed nerve containing both muscle and cutaneous afferents, implying that any changes in the N1-P1 amplitude may not be solely attributable to only one of these groups of afferents.Gandevia and Burke 34 demonstrated an intramuscular and percutaneous mixed nerve trunk stimulating technique, in which muscle afferents contributed to the recorded N1-P1 potential.In this study, we did not record N1-P1 potentials elicited by stimulation of only cutaneous afferents (e.g.stimulating the nerve of the digit).This is because the responses obtained by the stimulating nerve of the digit are small and require a large number of sample sizes 35,36 .Therefore, this study suggests that somatosensory signals to the brain are diminished during RHI, although the modality of afferents is not speci ed.Further studies are needed to solve this problem.
In summary, our results suggest that attenuation of somatosensory inputs from the participants' hand to the brain occurs at the primary somatosensory cortex during RHI.Furthermore, the attenuation starts before the occurrence of the illusion.We consider that attenuation of somatosensory signals from the body parts at the entrance of the cerebral cortex is one of the factors associated with the occurrence of changes in feelings of limb ownership.This study has gone some way toward enhancing our understanding of the neural mechanisms underlying the occurrence of the RHI.

Participants
Thirty male volunteers aged 20 to 24 years, naïve to the purpose of the experiments, participated in this study.Sixteen participants participated in Experiment 1, and the remaining 14 participated in Experiment 2. All participants had normal ndings on physical and neurological examinations and provided written informed consent.This study was approved by the Human Research Ethics Committee of the Faculty of Education, Kumamoto University.The experiments were conducted in accordance with the Declaration of Helsinki.
Recording Electroencephalographic (EEG) signals for determining SEP were recorded from C4' (2 cm behind C4) referenced to Fpz' (2 cm caudal to Fpz) in accordance with the international 10-20 system.This arrangement of recording electrodes was introduced by Brooke et al. 19 , and was reported to successfully determine SEP during pedaling [37][38][39] and sustained nger muscle contraction with fatigue 36 .Vertical and horizontal electrooculograms (EOGs) were also recorded above and below the right orbital fossa.The EEG and EOG signals were ampli ed and ltered bandwidth of 5-100 Hz and 0.5-120 Hz, respectively.
All data were stored on a hard disk with a sampling rate of 1 kHz.

Electrical stimulation
The left median nerve was stimulated on the palm side of the wrist with surface Ag/AgCl disk electrodes (Ø 1.5 cm).The cathode was placed 2 cm proximal to the anode.The electrode was xed on the median nerve so as not to move during recording.Constant current square wave pulses (duration, 0.2 ms) were provided, and the intensity was adjusted to produce a slight twitch of the abductor pollicis brevis muscle.The inter-stimulus interval was approximately 10 s, which was manually controlled by the experimenter (see below).This was intended to prevent the participants from anticipating the timing of the electrical stimulation.

Experiment 1
The experiment was conducted in a dimly lit room.Participants sat on a chair throughout the experiment.Both the participants' left hand and the fake left hand (see below) wore identical light blue coloured rubber gloves to eliminate differences in appearance between them 40 .Participants put their left hand and forearm inside a wooden frame with the forearm in the prone position.A fake left hand constructed of rubber was placed in a prone position 19 cm medial to the participants' unseen left hand.The experimenter put black clothes on both the left upper arm of the participants and the forearm of the fake hand.Therefore, participants were able to see only the fake hand throughout the experiment.
Electrical stimulation was applied to the participants' left median nerve under one of the following four conditions: The congruent stroking condition The experimenter asked the participants to view the fake hand, and manually delivered tactile stimulation for 4 min with the use of two identical paintbrushes.At this time, both the participant's hand and the fake hand were stroked simultaneously and at the same location.During the congruent tactile stimulation, the experimenter applied electrical stimulation by pressing the foot switch at his foot.The interstimulus interval of the electrical stimulation was approximately 10 s, which was manually adjusted by the experimenter.

The incongruent stroking condition
The timing and location of stroking did not match between the participant's hand and the fake hand.Other procedures were the same as those in the congruent stroking condition.

The tactile stimulation condition
The fake hand was removed and a small cube with a side of 1 cm was placed where the fake hand was.The experimenter asked the participants to view the cube, and provided the tactile stimulation to the participants' left hand for 4 min.During the tactile stimulation, the experimenter applied electrical stimulation in a manner similar to that of the congruent stroking condition.

Control condition
The participants viewed the small cube that was the same as the used in the tactile simulation condition.Tactile stimulation was not provided to the participants' left hand for 4 min.However, only the electrical stimulation was applied to the participants' median nerve in a similar manner to that of the congruent stroking condition.
The four experimental conditions were repeated twice each in a random order.Between each condition, there was a resting period of 5 min.
After completing each condition, participants were also asked to answer the RHI questionnaire.The questionnaire consisted of eight statements that were adopted from Botvinick and Cohen's 1 original report.The questions were as follows: (Q1) it seemed as if I were feeling the touch of the paintbrush in the location where I saw the rubber hand touched, (Q2) it seemed as though the touch I felt was caused by the paintbrush touching the rubber hand, (Q3) I felt as if the rubber hand were my hand, (Q4) it felt as if my (real) hand were drifting towards the rubber hand, (Q5) it seemed as if I might have more than one left/right hand or arm, (Q6) it seemed as if the touch I was feeling came from somewhere between my own hand and the rubber hand, (Q7) it felt as if my (real) hand were turning 'rubbery', (Q8) it appeared (visually) as if the rubber hand were drifting towards my hand.The participants responded by choosing a value on a 10-point scale ranging from 1 to 10, with 1 corresponding to 'strongly disagree' and 10 to 'strongly agree'.

Experiment 2
In this experiment, only the control condition and the congruent stroking condition in Experiment 1 were performed.In the congruent stroking condition, the experimenter provided both tactile and electrical stimulations, as in Experiment 1.The participants were asked to press a button that was put under their right hand with the right hand when they felt the rubber hand as their own.This allowed the experimenter to know the timing of occurrence of the RHI 41 .The experimenter stopped the synchronous tactile stimulation about 20 s after the participants pressed the button.With a 10-20 s break, the experimenter provided both tactile and electrical stimulations to the participants until 20 s after pressing the button.These procedures were repeated 40-50 times.Other procedures were the same as those performed in Experiment 1.The two experimental conditions were conducted in random order.

Figures
Figures

Figure 1 Grand
Figure 1

Figure 3 Group
Figure 3