Methods .1 Participants

Seventeen participants aged 4 years (M = 4.55 years, SD = .36 years) took part in this study. Three males and 14 females took part. An additional 5

participants (all male) were excluded prior to analyses for failing to understand task instructions (3 participants) or repeatedly removing their hands out of the assigned posture (2 participants).

3.6.2 Apparatus and Materials

This study used the same apparatus as Experiment 1, however there were a few differences with additional equipment. For example, the small ledge on which the stuffed animals were attached was replaced with a larger ledge (measuring 32 cm by 42 cm by 11 cm) to accommodate the length of the fake rubber hands (see Figure 3.4). An additional length of faux fur (measuring 30 cm by 46 cm) was used to cover the ends of the rubber hands.

The experiment was conducted in a quiet room at either a nursery or at the Goldsmiths InfantLab. Although in the previous study it was found that children could not locate the tactors using only auditory cues (see Experiment 1), as an added precaution a centrally placed speaker playing grey noise was used to mask any sound the tactors made.

(A) (B) (C)

Figure 3.4: Experimental set up of Experiment 2. Panel A shows the No Rubber Hands condition. Panel B shows the Uncrossed Rubber Hands condition and Panel C shows the Crossed Rubber Hands condition. Participants’ hands are beneath the black cover.

3.6.3 Procedure

This experiment used the same procedure for the practice phase as in Experiment 1 (see Section 3.2.3). The participant's hands were placed on the buzzers in the uncrossed posture, underneath the black covering cloth.

Once again, each participant was required to correctly locate the stimulus on 5 consecutive trials in order to continue to the experimental trials.

The experimental trials took a similar format to the practice trials except that the experimenter was blind to the accuracy of the child’s answer during the experimental trials and some experimental trials included the presence of rubber hands. Participants’ verbal responses were recorded by the experimenter on a computer. In the experimental phase, there were six separate blocks of trials: (i) Uncrossed-hands, No rubber hands; (ii) Crossed-hands, No rubber hands; (iii) Uncrossed-hands, Uncrossed rubber hands; (iv) Crossed-hands, Uncrossed rubber hands; (v) Uncrossed-hands, Crossed rubber hands; and (vi) Crossed-hands, Crossed rubber hands. The order of the blocks was counterbalanced across participants, yielding 12 separate order conditions (see Appendix B).

Between blocks, the experimenter always moved the child’s hands into a different posture. Additionally, the children were asked to close their eyes between all blocks (both when the rubber hands were either introduced to, or removed from, the ledge and also between blocks in which there were no rubber hands). Each block consisted of 20 trials (10 vibrotactile stimulations to each hand), across which the order of left and

right stimuli was randomized. Thus, across the whole experimental session (6 blocks) there were 120 trials.

3.7 Results

A measure of the children’s tactile localisation accuracy was derived by calculating the percentage of trials on which they made a correct response in each condition. One-sample t-tests of this percentage accuracy score showed that participants were performing reliably above chance (50%) in all conditions, except the Rubber Hands Crossed / Real Hands Crossed condition where the difference between chance and performance was marginally significant (see Table 3.3 below).

Table 3.3: Results from one sample t-tests comparing children’s tactile localisation accuracy with 50% (chance performance) across experimental conditions.

Rubber hand Posture

Real hands Posture

t df p d

None Uncrossed 5.08 16 <.001 2.54

None Crossed 2.48 16 .024 1.24

Uncrossed Uncrossed 2.64 16 .018 1.32

Uncrossed Crossed 3.68 16 .002 1.84

Crossed Uncrossed 3.54 16 .003 1.77

Crossed Crossed 2.08 16 .054 1.04

I used a repeated measures 2 x 3 ANOVA to investigate the effects of the within-subjects factors of Posture (Uncrossed-hands / Crossed-hands) and Rubber Hand (None / Uncrossed / Crossed). This revealed a main effect of Posture [F(1, 16) = 4.59, p = .048, !^p2 = .22], showing that, across all conditions, children were more accurate when localizing tactile stimuli in the uncrossed-hands posture compared to the crossed-hands posture. Additionally, an interaction effect of Posture x Rubber Hand condition was also significant [F(2, 32) = 3.41, p = .046, !^p2 =

.1]. No other main effects were found.

As I had quite specific hypotheses to test, I conducted three planned paired sample t-tests looking at the effect of Posture in each of the No Rubber Hand and Rubber Hands Uncrossed conditions and so the alpha value was bonferroni corrected to p = .017. First I conducted a one tailed comparison looking at the effect of Posture on tactile localization in the No Rubber Hand condition; as expected I replicated the results of Experiment 1 - the children had poorer tactile localization accuracy in the Crossed-hands posture compared to the Uncrossed-hands posture [t(16) = 2.57, p = .003, d = .72], demonstrating the “crossed-hands” effect (see Figure 3.5). However, two further planned comparisons (also one-tailed) revealed that this effect was not present when participants viewed Uncrossed Rubber Hands [t(16) = .45, n.s.] or Crossed Rubber Hands [t(16)

= 1.08, n.s.] with means indicating that children were performing comparably across postures when they viewed either Uncrossed or

As I was particularly interested in whether viewing the Rubber Hands in an Uncrossed posture would disrupt children’s tactile localization accuracy when their hands were in that same uncrossed posture, I ran a one-tailed comparison between the Uncrossed posture in the Rubber Hand Uncrossed condition with the Real Hands Uncrossed condition in which participants did not have visual information of any hands (i.e. the No Rubber Hands condition). I found a significant difference in tactile localization accuracy between these conditions [t(16) = 2.32, p = .017, d = .51), with means indicating that children’s performance in the tactile localization task in Uncrossed postures was significantly impaired when they viewed Uncrossed Rubber Hands.

I now go on to a more exploratory analysis of the comparisons between other conditions in this study. First, I examined the role of seeing Rubber Hand (No Rubber Hands and Rubber Hands Uncrossed) on localization performance when the hands were Crossed (as in Azañón &

Soto-Faraco, 2007) and found no significant difference [t(16) = 1.43, n.s.]. I then investigated whether having visual cues to the hands in a novel posture affects performance, examining the effect of Crossed Rubber Hands (compared to hidden hands) on tactile localization in both Postures (Uncrossed and Crossed). The comparison between Crossed Rubber Hands and No Rubber Hands in the Crossed posture yielded no significant difference in performance [t(16) = .15, n.s.]. However, a trend towards decrease in localization performance between conditions in which participants observed Crossed Rubber Hands compared to No Rubber

Hands when their real hands were in the Uncrossed posture produced a marginally significant effect [t(16) = 1.87, p = .08].

Figure 3.5: Mean tactile localization accuracy (percentage correct) of 4-year-olds in the crossed-hands and uncrossed-hands posture across conditions where they viewed rubber hands in an uncrossed- or crossed-hands posture or no rubber hands. Error bars indicate the standard error of the mean.

3.8 Discussion

As hypothesized, this study found that, when sight of the hands is not available, 4-year-olds demonstrated the “crossed hands” deficit (i.e. were not as accurate at locating the site of a tactile stimulus in the crossed hands posture, relative to the uncrossed hands posture). This confirms the findings from the previous study (Experiment 1), demonstrating that children of 4 years use an external reference frame to code touches to the body. Further to this, a significant interaction of Posture and Rubber Hand was also found in this age group indicating that sight of fake hands modulates tactile localization accuracy: when participants had visual information of a pair of (fake) hands in the uncrossed posture, the “crossed hands” effect disappeared.

What was interesting about the reduction of the effect of posture when participants could see uncrossed rubber hands, was that rather than this being the result of improved performance in the crossed posture (as was found in a study of adults by Azañón & Soto-Faraco, 2007), it was the result of a reduced accuracy in the uncrossed posture. This is consistent with the findings of Experiment 1, in that when children see hands in the uncrossed posture, tactile localization accuracy in this posture decreases.

In Experiment , I argued that this pattern of results was due to the fact that children of 4 years demonstrate visual interference when they see their arms. I proposed that when 4-year-olds have current sight of hand posture (specifically the uncrossed posture), they no longer rely on

prior information regarding the canonical posture of their bodies, and as a result of this, they do not benefit from having their hands in this posture.

It has been argued that this also means that young children are unable to incorporate visual cues of their own body within their body schema (Begum Ali et al., 2014). The current study extends this finding, in showing that even artificial hands placed in the uncrossed posture disrupt the advantage children would usually have when their hands are in anatomical locations in space. It appears that visual information specifying the position of the hands, rather than benefitting performance as it does in adults (Longo & Haggard, 2011; Longo et al., 2011) actually impairs it in early childhood.

A surprising element of this disruption is that it occurred without ownership of the limb being manipulated i.e. without experimental conditions in which the researchers prompted children to adopt the artificial hands as part of their own bodies. When designing the study, it was decided not to induce the rubber hand illusion by way of contingent visual-tactile information. This was decided in order to avoid any carry over effects (i.e. the effects of the contingent visual-tactile stimuli could affect not only the condition that immediately followed this stimulation, but other conditions as well) between conditions. Carryover effects would make it difficult to tease apart potential explanations for different patterns of findings.

Additionally, a further reason for not inducing the rubber hand illusion was in order to make the current study more comparable with

Azañón and Soto-Faraco (2007), as those researchers did not induce the rubber hand illusion, via contingent visual-tactile stimulation, with their adult participants.

There are several potential explanations for why seeing uncrossed rubber hands disrupted the 4-year-olds' performance even when the rubber hand illusion was not induced. It is possible that in this young age group, children are more susceptible to accepting artificial limbs as their own and do not need to see and feel contingent stroking of their own and the rubber hand. Given the main effect of age group in Cowie et al. (2013), which suggests that children experienced the RHI, regardless of whether they received contingent visual-tactile stimuli or not, this is a plausible explanation.

Most likely, and most consistent with the findings from Experiment 1 and the current experiment, is that children are especially captured by visual information of the limbs, be that their own or artificial limbs. This is an assertion that has been demonstrated throughout the psychological literature from various areas (Hay, 1979; Ferrel-Chappus et al., 2002;

Smyth et al., 2004; Bremner et al., 2013).

Thus, it is my proposal that, at this young age, viewing uncrossed rubber hands prompts children to use representations of their current body posture, rather than the heuristics provided by the canonical body representation. Therefore, children are not bestowed with the enhancement in tactile localization accuracy that is specific to the uncrossed hands posture (due to the limbs being in a canonical location).

Without this enhancement in accuracy with hands uncrossed, the difference in performance across postures is dramatically reduced and the crossed hands deficit is no longer evident.

So a question for future research relates to the causes of this visual interference. Is visual interference the result of ownership of the hand or are visual cues to hand position sufficient to result in this interference?

Either of these explanations are plausible, given that both ownership of the limb and visual capture of hand position have been demonstrated by Cowie et al. (2013) in this age group. However, considering that children at 4 years of age are still relatively immature at using visual cues to locate the limbs, with this ability developing throughout infancy and childhood (see Cowie & Bremner, 2013), it may be that the answer to the cause of visual interference is not straightforward.

This also relates to the wider question of when children begin to successfully integrate visual cues of the limbs (either their own or artificial limbs) to locate the limbs in space, thus deriving a benefit from viewing uncrossed rubber hands whilst their own were in the crossed posture. A series of studies that could be conducted to further investigate the effect of vision of limbs could be to manipulate ownership over the rubber hands by inducing the rubber hand illusion. This would be interesting across all variations of posture (of both real and artificial hands), but particularly interesting in conditions in which participants’

real hands are in a posture that is incongruent to that of the artificial hands and the subsequent effect the illusion would have on tactile

localization accuracy. Again, investigating the developmental trajectory and the pattern of the crossed hands deficit would be especially interesting in terms of determining the progress of developing integration of visual cues to the limbs.

In summary, this study has confirmed that by 4 years of age, children use an external frame of reference in which touches to the body are coded. This is in accordance with previous research (Experiment 1;

Begum Ali et al., 2014). Further to this, it has been shown that when visual information of the uncrossed posture is provided (via artificial limbs), children demonstrated decreased tactile localization accuracy when their own hands were in the corresponding posture. It seems that when visual cues are provided, this disrupts the benefit children would gain from their limbs being in a canonical posture. This is the case when children either see their own hands (as in Experiment 1) or artificial hands. Therefore, this suggests that visual cues to hand position interferes with a representation of the layout of the body in external space.

The last two experiments have shown how current visual information of a limb (whether a participant’s limb or an artificial one) can modulate tactile localization accuracy across different arm postures.

Additionally, these experiments have shown that children as young as 4 years of age are able to code touches within an external frame of reference.

This is a younger age than that found by Pagel et al. (2009), however as previously stated, Bremner et al. (2008) have demonstrated that infants within the first year of life demonstrate the crossed-hands effect and thus

use an external reference frame to map touches on the body. The next study investigates exactly when an external reference frame emerges in infancy.

3.9 Experiment 3: The emergence of an external