School of Psychological Sciences

Transcription

1 A mixed method investigation of embodiment using the Rubber Hand Illusion A thesis submitted to The University of Manchester for the degree of Doctor of Philosophy in the Faculty of Medical and Human Sciences Elizabeth Lewis School of Psychological Sciences

2 2 List of Contents List of Tables... 5 List of Figures... 7 List of Abbreviations... 9 Abstract Declaration Copyright Statement Acknowledgements Chapter 1: General introduction Multisensory processes of embodiment in illusory and veridical contexts Embodiment and embodied perception The rubber hand illusion and embodiment change Bottom-up processes Top-down influences Interaction between top-down and bottom-up processes The role of individual sensory modalities Unique contributions of individual modalities to embodiment in the Rubber Hand Illusion The contribution of tactile signals in the Rubber Hand Illusion Multiple sensory outcomes of the rubber hand illusion Summary Research aims and overview of the Thesis Chapter 2: Development of subjective and perceptual measures Study 1: A qualitative investigation of embodiment in contextual variations of the Rubber Hand Illusion Introduction Method Results Discussion Study 2: Measuring multiple sensory outcomes of embodiment change using a simplified augmented reality version of the Rubber Hand Illusion Introduction Study A: Method Results Discussion... 66

3 Study B: Method Results Discussion of studies A and B Summary of studies one and two Chapter 3: Measuring veridical changes to embodiment using multiple sensory outcomes of the Rubber Hand Illusion Study three: The effect of Body Scan Meditation on embodiment Introduction Method Results Discussion Study four: the effect of anatomical dissection training on embodiment Introduction Method Results Discussion Summary of studies three and four Chapter 4: Development of the Embodiment Change Questionnaire Study 5: A psychometric analysis of the Embodiment Change Questionnaire Analysis one Method Results Discussion Analysis two Method Results Discussion Analysis three Method Results Discussion Study 5 discussion Chapter 5: General Discussion Review of findings and implications

4 4 5.2 Limitations Limitations of the RHI methodology Limitations of the body perception training methodology Future directions Conclusions References Appendix A Appendix B Appendix C Word count: 58,250

5 5 List of Tables Table 2-1: Items in the RHI questionnaire Table 2-2: Items in the Invisible Hand Questionnaire Table 2-3: Nine contextual variations of the RHI made through combination of visual and tactile stimuli. Visual stimuli presented as pre-recorded videos and tactile stimuli positioned either under or over participant s hand. Table displays a description of the visual-tactile discrepancy or consistency in each contextual variation. Participants completed each variation twice, once with synchronous stroking and once with asynchronous stroking to their own hand Table 2-4: RHI Questionnaire items Table 2-5: Mean and standard deviation scores for visual, tactile and proprioceptive judgements before and after the illusion and change scores within these modalities. Scores in the visual modality reflect the mean proportion of rubber hand judged to be like the participant s own hand, higher values indicate a larger proportion of rubber hand. In the tactile modality higher values indicate vibrotactile stimuli of a higher intensity. Scores in the proprioceptive modality reflect the position in which the participants located their index finger, higher scores indicate more leftwards positions closer to the location of the rubber hand Table 3-1: RHI Questionnaire items Table 3-2: Demographic information for participants in session 2 for each group (BSM vs. AC) separately and combined Table 3-3: One sample t-tests assessing changes to visual, proprioceptive and tactile sensory judgements in sessions 1 and Table 3-4: Means and standard deviations for the illusion score and HBD accuracy for each session (one vs. two) and each training group (AC vs. BSM) Table 3-5: Means and standard deviations of visual, proprioceptive and tactile change scores in sessions 1 and 2 for the BSM and AC groups Table 3-6: Means and standard deviations of long-term effects in visual, proprioceptive and tactile sensory modalities for the AC and BSM groups... 89

6 6 Table 3-7: Comparison of means and standard deviations of Personal Distress and Empathic Concern subscales between student type (medical vs. life sciences) and session (one vs. two) Table 3-8: Comparison of the means and standard deviations of questionnaire scores (illusion vs. control) for each training type (medicine vs. life sciences) and level of personal distress (high vs. low) in each session (one vs. two) Table 3-9: Comparison of the means and standard deviations of immediate (sessions one and two) and long-term visual effects of the RHI for each training type (medicine vs. life sciences) and level of personal distress (high vs. low) Table 4-1: Polychoric correlation matrix and mean values of each item in the Embodiment Change Questionnaire Table 4-2: Structure of factor loadings and correlations of the Embodiment Change Questionnaire: CFA of the three-factor model Table 4-3:Polychoric correlation matrix and mean values of each item in the revised Embodiment Change Questionnaire Table 4-4: Structure of factor loadings and correlations of the revised Embodiment Change Questionnaire: CFA of the three-factor model Table 4-5: Items summed for the Body Ownership, Body Extension, Perceived Causality, Embodiment Change factors and the standard illusion score Table 4-6: Spearman's correlations between factor scores from the one and three factor model and the standard illusion score and Visual, Tactile and Proprioceptive sensory measures of the RHI in session one (N=103) Table 4-7: Spearman's correlations between factor scores from the one and three factor model and the standard illusion score and Visual (N=90), Tactile (N=91) and Proprioceptive (N=91) sensory measures of the RHI in session two Table 4-8: Spearman s rank-order correlations between subjective questionnaire scores and subscales of the IRI (N=103)

7 7 List of Figures Figure 1-1 Schematic representation of the Rubber Hand Illusion. The participant (S) watches a fake rubber hand (FH) being stroked by the experimenter (E) whilst the experimenter strokes the participant s real hand at the same time, in the same anatomical location. During this procedure, the participant cannot see their real hand because it is hidden by a partition (P) Figure 2-1: Overhead view of the experimental set up. Pre-recorded videos of the RHI were displayed via a HMD. When the video was played, the experimenter stroked the participant s hand using an audio click track in the video as a guide for timing and duration Figure 2-2: Graph displaying the mean scores for illusion and control items during synchronous and asynchronous stroking for each participant Figure 2-3 Schematic representation of the augmented reality RHI. The line drawing shows the participant s position relative to the video image presented through the HMD. A snap-shot of the video is presented as an example of the video image seen during the illusion. The table seen through the HMD (i.e. in the video) is positioned so that the rubber hand protrudes over the table edge. The participant s hand is positioned 10cm rightwards of the rubber hand seen though the HMD, and the participant s index finger rests on a tactor inlayed into the table surface Figure 2-4: Figure 2-5: Examples of photo stimuli presented in visual (left) and proprioceptive (right) outcome measures. In the visual task participants were presented with images made by blending a photograph of the participant s hand and the rubber hand seen during the RHI. Proportion of rubber hand in each image (left to right): 100%, 75%, 50%, 25% and 0%. In the proprioceptive task participants were shown images of an arrow marker on a table. The distance in millimetres between arrow marker and the participant s index finger in each image (left to right): 160, 100, 0, -100, Figure 3-1 Study design and procedure Figure 3-2: Scatterplots of the long-term visual effect of the RHI and changes to HBD accuracy. These scatterplots show that in the BSM group there was a negative correlation between improvements in HBD accuracy and the long-term visual effect

8 8 of the RHI (r 2 =26.52), but no relationship between these variables in the AC group Figure 3-3: Study design and procedure Figure 3-4 Bar chart comparing mean % of rubber hand in images judged to be like own hand at each time point for each training type (Medicine vs. Life Sciences) and level of Personal Distress (PD; High vs. Low). For all participants mean proportions increase after the RHI in session one (time points 1 and 2) and two (time points 3 and 4). Students studying Medicine who are high in PD show greater changes across each time point, showing greater increases between time points 1 and 2 and between time points 2 and 3, than the other groups Figure 3-5: Bar chart displaying means and standard errors of HBD accuracy change score for Medicine and Life sciences students with high and low trait personal distress. Whereas most participants showed a slight increase in HBD accuracy in session two, medical students who were high in trait personal distress showed a decrease in HBD accuracy following the training interval Figure 4-1: Path diagram of three-factor model of Embodiment Change tested using confirmatory factor analysis. Note. All items begin It seemed as though Figure 4-2: Path diagram of the updated three-factor model of Embodiment Change tested using confirmatory factor analysis. Note. All items begin It seemed as though. Item 10 from first analysis excluded. Item 3 (highlighted in bold) replaced with less ambiguous item I could not separate the experience of my own hand from my experience of the rubber hand

9 9 List of Abbreviations AC Anatomy Control AIC Akaike's Information Criterion AR Augmented Reality BIC Bayesian Information Criteria BSM Body Scan Meditation CDS Cambridge Depersonalisation Scale CFA Confirmatory Factor Analysis CFI Comparative Fit Index EC Empathic Concern ECQ Embodiment Change Questionnaire Embodiment Change Questionnaire- ECQ-R Revised EEG Electroencephalogram fmri Functional magnetic resonance imaging FS Fantasy Scale HBD Heartbeat Detection task HMD Head Mounted Display IPA Interpretative Phenomenological Analysis IRI Interpersonal Reactivity Index ML Maximum Likelihood PCA Principle Component Analysis PD Personal Distress Parameter Estimation by Sequential PEST Testing PT Perspective Taking RHI Rubber Hand illusion Root Mean Square Error of RMSEA Approximation SRMR Standardized Root Mean square Residual VR Virtual Reality WLMSV Weighted Least Square Estimator

10 10 Abstract A mixed method investigation of embodiment using the Rubber Hand Illusion Elizabeth Lewis, The University of Manchester for the degree of Doctor of Philosophy (PhD) September 2015 Embodiment is the experience of one s own body. It is often studied using the Rubber Hand Illusion (RHI). This illusion varies the consistency between visual, tactile and proprioceptive signals to elicit a change to embodiment. Changes to embodiment are typically measured using a single sensory outcome measure of proprioceptive drift, which is interpreted as a proxy measure of embodiment. This approach obscures the unique contribution of other modalities such as vision and touch. The work presented in this thesis uses a mixed method approach to investigate the unique contribution of visual, tactile and proprioceptive modalities within the multisensory process of embodiment. In study one, a qualitative analysis showed that when visual-tactile discrepancies were present in the RHI, participants described both body ownership and body extension type changes to embodiment, and changes to tactile perception. In study two, psychophysical measurements of the RHI showed changes to visual, tactile and proprioceptive aspects of embodiment, suggesting that embodiment in the RHI could be measured using multiple sensory outcomes. Studies three and four assessed the utility of measuring multiple sensory outcomes of the RHI, by exploring changes to embodiment following internal and external forms of body perception training. Study three showed that brief body scan meditation, as a form of internal body perception training, reduced the longevity of the visual sensory outcome of the RHI and that this reduction was negatively correlated with improvements in interoceptive sensitivity. Study four showed that learning about the body through anatomical dissection training, as a form of external body perception training, reduced the longevity of the visual sensory outcome measure and decreased interoceptive sensitivity, but only in medical students who were high in trait personal distress. Collectively, these findings suggest that aspects of the multisensory processes of embodiment can become specialised and identify some unique contributions of individual sensory modalities to embodiment. The proprioceptive sensory outcome appears to be stable over time but the visual sensory outcome is a longer-term change to embodiment, which is susceptible to interference from body perception training. In study five, confirmatory factor analysis was used to assess the psychometric properties of an embodiment change questionnaire measuring body ownership, body extension and perceived causality in the RHI. Factor scores from the questionnaire were correlated with visual and proprioceptive outcome measures of the RHI and measures of trait empathy. The results suggested factor scores had better convergent validity than the standard illusion score used in previous research. This work has improved subjective and perceptual measures of the RHI and specified ways that individual sensory modalities provide a unique contribution to embodiment. The methods developed have further applications for studying the multisensory process of embodiment and investigating embodiment in a number of clinical groups.

11 11 Declaration No portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning.

12 12 Copyright Statement i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the Copyright ) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trade marks and other intellectual property (the Intellectual Property ) and any reproductions of copyright works in the thesis, for example graphs and tables ( Reproductions ), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see in any relevant Thesis restriction declarations deposited in the University Library, The University Library s regulations (see and in The University s policy on Presentation of Theses.

13 13 Acknowledgements First of all I would like to thank my supervisors, Ellen Poliakoff, Emma Gowen and Luke Jones for their encouragement and support. I am extremely grateful for all of their advice and expertise during the development of the project and thesis writing. I would also like to thank my advisor Christine Rogers for having such confidence in me for so many years. With her support I have been able to achieve much more than I realised I was capable of. I would also like to thank Josh Bradley for helping to collect data at the University open day for study five. I am very thankful to my friends and family for being so understanding and supportive during the writing of my thesis. I am especially grateful to my Mum and Sister Charlotte. Their love and support has been an essential part of this entire process. I would like to make a special mention of Donna Lloyd and Martin Farrell whose passion inspired me to pursue this project. They are for me the embodiment of everything I love about academia and I feel blessed for their support and friendship. I am very grateful to many friends but I want to make special mention of Lauren Reilly, Kylie Abplanalp, Helen Sutherland and Chris Royle for much patience and many good times, and my housemate Sarah Hopkins whose energy, support (and coffee) makes every day the best. I would like to thank all of my fellow PhD students for their enthusiasm and friendship. Many thanks to Stacey Humphries, Jenna (Ginibu) Roberts, Dan Poole, Oana Linder, Gemma Barnacle, Adam Laurence, Sam Couth, Gabriel Davies, Cliff Workman and Michelle Hall. They have helped me in so many ways and brought fun and excitement to everyday, especially Thursday. They have all contributed to making my PhD the most interesting and happiest time of my life. Finally, I would like to say thank you to Georgie Farmer for daily support, distraction and laughter.

14 14 1 Chapter 1: General introduction Multisensory processes of embodiment in illusory and veridical contexts The work presented here uses qualitative, psychophysical and questionnaire measures to investigate the unique contribution of visual, tactile and proprioceptive modalities within the multisensory process of embodiment. Embodiment arises from the integration of a number of body signals, however, the methods used to investigate embodiment do not assess the unique contribution of individual modalities within this multisensory process. In this thesis, existing methods for the study of embodiment are further developed so that the role of individual body signals can be specified in more detail. To evaluate the utility of this new approach, it was applied to investigate how visual, tactile and proprioceptive modalities contribute to the multisensory process of embodiment over time. In this chapter, current methods for investigating the multisensory process of embodiment are introduced in section 1.1. Section 1.2 considers methodological influences on the current conceptualisation of embodiment. Further methodological developments are considered and an empirical test of the utility of these developments is proposed. In section 1.3, the aims and methodology of the research presented in this thesis are outlined. 1.1 Embodiment and embodied perception The term embodiment refers to the experience of one s own body, including the experience of what one s own body is, or body ownership and what it is like (Longo, Schüür, Kammers, Tsakiris and Haggard, 2008; De Vignemont, 2010). The experience of having a body that is separate from the world and uniquely one s own is considered a fundamental aspect of self-awareness (Gallagher, 2000). The numerous phenomenological explorations of the experience of embodiment are a testament to the complexity of this experience (e.g. Merleau-Ponty, 2002,,, Leder, 1990). This complexity of embodiment arises in part because it is always present as a continuous feature of all perceptual experiences, that same old body always there (James, 1890, p.242). Yet despite this constant presence it is rarely the focus of attention, and in this sense, the typical experience of one s own body is one of absence from perceptual experience. Leder describes this self-forgetting as intrinsic

15 15 to body function, allowing attention to be focussed upon the external world and returned to the body only in times of need. When an issue with the body arises, such as illness or injury, pain signals bring the body into the focus of attention so that the cause of the issue can be resolved. Leder describes these instances as the body dysappeared, because the body is away from its usual state of marginal awareness (p.69-99). The importance of the body has often been overlooked in cognitive science but more recent theoretical approaches such as embodied cognition recast the role of body as central to cognition, perceptual and social experience. There are a number of different approaches to embodied condition that differ with respect to the extent that internal representations feature in their explanations of cognitive abilities (Clark, 1999). Nevertheless, they are unified by the notion that the body is a core influence structuring perceptual experience of the external world (Glenberg, Witt, and Metcalfe, 2013). This is demonstrated in numerous studies showing that physical properties of body influence the way that the external world is perceived. Perceived distance is affected by the energy required to travel a distance. For example, the same distance looks longer when tired, when wearing a heavy backpack, or when of low physical fitness (Proffitt, 2006). Changes to the perceived size of the body, using virtual reality or optical magnification alters perceived physical characteristics of objects in the environment, such that objects are perceived to be smaller in size when the body is perceived to be larger (Van der Hoort, Guterstam, and Ehrsson, 2011; Linkenauger, Ramenzoni, and Proffitt, 2010; Linkenauger, Witt, and Proffitt, 2011). Embodied approaches further describe the role of the body in social perception (Decety and Sommerville, 2003). A number of studies demonstrate that the perception of one s own body and the perception of other people s bodies are supported by overlapping neural mechanisms, for example, there is a common neural basis for feeling disgust and seeing expressions of disgust (Wicker, Keysers, Plailly, Royet, Gallese and Rizzolatti, 2003). These overlapping mechanisms are thought to support affect sharing such that by simulating another s emotional behaviours it is possible to vicariously experience their internal states (Lamm and Singer, 2010). The importance of the body for perception of environment and social world highlights the need to understand how the body itself is perceived.

16 The rubber hand illusion and embodiment change Recent approaches to the study of embodiment use body illusions to bring the body to the foreground of awareness so that it can be measured and described. A number of body illusions have been reported but the most commonly used is the Rubber Hand Illusion (RHI; Botvinick and Cohen, 1998). In this illusion, participants watch a rubber hand being stroked whilst their own hand, which is hidden a short distance away, is stroked at the same time. In this situation participants see a rubber hand being touched and they feel a tactile sensation at the same time in the same location on their own body. A schematic representation of the RHI is shown in figure 1.1. The co-occurrence of seen and felt touch leads participants to report feeling as though the rubber hand is their own hand. The description of subjective body ownership indicates that a change to embodiment has occurred because an external object has become incorporated into the participant s experience of their own body. A number of outcome measures show that following the illusion participants perceive the rubber hand as they would their own body. When the rubber hand is threatened participants show a skin conductance response (Armel and Ramachandran, 2003) and a neural response associated with anxiety, such as increase in responses in the insula and the anterior cingulate cortex (Ehrsson Weich, Weiskopf, Dolan and Passingham, 2007). Further, there are changes in the homeostatic regulation of the real hand such that participants have an increased histamine response (Barnsley, McAuley, Mohan, Dey, Thomas and Moseley, 2011) and reduction in skin temperature (Moseley et al., 2008). Though it is unclear precisely what these physiological changes indicate about embodiment, they demonstrate that a change to physiological regulation of the self has occurred. Collectively, these results indicate that the RHI is a useful method for the manipulation of embodiment. Using this manipulation, it is possible to investigate the multisensory process of embodiment including bottom-up sensory processes and top-down influences of body representations

17 17 Figure 1-1 Schematic representation of the Rubber Hand Illusion. The participant (S) watches a fake rubber hand (FH) being stroked by the experimenter (E) whilst the experimenter strokes the participant s real hand at the same time, in the same anatomical location. During this procedure, the participant cannot see their real hand because it is hidden by a partition (P) Bottom-up processes The RHI is a clear demonstration that the experience of embodiment arises from the integration of a number of body signals, particularly vision and touch (Tsakiris, 2010). The temporally co-incident seen and felt touch is a critical feature of the illusion. When the seen and felt touch are asynchronous, participants no longer report feeling that the rubber hand is their own hand (Botvinick and Cohen, 1998) and physiological changes are significantly lower (Barnsley et al., 2011; Moseley et al., 2008; Armel and Ramachandran, 2003). Asynchronous touch between the rubber hand and the participants hand is commonly used as a control condition in experiments using the RHI, so that instances of embodiment change and no embodiment change can be compared (Tsakiris, 2010). A further multisensory process can also be observed in the RHI. When there is a short distance between the location of the rubber hand and the participant s hand, participants perceive their hand to be closer to the location of the rubber hand after the illusion (Botvinick and Cohen, 1998). This is referred to as proprioceptive drift and is measured by covering

18 18 the RHI equipment and asking participants to locate their own index finger either by pointing or identifying its location on a visual marker or scale (for alternative measurement approaches see Rohde, Di Luca and Ernst, 2011). The subjective and proprioceptive changes to embodiment in the RHI are thought to reflect a three-way interaction between vision, touch, and proprioception, which maintains consistency across these senses (Botvinick and Cohen, 1998). The seen touch on the rubber hand captures the tactile sensation on the participant's own hand, and this visual capture results in a mis-localisation of the felt location of one's own hand towards the spatial location of the seen hand. fmri studies further support the role of multisensory integration for embodiment. Subjective and proprioceptive changes following the RHI are positively correlated with activity in brain regions shown to support multisensory integration such as the ventral portion of the bilateral premotor cortices, the anterior section of the left intraparietal sulcus (Ehrsson, Holmes and Passingham, 2005; Ehrsson, Spence and Passingham, 2004) and the right posterior insula (Tsakiris, Hesse, Boy, Haggard and Fink, 2007) Top-down influences Though multisensory integration of bottom-up sensory signals is considered to be necessary for the RHI to be elicited, there is less consensus over whether it is sufficient. Some researchers have argued that the statistical correlations between sensory signals is sufficient such that embodiment can be elicited for non-body objects simply using synchronous stroking. Armel and Ramachandran (2003) conducted the RHI but replaced the rubber hand with a table and reported that participants felt sensations from the table surface and showed a skin conductance response when the table was injured. Similarly, Press, Heyes, Haggard and Eimer (2008) observed enhanced somatosensory event-related potentials (specifically an enhanced N140) in response to tactile stimulation when comparing synchronous to asynchronous tapping of a rubber hand or a rubber object. Though these studies suggest that a sense of embodiment can be elicited purely through bottom-up signals, this has not been replicated in other studies using non-body objects such as a stick or wooden block,or when a rubber hand is presented in an anatomically implausible position such as rotated 180 degrees from the participant s hand orientation (Haans

19 19 Ijsselsteijn and de Kort., 2008, Tsakiris, Carpenter, James and Fotopoulou, 2009; Tsakiris and Haggard, 2005; Costantini and Haggard, 2007), or positioned >30cm away from the participants hand (Lloyd, 2007). These constraints on embodiment in the RHI demonstrate that external objects are incorporated when they match existing representations of the form and position of the body. Body representations exert a top-down influence such that current sensory body signals can be interpreted in the context of an internal model of the body's structure. Bottom-up sensory inputs are integrated according to a set of background conditions that maintain the coherence of bodily experience (Tsakiris, 2010). Body representations are typically divided into two forms: the body image and the body schema (Gallagher, 1986). The body image is a stable representation of form and characteristics of the body that are accessible to conscious awareness and form the basis of comparison between what one s own body is like in comparison to others. In contrast, the body schema is representation of body posture and location that updates online during movement providing information about where the body is in relation to the environment. Both of these representations are thought to influence embodiment change in the RHI. The body image constrains what type of objects can be incorporated during the illusion such that the object must resemble pre-existing knowledge about the shape of the body (Tsakiris et al., 2008; Tsakiris and Haggard, 2005; see also Graziano, Cook and Taylor, 2000; Holmes, Snijders and Spence, 2006; Haans et al., 2008; Tsakiris, Carpenter, James and Fotopoulou, 2009). The body schema constrains the spatial characteristics of objects that can be incorporated during the illusion such that the object must be within peri-personal space, i.e. the area around the body perceived as being within reaching distance (Lloyd, 2007), and congruent with the current body posture (Costantini and Haggard, 2007) Interaction between top-down and bottom-up processes The presence of constraints in the RHI suggest that embodiment arises from the interaction between top-down representations of the body s expected characteristics and bottom-up sensory information about the body s typical characteristics. Sensory outcomes of the RHI are thought to indicate a change to body representations based on current sensory input. Changes to perceived location arguably reflect a change to

20 20 the body schema, which constitutes knowledge of body location and posture. Though, studies examining more explicit measures of the body schema, such as pointing and grasping movements, have produced inconsistent results regarding whether the illusion influences the accuracy of motor actions (Botvinick and Cohen, 1998; Kalckert and Ehrsson, 2012; Holmes, Snijders and Spence, 2006; Heed, Grundler, Rinkleib, Rudzik, Collins, Cooke and O Regan, 2011; Kammers, Kootker, Hogendoorn, Dijkerman, 2010; Kammers, de Vignemont, Verhagen, Dijkerman, 2009). Whereas visual changes such as increases in the perceived similarity of the rubber hand and the participant s hand are thought to indicate a change to the body image, which constitutes knowledge of the form and appearance of the body (Longo, Schüür, Kammers, Tsakiris, Haggard, 2009). Although these body representations are considered by many to be distinct (Dijkerman and de Haan, 2007; Gallagher, 2005; Head and Holmes, 1911), they are also thought to interact and so should be considered as different aspects of embodiment rather than a strict dichotomy (Gallagher, 2005; de Vignemont, 2010; Holmes and Spence, 2007). The magnitude of sensory outcomes of the RHI is often interpreted as indicating the flexibility or malleability of body representations, that is, the degree to which body representations are influenced by current sensory signals. Individuals with anorexia, bulimia and schizophrenia show larger subjective and proprioceptive effects of the RHI suggesting that they have more malleable body representations (Mussap and Salton, 2006; Eshkevari, Rieger, Longo, Haggard and Treasure, 2012; Thakkar, Nichols, McIntosh, Park, 2011). One factor influencing the malleability of body representations is the ability to perceive the body from within, referred to as interoception (Tsakiris et al., 2011; Tajadura-Jimenez and Tsakiris, 2013). Interoception is the awareness of internal body sensations and the physiological state of the body, including the perception of temperature, itch, visceral sensations and cardiac awareness (Craig, 2002). An individual s awareness of interoceptive sensations, specifically heartbeats, has been shown to be negatively correlated with the magnitude of RHI effects, such that, participants who report stronger feelings of ownership over the rubber hand and show larger proprioceptive changes have poorer heartbeat detection performance (Tajadura-Jimenez and Tsakiris, 2013). An individual s interoceptive abilities are thought to influence the degree to which their body representations are influenced by bottom-up sensory stimulation, such that

21 21 external sensory information about the body has a greater contribution to embodiment when their interoceptive perception is poor. The multisensory process of embodiment and individual differences due to interoceptive perception have been described using a predictive coding account (Seth, 2013) which will be outlined in chapter three. 1.2 The role of individual sensory modalities Current models emphasise the consistency maintained between individual sensory modalities as a critical feature of embodiment, however, individual modalities contribute different types of information about what the body is like. The unique contribution of individual modalities is currently underspecified because of the way that embodiment is measured in the RHI. Evidence that individual modalities have a unique contribution to embodiment is presented in section The contribution of the tactile modality is explored in section Methodological developments that would allow the contribution of individual modalities to be measured independently are considered in section Unique contributions of individual modalities to embodiment in the Rubber Hand Illusion The RHI manipulates a number of body signals within a single experimental context. Some body signals are manipulated so that they are consistent, such as the cooccurrence of seen and felt touch used to elicit the illusion. Other body signals are manipulated to create discrepancies between the rubber hand and the participant s hand; for example, the rubber hand is seen in a different location, has a different appearance and a different texture to the participant s hand. Each of these discrepancies feature in subjective reports of the illusion; participants report feeling as though their hand is in the location of the rubber hand, the appearance of the rubber hand is similar to their own hand appearance and the skin on their own hand feels rubber-y (Lewis and Lloyd, 2010). The adoption of each of these features indicates that while synchronous stroking is necessary to elicit the illusion, the

22 22 subjective experience of the RHI is a sum of both the consistent and discrepant sensory information present in the illusion context. Despite the multitude of sensory influences in the RHI, embodiment change is generally assessed using a single sensory outcome measure of proprioceptive drift. And a subjective illusion score, composed of items describing perceived causality between seen and felt touch, feelings of ownership over the rubber hand and a change to self-location (other approaches to measuring he subjective experience of the RHI are discussed in chapter 4). These two measures are often positively correlated and have been interpreted as measuring a single multisensory process of embodiment (Botvinick and Cohen, 1998; Longo et al., 2008; Lopez, Lenggenhager and Blanke, 2010; Tsakiris, 2010). Subsequent studies show that subjective and proprioceptive aspects of embodiment can function independently. Subjective and proprioceptive aspects of embodiment can be experimentally dissociated in the RHI by reducing the number of body signals manipulated in the illusion. When the illusion is conducted without any stroking (i.e. participants see a rubber hand in an anatomically plausible position a short distance away from their hidden hand), participants no longer report feelings of ownership over the rubber hand but still judge their hand to be closer to the location of the rubber hand (Holmes, Snijders and Spence; 2007; Pavani, Spence and Driver, 2000). These studies show that proprioceptive change is predominantly caused by the discrepancy between the location of the participant s hand and the seen location of the rubber hand, and that the absence of proprioceptive changes following asynchronous stroking arises because the temporal discrepancy between seen and felt touch provide sensory evidence that the rubber hand is not the participant s hand. As such, the processes underlying proprioceptive drift differ from the processes giving rise to the subjective experience of embodiment (Ehrsson et al., 2004; Fiorio, Weise, Önal-Hartmann, Zeller, Tinazzi and Classen, 2011;Kammers et al. 2009; Rohde et al. 2011). Proprioceptive drift is not a behavioural proxy of embodiment but rather measures one aspect of the multisensory process of embodiment. Embodiment may be specified in greater detail by considering outcomes of the RHI as independent measures of different aspects of the multisensory process of embodiment. This approach would allow the unique contributions of individual modalities to be specified and may be useful for investigating individual differences in embodiment.

23 23 For example, when autistic participants experience the RHI they show similar subjective changes to controls, but reduced changes to proprioception (Paton, Hohwy and Enticott, 2012; Cascio, Foss-Feig, Burnette, Heacock, 2012). This result may indicate that individuals with autism have a more reliable sense of proprioception making it less susceptible to bias from other modalities such as vision and touch (Masterson and Biederman, 1983; Haswell et al., 2009). Alternatively, it may indicate that they are less sensitive to the global context of the illusion such that they maintain a consistent subjective experience of embodiment without maintaining consistency across sensory modalities (Happé and Frith, 2006; Simmons, Robertson, McKay, Toal, McAleer and Pollick, 2009). These interpretations are not distinguishable using a single sensory outcome measure but could be investigated further by assessing multiple sensory outcomes of the RHI. The potential utility of measuring multiple sensory outcomes of the RHI is explored in this thesis The contribution of tactile signals in the Rubber Hand Illusion A number of methods for measuring changes to proprioception following the RHI have been documented. In contrast, sensory outcome measures assessing visual changes following the RHI are less common. One previous study showed that participants judge the appearance of the rubber hand to be more similar to their own hand after the RHI but this effect was not assessed as a sensory outcome (Longo et al., 2009). A sensory outcome measure assessing visual changes following a related body illusion has, however, been developed and could be adapted to measure visual changes following the RHI (Tajadura-Jiménez, Longo, Coleman and Tsakiris, 2012; Sforza, Bufalari, Haggard and Salvatore, 2010). Changes to tactile perception following the RHI have not yet been explored but touch information is an important component of the illusion because it is part of the synchrony manipulation used to elicit a change to subjective embodiment. Previous studies examining the contribution of tactile signals to embodiment have focussed on this aspect of the RHI by manipulating the quality of synchronous stroking, particularly, the pleasantness quality of touch. Touch is rated as more pleasant when it is slow rather than fast. Slow stroking activates CT-afferent fibres, a subset of mechanoreceptors which are processed via a distinct cortical pathway associated

24 24 with interoceptive processing. Slow stroking typically used in the RHI, leads to stronger ratings of subjective embodiment than fast stroking (Crucianelli, Metcalf, Fotopoulou and Jenkinson, 2013), suggesting that pleasant touch may have a unique contribution to embodiment. However, ratings of subjective embodiment are not significantly different when comparing stroking applied to the back of the hand, which contains CT-afferent fibres, and the palm of the hand, which does not contain CT-afferent fibres (Lloyd, Gillis, Lewis, Farrell and Morrison, 2013). These results suggest that while pleasant touch increases the subjective experience of embodiment other forms of touch also contribute to embodiment. Touch also provides information about the physical boundary of the body whenever contact is made with objects in the environment. This tactile feature is consistent in the RHI because the participant s hand and the rubber hand are both in contact with a table surface, albeit in a different location. As such, it is unclear whether this aspect of touch contributes to embodiment in the RHI. Chapter two explores changes to tactile perception following the RHI that could be assessed using a sensory outcome measure. By developing visual, tactile and proprioceptive outcome measures of the RHI, it would be possible to measure multiple aspects of the multisensory processes of embodiment within a single illusion context Multiple sensory outcomes of the rubber hand illusion To evaluate the utility of measuring multiple sensory outcomes of the RHI, it was used to investigate how veridical body perception influences the multisensory process of embodiment. By producing feelings of body ownership, the RHI provides a useful approximation of embodiment so that the underlying multisensory process can be investigated. However, the experience of ownership measured in the RHI is a specific state of embodiment that is not typical of embodiment in general. As discussed in section 1.1, the body is typically the background of experience and rarely the focus unless an issue with the body arises. This structure poses a problem when studying embodiment through body illusions that use sensory discrepancies to bring the body into focal awareness. Once body experience is brought into focal awareness, it is no longer the tacit form of body awareness that it typical of embodiment. To investigate this structure of embodiment requires an indirect

25 25 manipulation of body awareness. One approach would be to assess susceptibility to the RHI before and after veridical changes to embodiment, that is, changes to embodiment that arise from non-illusory body experiences. Such changes could be elicited with body perception training activities and measured using multiple sensory outcomes of the RHI. Changes to outcome measures following training would indicate which aspects of the multisensory process of embodiment had changed during training and what type of body experiences contributed to those changes. Given that the body is present for all activities many forms of training could potentially have an effect upon embodiment. To fairly assess the utility of measuring multiple sensory outcomes of the RHI over time, we selected forms of body perception training that, based on previous literature, might be expected to influence susceptibility to the RHI. As discussed in section subjective and sensory outcomes of body illusions are negatively related to interoceptive sensitivity. Based on this evidence, increasing awareness of interoceptive sensations is expected to reduce the magnitude of all outcome measures of the RHI. One method of increasing awareness of interoceptive sensations is to instruct participants to practice focussing on their internal body sensations using body scan meditation (Mehling, Daubenmier, Price, Acree, Bartmess and Stewart, 2013). This manipulation was used as a form of internal body perception training. As a contrast to this study, an external form of body perception training was also examined by investigating susceptibility to the RHI in medical students before and after anatomical dissection training. During anatomical dissection students learn about the physical form of the human body by cutting a cadaver into separate parts. During this form of training medical students learn many new sensory properties that can apply to bodies. Further, some medical students show a reduction in empathy during their first year of medical training (for reviews see Neumann, Edelhäuser, Tauschel, Fischer, Wirtz, Woopen, Haramati and Scheffer, 2011; Youssef Nunes, Sa and Williams, 2014) and empathy has been found to be positively related to susceptibility to the subjective and proprioceptive effects of the RHI (Asai, Mao, Sugimori, Tanno, 2011). As such, this form of external body perception training might be expected to alter susceptibility to some effects of the RHI, though it is unclear whether altered susceptibility would be consistent across all outcome measures, or present for all medical students.

26 Summary Changes to embodiment in the RHI have typically been assessed using a questionnaire measure and proprioceptive drift. A number of studies have shown that these measures describe different aspects of embodiment that function independently, but their results are commonly interpreted as measuring the multisensory process of embodiment in general. This generalisation contributes to an incomplete model of the multisensory process of embodiment in which the unique contribution of individual modalities is underspecified. The unique contribution of individual modalities could, however, be investigated by measuring multiple sensory outcomes of the RHI such that many aspects of the multisensory process of embodiment are assessed in a single context. By using this method to investigate changes to the multisensory process of embodiment following veridical body perception, it may be possible to assess how individual sensory modalities contribute to embodiment over time. 1.3 Research aims and overview of the Thesis The research in this thesis investigated the contribution of visual, tactile and proprioceptive modalities within the multisensory process of embodiment. The rubber hand illusion was used to manipulate multisensory information about the body and elicit a change to embodiment. The first aim of this thesis is to develop a simple augmented reality approach to the RHI in which visual, tactile and proprioceptive body signals can be manipulated independently and to use this approach to explore the contribution of touch to the subjective experience of embodiment in the RHI. Evidence presented in section suggests that the presence of sensory discrepancies in the RHI context influences the subjective experience of the illusion and produces measurable changes to sensory judgments. Therefore, the second aim of this thesis was to develop multiple sensory outcome measures assessing visual, tactile and proprioceptive changes following the RHI. The third aim was to evaluate the utility of using multiple sensory outcomes of the RHI and explore veridical long-term changes to embodiment following internal and external forms of body perception training. Examining long-term changes permits investigation of the way that embodiment becomes specialised to particular contexts

27 27 and activities. Evidence presented in section suggested that internal forms of body perception training would reduce subjective and perceptual effects of the RHI. Potential changes to embodiment following external body forms of body perception training are less clear so are explored using the methods developed in this thesis. To address these aims, five studies were conducted to investigate changes to perceptual and subjective aspects of embodiment both within the RHI and before and after body perception training. Qualitative, psychophysical and questionnaire measures were used to assess sensory influences on embodiment. The development of augmented reality equipment and application to the measurement of multiple sensory outcomes of the RHI is introduced in Chapter two. This chapter presents the results of studies one and two. In study one a qualitative investigation of subjective embodiment under varied sensory conditions showed that when visual-tactile discrepancies were present in the RHI, participants described both body ownership and body extension type changes to embodiment, and reported changes to tactile awareness. Modality specific changes to self-perception were investigated in study two, which showed that the RHI produced measurable changes to visual, tactile and proprioceptive aspects of embodiment. Collectively, these studies showed that by manipulating visual, tactile and proprioceptive aspects of the RHI context a broader range of subjective experiences could be elicited and multiple sensory outcomes could be measured. Subsequent chapters evaluate the utility of methodological developments detailed in Chapter two. An evaluation of the utility of measuring multiple sensory outcomes of the RHI is presented in Chapter three, in which the effect of body perception training on multiple sensory outcomes of the RHI was investigated in two different ways. Study three examined internal body perception training though a one week body scan meditation intervention. Study four examined external body perception training in which medical students learn about the body through anatomical dissection. Together, these studies identified modality specific changes to sensory outcome measures suggesting that individual modalities contribute to embodiment in different ways over time. The qualitative analysis in study one was used to develop an embodiment change questionnaire. This questionnaire was evaluated using psychometric analyses in

28 28 chapter four. This chapter presents the results from study five which showed that the embodiment change questionnaire was a valid measure of subjective embodiment in the RHI. The findings from each chapter and relevant theoretical and methodological issues are discussed in Chapter five, along with suggestions for future research.

29 29 2 Chapter 2: Development of subjective and perceptual measures As discussed in chapter one, the body is generally a phenomenologically transparent feature of perceptual experience and so we have limited awareness of how sensory information contributes to embodiment. To measure the contribution of individual sensory modalities their contribution must be made salient. Salience is achieved in the RHI by creating sensory discrepancies, that is, bottom up sensory information that it similar but not identical to top-down expectations regarding what the body is like. Discrepant features can become incorporated into body awareness and measured as both subjective and objective changes to self-perception. In this chapter, two studies detail the development of subjective and objective measures of embodiment change that permit the contribution of individual modalities to embodiment. In study one, a qualitative method is used to explore how visual, tactile and proprioceptive modalities contribute to the subjective experience of embodiment. In study two, psychophysical methods are used to investigate visual, tactile and proprioceptive changes following the RHI and to evaluate the feasibility of using multiple sensory outcomes of the RHI to measure embodiment. In both studies one and two, a simple approach to augmented reality is used to conduct the RHI. The potential of this approach for eliciting and measuring subjective and sensory changes to embodiment is explored. 2.1 Study 1: A qualitative investigation of embodiment in contextual variations of the Rubber Hand Illusion Introduction Previous studies have used qualitative methods to investigate the subjective experience of embodiment in the RHI (Lewis and Lloyd, 2010; Moguillansky, O'Regan and Petitmengin, 2013). By allowing participants to freely report on their subjective experience of the illusion it is possible to identify which features of the illusion are phenomenologically salient. Phenomenological salience in the RHI appears to arise from mild discrepancies between bottom-up sensory information and top-down expectations. For example, there is a discrepancy between the seen location of the rubber hand and the current expectation of where one s own hand is,

30 30 producing a subjective experience in which one s own hand is in the location of the rubber hand. Some features of the RHI are consistent in that bottom-up sensory information present in the illusion is the same as current top-down expectations of what sensory information should be present. As discussed in Chapter 1, a consistent feature of the illusion is the presence of tactile sensations arising from physical contact with the environment. Physical contact is present for both the participant s hand and the rubber hand because they both rest on a table or box surface, albeit in different locations. This aspect of the illusion does not feature in spontaneous subjective reports of the illusion, yet does appear to change during the RHI. When asked directly to describe what they felt under their hand, participants reported feeling the table surface located under the rubber hand rather than under their own hand (Moguillansky, O'Regan and Petitmengin, 2013). This suggests that the perception of tactile contact changes in the RHI but is not phenomenologically salient. Tactile contact in the RHI could become phenomenologically salient by varying tactile contact and seen contact of the rubber hand to create tactile discrepancies. Tactile discrepancies could be added to the RHI context in a number of ways, for example, using clothing like coverings similar to clothing, different table surfaces and the absence of objects seen in contact with the rubber hand. The presence of a tactile discrepancy may alter subjective embodiment in the RHI but it is unclear precisely how the subjective experience would change. Some subjective aspects of the RHI are predictable because they are similar to the discrepancy in the RHI context. For example, a change to the perceived appearance of the rubber hand is conceptually related to the visual discrepancy between the appearance of the rubber hand and the participant s hand. Other discrepancies produce less predictable changes to embodiment in the RHI. Visual discrepancies regarding the absence of a body object in the illusion is one such example. In the invisible hand and invisible finger illusion, participants watch the experimenter trace the outline of hand or finger above a table surface whilst their own hand is stroked in a synchronous way (Guterstam, Gentile, and Ehrsson, 2013; Lewis, Lloyd and Farrell, 2012), following the manipulation participants report feeling ownership over an invisible body part. Qualitative analysis showed that participants also describe subjective experiences that are unexpected consequences of the manipulation. For example, participants

31 31 report paraesthesias such as numbness and tingling, and a flexible body boundary such that the perceived shape and size of an invisible finger changes to match the pattern of stroking observed (Lewis, Lloyd and Farrell, 2012). These results suggest that when the visual form of the body is absent in body illusion, the subjective experience of embodiment is very different to the typical experience of the RHI but can be explored using qualitative analyses. A qualitative method suitable for investigating the subjective experience of embodiment is Interpretative phenomenological analysis (IPA; Smith, Flowers and Osborn, 1997; Smith, Flowers and Larkin, 2009). IPA provides a structured analysis of interview data, by exploring participants experiences, cognitions, and meaningmaking (Smith, 1996). IPA is a two stage process initially subjective reports are analysed individually and phenomenological features are annotated as codes, i.e. short descriptions of the phenomenological content of participant s subjective reports. In the second stage of analysis, similarities and differences between codes identified in each participant s subjective reports are explored and described as themes that can describe the relationships between codes. These two stages are used iteratively until a large number of themes can be collapsed into fewer broad themes that demonstrate the structure of experience throughout the subjective reports analysed. This two-stage process allows the analysis to be consistent with the experience of each participant, whilst forming a generalizable structure of experience observable in all subjective reports. A further benefit of IPA is that the interpretation of the participant s subjective experience is achieved by exploration of contextual influences, that is, broader influences that may not be explicitly described by participants yet describe variation between subjective reports. The sensitivity of IPA to contextual features and individual differences makes it an ideal approach for exploring the subjective experience of embodiment in the RHI. The present analysis aims to explore how people make sense of their body experience during visual-tactile contextual variations of the RHI. Whereas previous qualitative investigations of the RHI have used a single context to characterise the subjective experience of embodiment, the present study compares subjective reports obtained during a number of contextual variations. Contextual variations of the RHI were created using a simple augmented reality approach in which pre-recorded videos are presented via a head mounted display. In each context, the illusory hand

32 32 was presented in the same location as the participant s hand so that unlike previous studies the discrepancy between the location of the participant s hand and the seen location of the rubber hand was minimal. The context varied according to the type of visual-tactile discrepancy present in the context. These were created by varying the physical objects under the participant s hand (either a hard table or soft pad) and on top of the participant s hand (a cloth covering), and how the illusory hand appeared through the head mounted display using pre-recorded videos. The videos showed either a rubber hand on a hard table surface, an invisible hand being stroked on top of the table surface, or a rubber hand not supported by any surface (i.e. not in physical contact with an external object). Participants were interviewed before, during and after experiencing contextual variations of the RHI, so that descriptions of their general body experience and experience of embodiment during the RHI could be obtained. Interview data was analysed using IPA to assess how people make sense of their body experience during contextual variations of the RHI. This analysis was used to assess how visual, tactile and proprioceptive signals contribute to embodiment in the RHI, potential changes to tactile perception and features of subjective embodiment that were generalizable across contextual variations of the RHI. Previous qualitative investigations of the RHI examining a single illusion context have used what would be considered relatively large samples for IPA research. With this approach, these studies have explored the breadth of subjective experiences reported during the illusion, distinguishing between experiences that are typically reported and more idiosyncratic experiences (Lewis and Lloyd 2010; Moguillansky, O'Regan and Petitmengin, 2013). Examining body experiences across a large number of contextual variations, results in a substantial amount of data for each participant. In this situation, to ensure an adequate depth of analysis it is more feasible to examine a small sample in detail. A potential issue that arises from using a small sample is that it becomes difficult to assess the meaning of variation between subjective reports. For example, would differences between subjective reports indicate variation in typical body experience or that individuals in the sample have atypical body experiences. To overcome this limitation, sample characteristics can be assessed so that variation in subjective reports can be meaningfully interpreted. To this end participants were

33 33 assessed for depersonalisation characteristics. Depersonalisation is an alteration in the perception or experience of the self so that one feels detached from, and as if one is an outside observer of, one s mental processes or body (e.g. feeling as if one is in a dream) (American Psychiatric Association, 1994). In addition to ineffable feelings of unreality, depersonalisation experiences include a number of abnormal body experience symptoms, such as emotional numbing, heightened self-observation, changes in body experience and changes in the feeling of agency (Lewis, 1932; Mayer-Gross, 1935; Ackner, 1954). Whilst depersonalisation is a syndrome in its own right, it is also common to a number of clinical disorders such as anxiety, depression and schizophrenia (Simeon, Gross, Guralnik, Stein, Schmeidler and Hollander, 1998). In this study, depersonalisation characteristics were assessed using the Cambridge Depersonalisation Scale (CDS; Sierra and Berrios, 2000), and used as a general measure of abnormal body experience such that sample characteristics could be inferred and utilised in the analysis of subjective reports Method Participants For IPA, samples are selected to achieve depth of analysis rather than generalisability to the general population and so small purposeful samples are preferred to large samples. A sample size of N=3 is considered optimal, though fewer participants allows deeper explorations of variation within a single participant s subjective report which is advantageous for this study. As such two students from the University of Manchester (Ben: male, aged 32 years, Jenny: female aged 27 years) were recruited via opportunity sampling. Both participants had taken part in the RHI before the study. Participants were free from tactile and proprioceptive deficits and had normal or corrected to normal vision. Apart from these sensory capabilities, no other criteria or characteristics were used to select the participants. This approach to participant recruitment is in line with previous research investigating embodied experience during the RHI (e.g. Lewis and Lloyd, 2010; Moguillansky, O'Regan and Petitmengin, 2013). However, in qualitative research it can be useful to know the homogeneity of the sample and so participants were assessed for depersonalisation characteristics using the Cambridge Depersonalisation Scale. Written informed consent was obtained from each

34 34 participant and data was anonymised. Research was approved by the University of Manchester Ethics committee Materials Equipment A Sony HMZ-T1 head mounted display (HMD) was used to display visual stimuli such that it covered the participant s body and field of view. The position of the HMD within the experimental set up is shown in figure 2.1. The HMD was modified by replacing the straps with a cycling helmet and adding a counter weight to the back. These adjustments improved image stability and reduced pressure on the participant s nose. Visual stimuli Visual features of the RHI context were varied in pre-recorded videos of the RHI. A camera was used to record three three-minute videos from the participants point of view in the experimental set up. In each video the experimenter stroked the illusory hand in a specific pattern at a rate of one stroke per three seconds. Each video had an audio click track that matched the speed and duration of strokes to the illusory hand. Video one showed a rubber hand on a table being stroked by the experimenter. Video two showed a rubber hand that was not in contact with a table surface being stroked by the experimenter. Video three showed an empty table over which the experimenter traced the outline of an invisible hand according to the same pattern used in videos one and two. To ensure that the stroking pattern was similar to videos one and two, a transparent outline of the rubber hand was placed on the table and used as a guide for the invisible hand shape. This guide was not visible in the video. Video stimuli were presented to the participant via a head mounted display. The experimental set up and screen shots from videos one, two and three are shown in figure 2.1.

35 35 Visual stimuli: Tactile stimuli: Soft pad/table/cloth Figure 2-1: Overhead view of the experimental set up. Pre-recorded videos of the RHI were displayed via a HMD. When the video was played, the experimenter stroked the participant s hand using an audio click track in the video as a guide for timing and duration. Tactile stimuli Three different types of tactile stimuli were used in the study. A cloth covering placed over the participant s hand to mimic tactile contact that arises from clothing, a soft pad (10 x 10cm Hollow fibre polyester filled pad) positioned under the participants hand and a hard, smooth wooden table under the participant s hand. Questionnaire measures RHI Questionnaire: Six items used in the original RHI study (Botvinick and Cohen, 1998) were used to assess subjective embodiment during videos one and two (i.e. when a rubber hand was present in the video). Questionnaire responses are marked on a seven point likert scale ranging from -3 (strongly disagree) to + (strongly agree). Items 1-3 describe items considered to describe the typical experience of the

36 Control Standard illusion 36 RHI. These items were averaged to form an aggregated illusion score. Items 4-6 describe body experiences that are not associated with the RHI. These items were averaged to form an aggregated control score. Items are displayed in table 2.1. Table 2-1: Items in the RHI questionnaire 1...the rubber hand was my hand 2 the touch I felt was caused by the touch I saw on the rubber hand 3 I felt the touch in the location that I saw the touch it seemed as though my real hand was drifting towards the 4 rubber hand 5 it seemed as though I might have more than one right arm it seemed as though the touch I felt came from somewhere in 6 between my hand and the rubber hand it seemed as though (visually) the rubber hand was drifting 7 towards my hand Note: All items begin with It seemed as though. Invisible Hand Questionnaire: Six items used to assess the invisible hand illusion (Guterstam, Gentile and Ehrsson, 2013) were used to assess subjective embodiment during video three (i.e. when no hand was visible). Items 1-2 were averaged to form a comparable standard illusion score and items 3-4 were averaged to form a comparable control score. These items are displayed in table 2.2. The Cambridge Depersonalisation Scale: (CDS; Sierra and Berrios, 2000) comprises 29 items associated with depersonalization disorder. The items describe abnormal experiences affecting some sensory modalities, an inability to experience some emotions, heightened self-observation, the lack of body ownership feelings, somatosensory distortions, out-of-body experiences, autoscopy, and the lack of agency (Sierra, Baker, Medford and David, 2005). Each item requires two responses on two Likert scales, the first describes the frequency of an experience (never = 0,

37 Control Standard illusion 37 rarely = 1, often = 2, very often = 3, all the time = 4) and the second describes the typical duration of an experience (few seconds = 1, few minutes = 2, few hours = 3, about a day = 4, more than a day = 5, more than a week = 6) of the experience. All responses are summed into a single score (ranging from 0 to 290). A cutoff point of 70 was shown to yield a sensitivity of 75.5% for patients with depersonalization disorder. An example of this questionnaire is presented in Appendix D. Table 2-2: Items in the Invisible Hand Questionnaire 1...it felt as if my right hand were located on the table where I saw the brush moving, as if I had an invisible hand. 2...I felt the touch of the brush in empty space in the location where I saw the brush moving. 3...I felt the touch of the brush directly on the table below where I saw the brush moving. 4...it felt as if my right hand disappeared, as if it had been amputated. Note: All items begin with It seemed as though. Interview schedule Three interview schedules provided guides for semi-structured interviews conducted before, during and after the RHI. Before RHI: what is your body experience like? What are unusual body experiences like? What was your previous experience of the rubber hand illusion like? ; during RHI: what is this experience like? ; after RHI: how would you describe your experiences in the illusions overall? Have these experiences changed the way that you think about your body experience? How did it feel to discuss your body experience during the illusions?. Further questions and general prompts, such as can you explain that further? were used to encourage detailed responses. All responses were audio recorded using an Olympus WS-802 digital voice recorder.

38 Procedure Participants initially completed an introductory interview in which they were asked to describe their general body experience. Once this was complete, participants sat at the table as shown in Figure 1.1 and put on the HMD. The HMD straps were secured and the optical settings adjusted until the picture was judged as clear by the participant. The position of the HMD was maintained by the participant placing their chin in a chin rest. The participants took part in nine contextual variations of the RHI described in Table 2.3. At the start of each illusion, participants were instructed to describe their experience as fully as possible. The experimenter played the video and stroked the participant s hand according to the pattern used in the video. Each contextual variation was completed as a synchronous and asynchronous condition. During synchronous conditions, the experimenter stroked the participant s hand at the same time as the stroking in the video, during asynchronous conditions the experimenter stroked the participant s hand 500ms after the stroke in the video had occurred. The timing of these strokes was done according to the audio click track on the video which had different tones for synchronous and asynchronous conditions. If the participant had not begun to describe their experience after 1.5 minutes, they were prompted with questions (e.g. what is this experience like? ). Once the video was finished, the participant removed the HMD and completed either the RHI questionnaire or invisible hand questionnaire, depending on which video was used in that condition. A three minute break was provided between each condition to allow the illusion to diminish before the next condition (e.g. McKenzie and Newport, 2015). Once all 18 conditions were complete, a final interview was conducted using the debrief schedule. All interview responses were audio recorded and transcribed verbatim but analysis of this data is not presented in this thesis.

39 39 Table 2-3: Nine contextual variations of the RHI made through combination of visual and tactile stimuli. Visual stimuli presented as pre-recorded videos and tactile stimuli positioned either under or over participant s hand. Table displays a description of the visual-tactile discrepancy or consistency in each contextual variation. Participants completed each variation twice, once with synchronous stroking and once with asynchronous stroking to their own hand. Video 1 Table contact Visual contact Video 2 No contact Video 3 No hand Table contact Consistent Contact felt but not visible Contact but no hand form visible Tactile contact Soft pad contact Cloth covering tactile reduced Tactile increased Reduced felt contact but no contact visible Increased felt contact but no contact visible Reduced felt contact but no hand form visible Increased felt contact but no hand form visible Analysis Transcripts were analysed using the iterative two stage process of IPA (Smith, 1996; Smith, Flowers and Larkin, 2009). Validity and reliability are important considerations in qualitative research. Smith (2015) suggested two important criteria to address this: internal coherence, which refers to whether the argument presented is internally consistent; and presentation of evidence, which refers to the availability of sufficient data for the reader to assess the interpretation. To meet these criteria, each theme is supported by extracts of the participants actual discourse Results Questionnaire data The distribution of questionnaire data was non-normal so Wilcoxon signed rank tests were used to assess whether responses to illusion items and control items were significantly different in synchronous and asynchronous conditions for each

40 40 participant. For Ben, responses to illusion items were significantly higher in the synchronous than asynchronous conditions (z= , p=.011), but responses to control items were not significantly different (z= , p=.482). For Jenny, responses to illusion items were not significantly different between synchronous and asynchronous conditions (z= , p=.063), and responses to control scores were significantly higher in the asynchronous than synchronous condition (z= -2.41, p=.016), mean values are presented in Figure 2.2. Comparison of CDS scores showed that Ben had a low score for depersonalisation (CDS=5) whereas Jenny had a high score (CDS=89) that exceeded the cut off for depersonalisation syndrome Qualitative data An IPA analysis addressed the following question: How do people make sense of their body experience during contextual variations of the RHI? The analysis showed that participants described a change to embodiment in each of the contextual variations. Their reports were similar to findings from previous qualitative studies showing that the subjective experience of the RHI is composed as general features in which there is a global change to embodiment reflecting the degree that the illusory hand is incorporated into self-perception and a number of modality specific changes such that characteristics of the illusory hand become more salient. The analysis resulted in one superordinate theme: The expectation of ambiguous body experiences influences the granularity of embodiment across contextual variations of the RHI. This theme described the relationship between three subthemes: Ambiguity in body signalling experiences; Tactile awareness and the granularity of embodiment in the RHI; and, Association with the illusory hand as a distinct form of embodiment in the RHI. These themes are discussed and summarised in conjunction with the questionnaire data.

41 41 Figure 2-2: Graph displaying the mean scores for illusion and control items during synchronous and asynchronous stroking for each participant. The expectation of ambiguous body experiences influences the granularity of embodiment across contextual variations of the RHI The expectation of ambiguous body experiences emerged as a key factor that could account for the different ways that participants made sense of their body experiences across contextual variations of the RHI. The expectation of ambiguous body experience reflected the need to interpret common physical sensations in general, and in the RHI, this was reflected as the degree of granularity of embodiment (i.e. whether participants gave general descriptions of body experience that were invariant across contextual variations or more specific interpretations of sensory information that vary across contextual variations). The granularity of embodiment described differences between the participants in terms of how varied their body experiences were across different contextual variations. Varied body experiences arose when awareness of discrepant tactile information was utilised to make sense of embodiment in different ways depending upon the context, while a lack of variation occurred when awareness of discrepant tactile sensations were reduced regardless of the context. Awareness of discrepant tactile sensations was associated with a form of embodiment distinct from ownership or incorporation of the illusory hand. This experience was described as an association with the illusory hand without a change

42 42 to self-experience. This theme accounted for the relationships between three subthemes: Ambiguity in body signalling experiences; Tactile awareness and the granularity of embodiment in the RHI; and, Association with the illusory hand as a distinct form of embodiment in the RHI. Subthemes are discussed and supported with extracts from the participants subjective reports. Ambiguity in body signalling experiences When describing their general body experience, both participants described their body as neutral and that they did not really pay attention to their body sensations unless deliberately focusing on the body (e.g. during relaxation), or when a body signalling sensation occurred in which case attention was directed to the body. For Ben, signalling experiences were clearly defined by their cause. For Jenny, signalling experiences could be ambiguous and required interpretation using contextual knowledge about previous or upcoming events. These extracts compare different experiences of body signalling. Ben: I would say [my body experience] is just like an absence of anything particular, if I was feeling something particular then that would be abnormal. I think the normal experience is just to kind of not notice it really, I just notice anything kind of bad or painful and uncomfortable. I wouldn t notice it a lot. I guess it s just a warning that something is wrong with you and to take it easy. I go from, it s either its negative there s something wrong or it s kind of completely neutral and everything is fine. The scale seems to go from bad to neutral. There doesn t seem to be much above that Jenny: Unless I m trying to relax or meditate or something like that, I don t usually pay attention to my own bodily functions and stuff. I only do that when there s some kind of imbalance going on when my body is signalling me hello take care of me. Its majorly biological ones but I can also have this warning signal at a more emotional level like for example when I m feeling anxious or upset I also have bodily or physiological symptoms or signs. Sometimes they are obvious

43 43 sometimes they aren t, honestly speaking. Most times they are in the sense that my stomach is growling or my stomach is a little bit tight and I m also feeling the growl, having the growling signs and I feel a little bit dizzy and these all three kind of symptoms put together signify that I m hungry. Whereas if I know that if I have already eaten or I have a very anxiogenic event coming up, I know that its due to that There is a bit of interpretation or knowledge regarding the events that are coming up, so that is part of the information that helps me identify or label some of my body sensations. These extracts differ in terms of the amount of interpretation required to understand signalling experiences. For Ben, body signalling was related to physical issues with the body such as injury or illness, whereas for Jenny they were related to more common everyday experiences such as emotions and hunger. She described them as symptoms, ambiguous experiences that needed to be consciously interpreted using knowledge of past or future events. The expected ambiguity of body experience featured in the way that each participant made sense of their body experience during contextual variations of the RHI. Tactile awareness and the granularity of embodiment in the RHI Jenny s expectation of ambiguous body experiences in general, was associated with low granularity of body experiences across the contextual variations of the RHI. She reported feeling ownership over the illusory hand in all nine conditions. In the eight conditions with visual-tactile discrepancies, she described reduced awareness of tactile signals and, specific to the no hand video conditions, reduced awareness of hand presence. In contrast, Ben s low expectation of body ambiguity was associated with high granularity of body experiences such that he gave varied descriptions of body experience that were specific to each condition. In some conditions, particularly the no discrepancy and cloth covering conditions, he reported feeling ownership and agency over the illusory hand, suggesting the illusory hand had been incorporated into his body experience. In other conditions, Ben described feelings of association with the illusory hand without a change to his body experience (e.g. I ve got this association with [the illusory hand], but I wouldn t say it feels like

44 44 my hand particularly ). In all contexts, Ben described becoming aware of the sensory features of the illusory hand, either as his own hand or as an associated object. These sensory features varied depending upon the type of visual-tactile discrepancy in the illusion context (e.g. looks like a dead hand in the soft pad conditions, my hand feels like it s made out of cloth in cloth covering conditions). The following quotes were taken from the contact felt but not visible condition and show the different uses of the tactile discrepancy in their interpretation of their body experience. Ben interprets his awareness of the tactile discrepancy as evidence that the rubber hand is not part of his body but becomes aware of the stiff or wooden quality of the rubber hand as an object. In contrast, Jenny interprets the tactile discrepancy as a reduction to her own tactile awareness, which she describes as a typical body experience that can occur when tired. Ben: It feels slightly as though I ve got a wooden arm because what I can see, the arm is quite stiff. It s projecting over the edge of the table so it feels as though it should be bending but it s not so it feels a bit wooden. Overall I feel less ownership over this hand because there s nothing underneath it It s more like I m watching something else than its mine. So the sensation of the table on my actual hand is quite strong, like it s quite noticeable Jenny: At first I was looking at it and thinking, it s not my hand And then I started to feel more like my hand was in thin air, like you know when you sit your whole body on the bed and you re really really tired and you kind of feel like you re falling in the bed, because you re really tired. It felt exactly like that in my hand just because there s nothing [seen] under it and it feels like there s no table kind of it s not like something is holding my hand or anything like that. I would like to show you. Like for example if you hold your hand at a certain angle over a table or over something it kind of sits like this. It feels as if its sitting naturally a little bit more straight without me having to do any effort. [After resting her hand over the end of the table] Look its floppy, it shouldn t be floppy! I m not even considering the basic laws of physics anymore, of course if I put my hand like this, of course it s going to be floppy I need to actually do stuff to disprove myself

45 45 The quotes show that Ben utilises tactile presence and interoceptive information in conjunction with visual information to interpret this RHI context. Collectively the sensory information is interpreted as non-relevant to self-perception, but rather an indication of the properties of the rubber hand. This description implies some tacit awareness that a hand held in the position of the rubber hand seen in the video would require a quality of rigidity, whereas Jenny does not make this inference until she enacts the position of the rubber hand. Through this enactment, she becomes aware of the effort that would be required. During the illusion, Ben maintained awareness of tactile and interoceptive signals and they featured in his description of the illusory hand resulting in a variety of different body experiences, but Jenny had a reduced awareness of tactile and interoceptive signals resulting in generic incorporation of the illusory hand that was consistent across contextual variations of the RHI. The participants interpretation of their body experience was influenced by visual-tactile discrepancies in different ways. For Jenny the discrepancy was resolved through reduced awareness of tactile information and with this resolution, the illusory hand could become incorporated into self-experience. For Ben, the discrepancy was not resolved and instead reduced the extent of incorporation, such that the illusory hand was associated with but not part of his body. By not resolving the discrepancy, Ben reported greater granularity of body experiences, firstly because the illusory hand could be associated with or incorporated in his self-experience and secondly, because the visual-tactile discrepancy provided information about the qualities of the illusory hand as an object. Association with the illusory hand as a distinct form of embodiment in the RHI Incorporation of the rubber hand is the typical experience of the RHI. The present analysis identified an additional form of embodiment change in which Ben described feeling an association with the illusory hand as an object separate from his body. Association-type changes to embodiment did not include changes to self-perception, but they did include descriptions of body awareness under certain circumstances, particularly through comparison of the illusory and real hands during the illusion and after the illusion was over. Following each condition, both participants described changes to the way they perceived their own hand, suggesting that both

46 46 incorporation and association-type experiences produce sensory after-effects. The following excerpts show Ben s awareness of his own body when comparing between two separate hand perceptions, his own hand and the illusory hand. Further excepts describe after-effects from both participants. Ben: It s strange; I don t feel a great sense of ownership [over] it. But when there s something slightly not what I expect, I do notice it I kind of expect to feel what I see, but then as you re reaching, I m expecting a certain finger to be stroked and there was one instance where I kind of thought my finger s not in that position, which is odd because I didn t especially feel like it was my hand being stroked it s kind of like I ve taken ownership of it and it s become a bit too normal and so it was my finger not being where it should be that was surprising It s like I ve got used to this idea that there s this thing I m looking at that I ve got this connection with and it s not surprising anymore [during a break between conditions] I have noticed that every time I take [the HMD] off, I m starting to think that my hand looks more and more weird Every time, I feel slightly more revolted by my actual hand and more accustomed to the [rubber] hand And then when I look at my own hand it looks a bit sort of blotchy and thin and pink See this question feels like the wrong question now, the rubber hand began to resemble my own hand. It s like the other (way around); my real hand is beginning to not represent the rubber hand. Jenny: It s so strange. Now, after the conditions over, I still feel a bit funny and numb. It s like it didn t cancel itself out. I actually felt like not wanting to move my hand Didn t you see that? I was holding the pen like that without moving my hand. The fact that incorporation and association-type experiences were associated with changes to self-perception after the illusion suggests that both types of description indicate a change to embodiment. Further during both forms of embodiment change participants emphasised the importance of expecting the seen and felt touch to occur at the same time for these experiences. While incorporation experiences are an explicit change to self-perception, association experiences are implicitly related to

47 47 self-experience because the body is part of marginal awareness. Marginal awareness of the body appears to occur when the seen and felt touch used to elicit the illusion are expected. Nevertheless, during these experiences the body can become the focus of awareness when a discrepancy between the real hand percept and illusory hand percept arises Results summary Collectively the questionnaire data and IPA data suggest that the expected ambiguity of body experience influenced how participants made sense of their body experience in contextual variations of the RHI. Jenny reported an atypical body experience on the questionnaire measures. Her CDS score exceeded the cut off for depersonalisation syndrome indicting that she frequently experienced unusual body sensations, and her responses to questionnaires assessing subjective embodiment indicated an atypical experience of the RHI. The IPA indicated that her body experiences sometimes required conscious interpretation because they could be ambiguous. By expecting ambiguous body experiences, she incorporated discrepant sensory information in the RHI into her self-experience and reported a generic change to her body experience, such that, awareness of discrepant tactile and interoceptive information was reduced. In contrast, Ben showed a more typical body experience on the CDS and subjective embodiment questionnaires, and did not describe ambiguous body experiences in general. In the absence of an expectation of ambiguous body experiences, specific visual-tactile discrepancies reduced the extent of incorporation leading to a more nuanced form of embodiment in which the illusory hand was perceived to be associated with his body but not part of it Discussion The IPA resulted in one superordinate theme, The expectation of ambiguous body experiences influences the granularity of embodiment across contextual variations of the RHI. This theme described the relationship between three subthemes: Ambiguity in body signalling experiences; Tactile awareness and the granularity of embodiment in the RHI; and, Association with the illusory hand as a distinct form of

48 48 embodiment in the RHI. The analysis suggests that visual-tactile discrepancies in the RHI context may be useful for investigating individual differences in embodiment. The analysis was further used to assess how visual, tactile and proprioceptive signals contribute to embodiment in the RHI, potential changes to tactile perception and features of subjective embodiment that were generalizable across contextual variations of the RHI. These issues are addressed in this discussion. The IPA results provide a detailed account of the subjective experience of embodiment in contextual variations of the RHI. In IPA, depth of analysis is enhanced by identifying structural features that are generalizable across accounts, taking into account not only consistency between accounts but differences between them too (Smith, Flowers and Larkin, 2009). Differences between accounts are of particular importance because they highlight descriptive dimensions that account for variation between individuals and contexts. However, to meaningfully interpret variation between accounts it is necessary to refer to the sample characteristics. Scores on the CDS indicated that Jenny frequently experienced unusual body experiences, whereas Ben experienced more typical body experiences. As such, the participants in this study did not form a homogenous sample. Homogeneity of sample is often a desirable quality in qualitative research because findings can be considered to be representative of a specific population (Patton, 2015; Robson, 2002). From this perspective, it could be considered that the inclusion of a participant with abnormal body experience in the present study is invalid. However, alternative sampling approaches that utilise purposeful heterogeneity can also be advantageous in qualitative research. Of particular relevance to the present study, is the extreme case sampling approach in which extreme cases are examined in order to highlight and understand characteristics of more typical experiences (Ulin, Robinson and Tolley, 2012). This approach is beneficial, particularly for exploratory analyses, because it magnifies the characteristics that define what is considered to be typical and abnormal. Following from this, the IPA in this study is best considered as a case study comparison, in which extreme cases are studied in greater detail than is typically possible in larger IPA studies, such that a more detailed comparison of individuals can be achieved. This case study approach epitomises the idiographic rationale of IPA and is encouraged by leading proponents of the method (Smith, Flowers and Larkin, 2009). By comparing between participants with typical and

49 49 abnormal body experiences, it has been possible to identify phenomenological characteristics that distinguish typical and abnormal embodiment, such as the granularity of body experience. When asked to describe their typical body experience their descriptions shared some structural features of typical body experience (Leder, 1990), such that they described marginal awareness of their bodies unless something was wrong, these situations were described as body signalling. Body signalling experiences occurred in different forms for each participant; whereas Ben described infrequent body signalling indicating pain or illness, Jenny described much more common experiences of emotions and hunger leading to body awareness. These differences implied that Jenny was more likely to become aware of her body. The participants also differed in terms of the interpretability of body signalling experiences. Jenny in particular described hunger and emotions as having similar physical manifestations and so they were ambiguous experiences requiring interpretation using contextual knowledge of past and future events. Differences between the participants in terms of the expected ambiguity of body signalling experiences was observed in the RHI as differences in the granularity of embodiment, with high ambiguity being associated with low granularity and vice versa. Low granularity was the tendency to interpret each contextual variation in the same way, as incorporation of the illusory hand and reduced tactile awareness, and high granularity was the tendency to report varied body experiences across contextual variations. Individual differences in the granularity of embodiment were observable in this study by comparing subjective reports from multiple contexts. This approach is not feasible for assessing individual differences in embodiment in larger samples because of the time required to collect and analyse data. Nevertheless, the analysis could be used to inform methods that may be useful for investigating individual differences in embodiment further. Increased granularity of embodiment was indicated by association-type experiences during contexts with visual-tactile discrepancies. This form of embodiment change is not currently described in questionnaire measures of subjective embodiment in the RHI. The current questionnaire assesses only incorporation type changes to embodiment. Incorporation is assessed using items describing ownership over the rubber hand, feelings of agency over the rubber hand, perceiving one s own hand in the location of the rubber hand and the perception of causality between the seen and

50 50 felt touch used to elicit the illusion. In contrast, association-type experiences described feelings of association and connection with the rubber hand as a joined object rather than part of the body. In the analysis, the experiences were distinguishable by the ability to perceive one s own hand as a distinct object to the illusory hand, and awareness of one s own hand. These features of distinctness and body awareness are similar to the distinction between body ownership and body extension (De Preester and Tsakiris, 2009). The present results suggest that it may be possible to measures these forms of embodiment in a contextual variation of the RHI, using an expanded questionnaire measure. The potential to measure both body ownership and body extension in the RHI requires further empirical investigation. The IPA results of the present study cannot be generalised because only a small number of participants who may not be representative of the general population were included. This is typical of IPA because depth of analysis is prioritised over generalisability (Smith, Flowers and Larkin, 2009). The present results can, however, be used to inform the development of subjective questionnaire measure that can be quantitatively assessed in further research. The results suggest two potential changes to the exiting questionnaire that could be useful for measuring body ownership and body extension-type experiences. Firstly, items describing the experience of association as defined above could be added to the questionnaire. Secondly, items describing modality specific changes to object perception (such as stiffness) could be added to the questionnaire. However, items describing modality specific changes may have limited utility for the measurement of embodiment. Modality specific items are included in the current questionnaire but these are not highly endorsed and so are rarely included in statistical analyses of embodiment outcomes. The main exception to this is the modality specific item describing a change to self-location. Though this item is commonly endorsed, the present results suggest that the utility of this item as a measure of embodiment may also be limited. Participants in this study did not spontaneously describe feelings of self-location. However, the contextual variations explored did not contain a discrepancy between the seen location of the illusory hand and the felt location of the participant s hand. Changes to self-location may not be perceptually salient unless a discrepancy is present, this will be discussed further in study two of this chapter.

51 51 The present results suggest that the subjective experience of embodiment includes experiences that are specific to the contextual features of the illusion and generalizable experiences that can occur across multiple contextual variations of the illusion. By using items that describe only generalizable experiences, it may be possible to develop a questionnaire for assessing changes to embodiment in different illusion contexts. The generalizable experiences reported in this study described changes to the perceived relationship between the real hand and illusory hand, specifically body ownership and body extension, as well as the perception of causality. These experiences may represent underlying factors of subjective embodiment that could be assessed using psychometric methods such as factor analysis (see Chapter four). This questionnaire would be useful for comparing the extent of incorporation that occurs in each illusion context. For example, the influence of visual-tactile discrepancies (arising from seen and felt object contact discrepancies) on the extent of incorporation could be quantitatively assessed using conditions in which visual-tactile aspects of the RHI were consistent or discrepant and comparing responses to items describing body ownership and body extension. When a visual-tactile discrepancy was present in the RHI context, participants also reported changes to tactile awareness such that the perceived contact between the participant s hand and an external object was either more or less apparent. This subjective experience may indicate that a change to tactile perception had occurred which could be measured as a sensory outcome of the RHI. Other discrepancies in the illusion, such as the proprioceptive discrepancy between the seen location of the rubber hand and the felt location of the participant s hand, have been shown to produce measurable after-effects and so the presence of a tactile discrepancy might be expected to do the same. Changes to tactile perception could be measured as changes to tactile sensitivity, though it is unclear precisely how tactile perception might be expected to change. Tactile sensitivity may be reduced through inhibition of discrepant tactile signals or it may be increased by increasing the gain of tactile signals to obtain more reliable tactile information for resolving the visual-tactile discrepancy. Further, examination of tactile sensitivity following the RHI is required to assess whether subjective changes to tactile awareness are indicative of objective changes to tactile perception, this is discussed further in study two of this chapter.

52 52 The development of a tactile outcome measure of the RHI and expanded questionnaire assessing the degree of subjective incorporation in the RHI, may provide new tools for examining individual differences in embodiment, in particular, for evaluating the amount of sensory ambiguity that is expected during selfperception. Ambiguous body experience was associated with reduced granularity of subjective embodiment in the RHI. Reduced granularity appeared to arise from reduced awareness of tactile and interoceptive signals, such that they did not feature in the overall interpretation of body signals. Previous studies have shown that participants with low interoceptive sensitivity are more susceptible to the effects of body illusions, that is they are more likely to incorporate external body objects into their self-perception (Tsakiris et al., 2011; Tajadura-Jimenez and Tsakiris, 2013). The present results suggest that when participants have a low awareness of interoceptive and tactile signals they use a one-size-fits all interpretation of the RHI, in which the synchronous seen and felt touch used to elicit the illusion is always interpreted as relevant to self-perception. While the analysis has generated some useful observations, it is important to be mindful of the role of the researcher in both the generation and analysis of the data. The observations are not objective impressions of the phenomena studied but rather active constructions that arise from the researcher s engagement with the participants and the subjective reports (Banister, Burman, Parker, Taylor and Tindall, 1994). It is explicitly acknowledged in IPA that there is a double hermeneutic process, such that that the resulting analysis reflects the researcher s attempts to make sense of the participants attempts to make sense of their experience (Smith, 2004). As such, the results cannot be separated from the influence of the researcher. However, to minimise the biasing effect of the researcher upon the final analysis it is useful to engage in reflexive practices, in which the researcher s pre-existing expectations and beliefs are made explicit so that they can be compared to the data generated (Murray, 2014). The reflexive practices employed in this study were the generation of an initial reflexive statement and reflexive diary. Though it is beyond the scope of the present study to consider all reflexive instances that occurred, it is useful to compare initial preconceptions to the final output of the

53 53 analysis. During the initial stages of the study, I had a number of preconceptions about the data that this study could generate. These preconceptions were largely influenced by my experience conducting a previous qualitative investigation of the RHI (Lewis and Lloyd, 2010). In the previous study, subjective reports had a consistent structure such that participants would report a global description of their body experience as well as modality specific features, such as visual and proprioceptive changes to self-perception. Descriptions of modality specific features were more idiosyncratic and appeared to me to be modality specific, that is, individual participants tended to describe modality specific changes in a particular modality. Based on these experiences I expected that when participants experienced contextual variations of the RHI, their descriptions would indicate a bias towards a particular modality that was consistent across the variations. Contrary to these expectations, the present analysis indicated that the tendency to experience modality specific changes was driven by the granularity of body experience rather than a modality specific bias. Though making these expectations and their origins explicit cannot remove their influence upon the data, from a personal perspective, this process can make apparent where novel insights in the analysis have emerged. Further, it can to some extent, make the researcher s influence upon the data more apparent to the reader. In the present study, subjective reports were collected from two participants with typical and atypical body perception and explored using IPA to assess how people make sense of their body experience during contextual variations of the RHI. The results suggested that participants resolved visual-tactile discrepancies through either reduced awareness of both discrepant tactile and interoceptive sensations, or reduced incorporation of an illusory hand. The tendency to resolve discrepancies in a particular way was associated with individual differences in the expected ambiguity of body experience. Individual differences in embodiment could be investigated further by increasing the forms of embodiment change assessed by questionnaire measures and measuring changes to tactile and visual perception.

54 Study 2: Measuring multiple sensory outcomes of embodiment change using a simplified augmented reality version of the Rubber Hand Illusion Introduction As discussed in Chapter one, embodiment arises from the integration of multiple signals including vision, touch, proprioception and interoception (Seth, 2013). The way that these sensory modalities contribute to embodiment can be examined using sensory outcome measures of body illusions such as the RHI. Previous studies have used a singular sensory outcome measure to assess embodiment in body illusions and then generalised findings to the multisensory process of embodiment. This approach may have led to an under-specification of the unique contribution of individual modalities. For example, there is a strong positive correlation between proprioceptive measures of the RHI conducted six months apart. This result has been interpreted as an indication that embodiment is perceptually stable over time (Bekrater-Bodmann, Foell, Diers and Flor, 2012). Nevertheless, it is possible that this result is specific to the contribution of the proprioceptive modality. Other modalities such as vision and touch may contribute to embodiment in different ways over time. This alternative interpretation seems likely given that outcomes of the RHI are considered to measure different aspects of body representations, and have been shown to measure dissociable mechanisms within the multisensory process of embodiment (Rohde et al., 2011). An alternative approach is to measure multiple sensory outcomes within a single illusion context. The RHI provides an ideal context because it contains visual and proprioceptive discrepancies that lead to visual and proprioceptive sensory outcomes. Proprioceptive outcomes of the RHI are widely reported and have been measured using a number of psychophysical methods. Most commonly, a method of adjustment has been used in which the experimenter adjusts a location marker until the participant indicates that it is in-line with the perceived location of the participant s index finger (e.g., Botvinick and Cohen, 1998). Other studies have used a two alternative forced choice procedure with adaptive staircase method, in which a location marker appears and participants judge whether it is to the left or right of their index finger (Rohde et al., 2011). The visual outcome of the RHI has been assessed using likert scale ratings of the perceived similarity of appearance between

55 55 the rubber hand and their own hand, which has been shown to increase following the RHI (Longo et al., 2009). Study one suggested that it may also be possible to measures a tactile outcome of the RHI. When the RHI context included a tactile discrepancy (e.g., conflicting information regarding the presence and absence of tactile signals arising from physical contact with objects) participants reported changes to tactile awareness. These subjective reports may indicate that participants had a change to tactile sensitivity following the illusion which could be measured by assessing changes to their tactile sensitivity threshold. Collectively, these studies suggest that it may be feasible to measure visual, tactile and proprioceptive outcomes of the RHI. A psychophysical approach to measuring visual outcomes of body illusions has been developed in a related body illusion referred to as enfacement (Tajadura-Jimenez et al., 2012; Tsakiris, 2008; Sforza et al., 2010). In the enfacement illusion, participants watch a video of another person s face being stroked whilst their own face is stroked at the same time. This video is presented on a computer monitor giving the impression of looking into a mirror but seeing another person s face. The cooccurrence of seen and felt touches leads participants to report feeling as though they are looking at their own face (Tajadura-Jimenez et al., 2012). Following the illusion, participants also incorporate the appearance of the other s face in the video into their perception of their own face. This effect is measured by showing the participant blended images that are composites of their own face and the face used in the video. After presenting a blended image, participants are asked to judge whether the image looks more like their own face or the face seen in the video. This outcome measure assesses the proportion of other s face in images judged as being like the participants own face, following the illusion participants judge blended images with a higher proportion of the other s face as being like their own face (Tsakiris, 2008; Sforza et al., 2010; Tajadura-Jimenez et al., 2012). This visual outcome measure could be adapted for the RHI. To measure a visual outcome of the RHI in a way that is comparable to veridical self-perception, requires the presentation of blended images over the location of the participant s body. This is not feasible with the standard RHI set up but could be achieved using Augmented Reality (AR) or Virtual Reality (VR) equipment. In VR it is possible to digitally create an entire environment and avatar body which is often

56 56 presented to the participant through a head mounted display (HMD; Yuan and Steed, 2010). HMDs hold two digital screens over the participant s eyes to cover the entire visual field. The participant feels as though they are immersed in the new environment presented through the display and that their body has been replaced with an avatar body. This approach has been used in a number of previous embodiment studies (e.g. Perez-Marcos, Slater and Sanchez-Vives, 2009; Slater, Ehrsson and, Sanchez-Vives 2009; IJsselsteijn, de Kort and Haans, 2006; Raz, Weiss and Reiner, 2008). As the avatar body is digitally created visual characteristics can be varied. For example, using an avatar with a larger body leads to increases in perceived body size (Normand, Giannopoulos, Spanlang and Slater, 2011). Manipulations of size and shape are suited to VR but other visual features such as skin surface appearance and textures are not easily reproduced in VR without advanced digital rendering techniques. These visual aspects are important for assessing the visual outcome of the RHI because the blended images presented should look like the participant s hand. As such, the present study used a photorealistic display that can engage the visual system in a more ecological way (Ferwerda, 2003). Body illusions using photorealistic AR typically present a live camera feed of an alternative body to the participant via a HMD. In these studies the participant sees a real or fake body in the location of their own body (Schmalzl, Thomke, Ragno, Nilseryd, Stockselius and Ehrsson, 2011, Guterstam and Ehrsson, 2012, Lenggenhager, Tadi, Metzinger and Blanke, 2007; Ehrsson, 2007; Petkova and Ehrsson, 2008, Slater, Splanlang, Sanchez-vives et al, 2010). This approach has been useful for creating full body illusions in which participants report feeling as though an alternative body is their own (Leggenhager, Tadi, Metzinger and Blanke, 2007; Ehrsson, 2007, Petkova and Ehrsson, 2008; Slater, Splanlang, Sanchez-vives et al, 2010). The use of real-time camera feeds allows features of the illusion to be varied in unusual but visually realistic ways, however, it lacks some of the flexibility that can be achieved with VR because what the participant sees is a physical rather than digital environment. Combined approaches, such as the MIRAGE box, use real-time image augmentation of live camera feeds. The MIRAGE box can capture realistic images of the participant s own hand which can be digitally augmented, for example, by stretching finger or presenting multiple hands during the illusion (Regenbrecht, Franz, McGregor et al., 2011; Newport and Preston, 2010; Newport, Pearce and

57 57 Preston, 2010; Newport and Gilpin, 2011). A simple alternative employed in the current work is to use pre-recorded images and videos that are displayed via a HMD. Pre-recorded images are photorealistic but can be digitally altered prior to the experiment providing an ideal balance between image quality and flexibility. Further, this approach is cost effective and easy to use. By presenting pre-recorded videos via a HMD a novel version of the RHI was created in which there were sensory discrepancies to vision, touch and proprioception. Visual aspects of the illusion were created using a pre-recorded video of the RHI. Like most RHI studies, the rubber hand had a different appearance and appeared in a different location to the participant s hand. Unlike previous studies, the rubber hand protruded over the edge of a table surface so that it was not in contact with any object. This was in contrast with the participant s hand that rested on a table surface during the illusion. This created an additional visual-tactile discrepancy regarding the presence/absence of tactile signals arising from contact with the table. Study one explored subjective descriptions of this RHI context, showing that participants report a change to embodiment and changes to tactile awareness. There were two main aims of the study. Firstly, to assess whether multiple sensory outcomes can be measured following the RHI, including visual, tactile and proprioceptive changes. Secondly, to assess the feasibility of using prerecorded videos and images presented via a HMD to eliciting the RHI and measuring sensory outcomes. By addressing these aims this study develops an accessible methodological approach, which could be used to investigate the unique contribution of individual sensory modalities within the multisensory process of embodiment Study A: Method Participants 36 undergraduate students (male: female = 16: 20, mean age = 26 years, age range = years, right-handed: left-handed = 32: 4) were recruited from the University of Manchester. All participants were free from tactile and proprioceptive deficits in their right hand and had normal or corrected to normal vision. Twenty participants had experienced the RHI before. Written informed consent was obtained from each

58 58 participant. The research was approved by the University of Manchester Ethics Committee Design Two testing sessions were completed seven days apart. In session one, participants took part in the RHI once and completed the subjective questionnaire measure. In session two, participants took part in the RHI three times and completed visual, tactile and proprioceptive outcome measures of the RHI. Equipment Participants sat at a table with their right index finger resting on a tactor (with a 1.6cm x 2.4cm vibrating surface, Oticon Limited, B/C 2-PIN) that was inlayed into the table surface, and their chin in a fixed height chin rest to maintain their midline body position throughout the experiment. The height of the chair could be altered to accommodate different body sizes. Participants wore a Sony Head mounted display (HMD-T2) which was used to display all stimuli during both the illusion and sensory tasks. The HMD has a diagonal field of view of 45 degrees and a display resolution of 1280 x 720 pixels. The straps of the HMD were replaced with adjustable headgear straps used in welding helmets. This modification improved the stability and comfort of the HMD. A schematic diagram of the experimental set up is shown in Figure 2.3.

59 59 Figure 2-3 Schematic representation of the augmented reality RHI. The line drawing shows the participant s position relative to the video image presented through the HMD. A snap-shot of the video is presented as an example of the video image seen during the illusion. The table seen through the HMD (i.e. in the video) is positioned so that the rubber hand protrudes over the table edge. The participant s hand is positioned 10cm rightwards of the rubber hand seen though the HMD, and the participant s index finger rests on a tactor inlayed into the table surface. RHI video During the illusion, participants watched a video of a rubber hand being stroked for 2.5 minutes. In line with previous research the stroking pattern did not appear predictable to the participant as this is thought to make the illusion more compelling (Armel and Ramachandran, 2003). Strokes covered the entire hand area and were a mixture of long strokes of the entire finger, short strokes of half the finger and taps to the end of the finger. To ensure that the video closely matched the participant s unmediated view of the laboratory (i.e. without the HMD), the video was recorded from the participant s seated position using a 50mm focal length which gives a similar perspective to the human eye. The video contained three differences from the true laboratory setting. The table was shifted backwards towards the camera so that there was no table surface under the location of the tactor. A rubber hand to replace

60 60 the participant s hand was placed over the edge of the table so that it was not supported. An image from the video is displayed in figure 2.3. There was a 10cm horizontal separation between the index finger of the rubber hand seen in the video and the physical location of the tactor. All images of the lab and rubber hand were captured from the same location so that image properties were consistent. A guide image displaying relevant locations of the table, rubber hand and tactor was created to ensure that all images displayed were spatially aligned with the true environment. To prevent image distortions, the display properties (e.g. image dimensions and pixel density) of all devices and programs were matched to the settings used to capture the images. Sensory outcome measures: materials and procedure The visual, tactile and proprioceptive sensory judgement tasks were all run via Eprime software. Each task followed a two alternative forced choice Parameter Estimation by Sequential Testing procedure (PEST; Taylor and Creelman, 1967). The PEST technique determines the value of thresholds in as few trials as possible. It is a binary forced choice procedure that makes online judgements about the next step size using the participant s previous responses. The algorithm decides whether to change the next stimulus intensity based on the participant s responses at the current level, taking into account the target probability. The step size was doubled after three responses in the same direction. After the first reversal in response direction the step size was halved (i.e., each time the participant s threshold level was passed smaller changes in stimulus intensity were presented). The procedure ended when a minimum step size was reached and the stimulus intensity presented at this stage was recorded as the participant s threshold judgement. Alternatively, the procedure ended when the maximum number of trials was reached. In this case the participant s threshold judgement was approximated by averaging over the final trials. Further details about each of the sensory judgement tasks are described below. All responses were collected using laptop buttons marked with raised tactile coverings.

61 61 Figure 2-4: Figure 2-5: Examples of photo stimuli presented in visual (left) and proprioceptive (right) outcome measures. In the visual task participants were presented with images made by blending a photograph of the participant s hand and the rubber hand seen during the RHI. Proportion of rubber hand in each image (left to right): 100%, 75%, 50%, 25% and 0%. In the proprioceptive task participants were shown images of an arrow marker on a table. The distance in millimetres between arrow marker and the participant s index finger in each image (left to right): 160, 100, 0, -100, Visual task: 100 images were created by blending a photograph of the rubber hand (seen during the RHI) and a photograph of the participant s hand (taken at the end of session one), using the open source image blending software SmartMorph (MeeSoft; examples are shown in Figure 2.5. An image was presented for one second followed by the question is this more like the rubber hand or more like your own hand?. The available responses were either more like the rubber hand or more like my own hand. The first photograph was always 100% rubber hand. The initial step size represented a 10% decrease in the proportion of rubber hand in the photo and the program completed once a step size of 1% was reached. The target probability was 50% so the program was complete when the stimulus level at which the participant responded more like my own hand 50% of the time was reached. Alternatively, to avoid excessive testing time, if the maximum number of 100 trials was reached then the stimulus level used in the last 50 trials was averaged to give the output value. Proprioceptive task: 65 photographs showing a red arrow on top of a table surface were taken. The table was larger than the table in the laboratory. In the first photograph, the arrow was positioned so it pointed towards the participant s midline and centre of the HMD. In each photograph the position of the arrow was moved rightwards by 5mm covering a distance of 320mm, examples are shown in Figure

62 During the task an image was presented for one second followed by the question is the arrow positioned to the left or right of your index finger?. Participants could respond left or right using the keyboard buttons. The initial step size represented a 25mm rightward shift in the arrow position and the program completed once a step size of 5mm was reached. The target probability and maximum number of trials were identical to the visual task. Tactile task: participants were presented with vibrotactile stimuli of a range of intensities via the tactor. The tactor was controlled via the volume output in Eprime (Eprime output value: 0 (max volume) to (min volume) which was used to vary the stimulus intensity. On each trial participants saw a number 1 and 1000ms later a number 2. A vibration followed one of these numbers. Following the stimulus presentation the question when did you feel the vibration? appeared on screen. Participants could respond time 1 or time 2. The procedure started with a supra-threshold vibration (0, max volume output in Eprime). The initial step size was 800 db and the program was complete once a step size of 50 db was reached. This target probability was 75%, since this task was a 2AFC, this represents correctly detecting the tactile stimulus 50% of the time. Alternatively, if the maximum number of 250 trials was reached then the stimulus level used in the last 100 trials was averaged to give the output value. The maximum number of trials in this task was larger in this task than the other two because the tactile scale contained substantially larger number of stimulus intensities. Questionnaire measure RHI Questionnaire: Participants completed a questionnaire to assess their subjective experience of the RHI. The items are presented in table 2.4. Items 1-7 were taken from the questionnaire used in the original Botvinick and Cohen (1998) study. Items 1-3 were averaged to form an illusion score and items 4-7 were averaged to form a control score. Items 8-16 were included to provide data for Study five but are not analysed in this study. Responses were collected on a 7-point Likert scale ranging from -3 to +3. Items were presented in a randomised order for each participant. An example questionnaire is presented in Appendix A.

63 Procedure Session one: Participants were seated in the equipment and the height of the chair was adjusted to a comfortable level. They then put on the HMD and adjustments for image quality were made. The participant s right hand was placed on the table with their index finger resting on the tactor, this acted as a position marker throughout the experiment. Participants took part in the RHI and then completed the embodiment change questionnaire. During the RHI they watched the pre-recorded video whilst their hand was stroked in the same pattern shown in the video. The experimenter listened to an audio guide which signalled the beginning and end of each stroke so that the visual and tactile touch cues occurred simultaneously. Once they had completed the questionnaire, a photograph of the participant s hand was taken so that it could be merged with a photograph of the rubber hand. Sensory outcome measures were collected on a different day. This provided time to blend the photograph of the participant s hand and the rubber hand for the visual task, and reduced the burden on the participant as the second session had a duration of 1.5 hours. Session two: The experimental set up was identical to session one except that the participant s left hand was placed onto marked keyboard buttons which were positioned outside of the field of view seen through the HMD, i.e., so it was not positioned under the HMD image. The participant then completed each of three sensory blocks: visual, tactile and proprioceptive. Within each block, the participant completed a baseline sensory measure followed by the RHI, finally they completed the sensory measure for a second time as a post-illusion measure. There was a 15 minute break between each sensory block to minimise carry-over effects and the order of the blocks was counterbalanced across participants. A brief calibration task was completed before the baseline visual and proprioceptive measures. Before beginning the proprioceptive task, participants were shown the midline arrow photograph and asked to indicate whether it was on the left or right side of their index finger. This was repeated for the furthest right arrow location. This ensured that the participant felt that their finger was positioned somewhere in-between the two furthest arrow locations. Before beginning the visual task, participants were shown the photograph of the rubber hand and the photograph of their own hand. This

64 Study five items Control Standard illusion 64 ensured that the participant could recognize their own hand and provided them with a benchmark of what the rubber hand looked like. Table 2-4: RHI Questionnaire items 1...the rubber hand was my hand* the touch I felt was caused by the touch I saw on the rubber 2 hand* 3 I felt the touch in the location that I saw the touch it seemed as though my real hand was drifting towards the 4 rubber hand 5 it seemed as though I might have more than one right arm it seemed as though the touch I felt came from somewhere in 6 between my hand and the rubber hand it seemed as though (visually) the rubber hand was drifting 7 towards my hand 8...I could move the rubber hand* 9...the experience of my hand was less vivid than normal*...there was some form of association between myself and the 10 rubber hand* there was some form of connection between myself and the 11 rubber hand* 12...the rubber hand was an object that had become joined to me* 13...I became more aware of the attributes of the rubber hand* 14...the touch I saw and the touch I felt were the same event* 15...the touch I saw caused the touch that I felt* I expected to feel a touch when I saw the experimenter s finger 16 approaching* Note: except item 16 all items begin with It seemed as though. *items from the Embodiment Change Questionnaire

65 Results RHI Questionnaire To assess whether participants had reported the typical experience of the RHI a paired t-test was conducted on the aggregated illusion and control scores. The analysis showed that the ratings of illusion items were significantly higher than the ratings of control items (1.97 vs ), t(35)= , p<.001. This suggests that participants were more likely to positively endorse items that specifically described embodiment change during the RHI than items describing odd body experiences not specifically associated with the RHI Sensory outcome measures Sensory change scores were obtained for each participant by subtracting the baseline score from the post illusion score. Tactile thresholds could not be obtained for three participants either due to non-compliance with task instructions or because the participant could not detect the initial stimulus level. A proprioceptive threshold could not be obtained from one participant as they perceived their right index finger to be positioned to the left of their midline before starting the task. These participants were excluded from the relevant analysis. Values that were greater than 2.5 standard deviations from the mean were removed from the data set. Following this procedure one participant was removed from the tactile analysis. Following this exclusion all data sets had a normal distribution. The mean and standard deviation values for these as well as the pre and post illusion scores for each sensory outcome are presented in Table 2.5. Three one sample t-tests were conducted to assess whether the sensory change scores associated with each modality were significantly different to zero. A Bonferroni corrected criterion for significance of p=.0167 was used to correct for multiple comparisons. There was a significant visual change such that participants identified photographs containing a higher proportion of rubber hand as being like their own hand after the illusion than before, t(35)=-5.04, p<.001, d=0.67. There was also a significant tactile change such that participants could detect vibrations of a lower intensity after the illusion, t(31)=2.941, p=.006, d=0.30, but changes to proprioceptive judgements were not significant, t(34)=-1.496, p=.144.

66 66 Table 2-5: Mean and standard deviation scores for visual, tactile and proprioceptive judgements before and after the illusion and change scores within these modalities. Scores in the visual modality reflect the mean proportion of rubber hand judged to be like the participant s own hand, higher values indicate a larger proportion of rubber hand. In the tactile modality higher values indicate vibrotactile stimuli of a higher intensity. Scores in the proprioceptive modality reflect the position in which the participants located their index finger, higher scores indicate more leftwards positions closer to the location of the rubber hand. Modality N Pre-illusion Post-illusion Change score Visual (11.37) (13.31) 8.58 (9.79) Tactile) (447.50) (515.33) (289.57) Proprioceptive (60.02) (64.96) 6.13 (32.32) Note: metrics for each test are - visual (% rubber hand), tactile (vol. output) and proprioceptive (mm) Discussion Participants reported the typical subjective experiences of embodiment change associated with the RHI. Sensory changes to vision and tactile sensitivity were measured. After the illusion, participants recognised images with a higher proportion of rubber hand as being like their own hand and could detect vibrations of a lower intensity. The mean values for proprioceptive judgements before and after the illusion indicated that participants perceived their hand to be closer to the location of the rubber hand after the illusion, but this change was not statistically significant. The proprioceptive outcome of the RHI has been extensively replicated, as such, it is unlikely that no proprioceptive change occurred during the study. The lack of effect in this study is most likely due to the AR approach used. To create an AR display, pre-recorded videos were presented via a HMD. As the videos were pre-recorded there was no association between a participant s head movements and the image displayed. Visual-motor association is possible in VR and creates a compelling sense of presence but is not possible with the static display set up in this study. For this

67 67 static display, the participants head movements were minimized using a chin rest. The chin rest maintained the participants body midlines but did not prevent all head movements. Any head movements would alter the relationship between the arrows seen during the proprioceptive task and the location of a participant s hand, introducing unsystematic errors to the measurement. To test whether head movements may have affected the measurement of proprioceptive drift, we conducted a further study where the HMD was fixed to the table surface preventing any changes to the visual field during the collection of proprioceptive measurements Study B: Method Participants 15 students (male: female = 7: 8, mean age = 27 years, age range = years, Right-handed: Left-handed = 14: 2) were recruited from the University of Manchester. All participants were free from tactile and proprioceptive deficits in their right hand and had normal or corrected to normal vision. Seven participants had experienced the RHI before. Written informed consent was obtained from each participant. The research was approved by the University of Manchester Ethics Committee Materials The experimental set up was the same as in study A except that the HMD was fixed to the table so that the height and position of the HMD was the same for all participants and could not be moved in any direction. The illusion and proprioceptive task was identical to Experiment one Procedure Participants were seated in the equipment and completed initial visual checks. They completed the proprioceptive task, took part in the RHI and then completed the proprioceptive task again. The procedure used was the same as the proprioceptive block used in Study A.

68 Results To assess whether participants had reported the typical experience of the RHI a paired t-test was conducted on the aggregated illusion and control scores. The analysis showed that the ratings of illusion items were significantly higher than the ratings of control items (1.8 vs ), t(14)= 7.58, p<.001. This suggests that participants were more likely to positively endorse items that specifically described embodiment change during the RHI than items describing odd body experiences not specifically associated with the RHI. A one sample t-test was conducted to assess whether the proprioceptive difference score (mean = 24.33) was significantly different to zero. Results showed that the participants felt their hand to be significantly closer to the location of the rubber hand after the illusion than before, t(14)= 2.506, p=.025, d= Discussion of studies A and B The results from these two studies show that participants reported the subjective experience typically associated with embodiment change in the RHI and that following the illusion significant changes to visual, proprioceptive and tactile aspects of embodiment were present. Participants judged the appearance of the rubber hand as being more like their own hand, they judged the location of their hand to be closer to the location of the rubber hand and they could detect vibrations of a lower intensity demonstrating an increase in tactile sensitivity. These results indicate firstly that multiple sensory outcomes of embodiment change can be measured within a single illusion context, and secondly that presenting pre-recorded videos via a HMD is a feasible AR approach suitable for embodiment research. With these developments additional lines of embodiment research can be pursued. The proprioceptive change is a well-established effect of the RHI and visual changes have been demonstrated in the enfacement illusion (Botvinick and Cohen, 1998; Tajadura-Jiménez at al., 2014; Sforza et al., 2010; Sforza et al., 2010). The present study replicates and extends these findings by measuring both effects in a single illusion context as well as an additional measure of tactile change. Visual and tactile changes have been previously reported in qualitative investigations of the RHI

69 69 (Lewis and Lloyd, 2010). The present results support the presence of these modality specific changes. Questionnaire items describing visual and tactile changes, such as the rubber hand resembled my own hand and the skin on my hand was turning rubbery, are often used as control items i.e. items describing body experience not associated with embodiment change in the RHI (e.g. Kaplan, Enticott, Hohwy, Castle and Rossell, 2014; Ehrsson, Holmes and Passingham, 2005; Ehrsson, Rosén, Stockselius, Ragnö and Köhler, 2008). Although these items are often only minimally endorsed they are not suitable as control items because they relate to a sensory effect which can be measured. The presence of visual, tactile and proprioceptive changes in the RHI demonstrate that there is a top-down expectation that body signals should to some extent be consistent, that is, that sensory impressions should relate to a single body object. In this way sensory contributors to embodiment are unified, however, each one confers a specific type of information about what the body is like and so also has a unique contribution to embodiment. The contribution of each sensory modality has so far been investigated in isolation but by using multiple sensory outcomes it is possible to compare how each sensory modality contributes to embodiment. This provides a broader specification of the sensory interactions that give rise to embodiment and the opportunity to further specify previous findings about how embodiment functions. For example, previous research indicates that proprioceptive measures are comparable over a six-month interval. It is currently unclear whether this result is generalizable to visual and tactile aspects of embodiment, in which case embodiment as a whole would be perceptually stable, or specific to the proprioceptive component of embodiment. Further research using multiple sensory outcomes allows the characterisation of multisensory embodiment whilst preserving what is important about individual modalities. Individual modality measures can then be selected on the basis of how they contribute to embodiment overall. Examining multiple sensory outcomes also permits further specification of individual differences in embodiment. Some individuals are more susceptible to sensory effects of the RHI. These individuals may rely more on external body signals to compensate when other aspects of embodiment such as interoceptive signals or internal body representations are less reliable (Tsakiris et al., 2011; Tajadura-Jimenez and Tsakiris, 2013). Certain individuals may rely on some external body signals more than others. It has been

70 70 suggested that individuals with autism rely more on proprioception because this is a more reliable sensory modality (Paton, et al. 2012; Cascio et al., 2012). The relative weighting of sensory contributors to embodiment can be assessed by comparing individuals across multiple sensory outcome measures. This would provide further specification of the sensory profile associated with specific clinical groups. A novel outcome measure of tactile sensitivity change was used in this study. Participants were more sensitive to vibrotactile stimuli following the RHI. This novel finding suggests that tactile adaptation can also occur during embodiment change. This finding is most likely due to the tactile discrepancy present in the illusion context. The participants hands rested on a table surface whereas the rubber hand was unsupported. This visual-tactile conflict regarding the presence of tactile signals arising from object contact could potentially be resolved by increasing the gain of tactile signals, resulting in an increase in tactile sensitivity. An alternative explanation could be that this change is due to simply attending to the body or practice effects. However, a previous study examining tactile sensitivity following the RHI suggests that is not the case. Participants took part in the standard version of the RHI where there was no tactile discrepancy, and following the illusion showed a reduction in tactile sensitivity (Zopf, Harris and Williams, 2001). This suggest that in the absence of a tactile discrepancy, attending to one s body during embodiment change leads to a reduction in tactile sensitivity rather than the increase in tactile sensitivity reported in the present study. Further research is needed to examine how the tactile changes observed here are caused (e.g. comparing asynchronous and no stroking conditions), as well as further investigation of how touch contributes to embodiment as a whole. Previous research has suggested that touch is specifically related to affective body experience (Crucianelli, Metcalf, Fotopoulou and Jenkinson, 2013; Lloyd, Gillis, Lewis, Farrell and Morrison, 2013). The tactile measure presented here may provide another tool for investigating other ways that touch information contributes to embodiment. That is was possible to both replicate and extend existing measures of the RHI in this study demonstrates that embodiment research can be conducted using this simple approach to AR. Pre-recorded images are ideal for the display of skin appearance and texture because they have a photo-realistic image quality. These images can be augmented in a realistic way using image processing software prior to the session.

71 71 Using this approach, images of the rubber hand and the participant s hand were blended together for the visual task in this study. Blended images were presented over the location of the participants hand using the HMD so that the visual task was conducted without disrupting the illusion or creating an additional sensory discrepancy. With this approach to AR sensory features of interest can be varied whilst other aspects of the participant s body are accurately reproduced. This provides a high degree of control over the sensory context in which embodiment change occurs allowing the causes of embodiment change to be precisely defined. A further advantage of this approach is that it does not require the technical and financial investment necessary for other approaches such as real-time image augmentation (e.g. MIRAGE box) or VR. However, the utility of using a prerecorded image display is largely limited to studies of static embodiment because changes in image display cannot be associated with the participant s head and body movements. Where possible the participant s body movements should be minimized so that they cannot affect the data collected in the study. The quality of proprioceptive data collected was improved by fixing the HMD to a table surface in study 2. This AR approach could be developed further by using two cameras during image capture for stereoscopic image display and incorporating additional image processing software for image augmentation. This study demonstrates that embodiment can be studied using multiple sensory outcomes of the RHI. There are a number of potential applications for this, in particular, identifying how individual modalities contribute to embodiment and further specification of individual differences in embodiment. 2.3 Summary of studies one and two This chapter assessed the feasibility of using a simple augmented reality approach to elicit the RHI. In studies one and two, pre-recorded videos of the RHI were presented via a HMD to elicit the illusion, and photographic stimuli were used in sensory outcome measures. The results of study one showed that this simple approach to augmented reality was suitable for eliciting changes to subjective embodiment in the RHI, and suggested that participants also experienced changes to tactile sensitivity when a tactile discrepancy was present in the RHI context. In study

72 72 two, the augmented reality approach was developed further so that it could be used to measure sensory outcomes of the RHI. The results of study two showed that once the position of the HMD was fixed into place, rather than fixing the participants body midline with a chin rest, it was possible to measure visual, tactile and proprioceptive outcomes of the RHI. Collectively these results suggest that it is feasible to measure multiple sensory outcomes of the RHI using a simple augmented reality approach. To examine the utility of the methods developed in this chapter, they were used to investigate veridical changes to embodiment in studies three and four of chapter three. The results of study one also suggested that when a tactile discrepancy is present in the RHI context, some participants may report body ownership-type changes to embodiment whilst others may report body extension-type changes to embodiment. These experiences, though closely related in the RHI and both associated with perceived causality between the seen and felt touch used to elicit the RHI, were distinguishable in the qualitative analysis as the extent that participants could perceive their own hand as a separate object from the rubber hand. To assess the distinction between body-ownership and body extension-type experiences in the RHI, the questionnaire measure was expanded to include items describing bodyextension, body ownership and perceived causality. The underlying factor structure of the questionnaire was assessed using confirmatory factor analysis in Chapter four. Data for this analysis was obtained by aggregating data collected in studies three and four of Chapter three.

73 73 3 Chapter 3: Measuring veridical changes to embodiment using multiple sensory outcomes of the Rubber Hand Illusion As discussed in Chapter one, body experience typically occupies the background of awareness. The RHI is a useful tool for investigating embodiment because it brings body experience into the foreground of awareness. However, in doing so the RHI elicits a form of body experience which is different to its usual state. To examine veridical changes to embodiment, multiple sensory outcomes of the RHI could be examined before and after body perception training. By comparing sensory measures at two time points, changes in susceptibility to each sensory outcome may reveal what types of body perception activities produce changes to specific aspects of the multisensory process of embodiment. This possibility was assessed in this chapter across two studies. Study three examines changes to embodiment following body scan meditation. This is a form of internal body perception training because participants learn about their internal body sensations. Study four examines changes to embodiment following anatomical dissection training. This is a form of external body perception training because participants learn about the form of the human body as an object external to their body. Through comparison of internal and external forms body perception training, the studies in this chapter explore the potential for modality specific changes to embodiment. 3.1 Study three: The effect of Body Scan Meditation on embodiment Introduction As discussed in chapter one, embodiment arises from the integration of multiple sensory signals including vision, touch, proprioception and interoception. These sensory modalities can be broadly classed as interoceptive (i.e. arising from inside the body; Craig, 2009, Critchley, Wiens, Rotshtein, Ohman and Dolan, 2004, Damasio, 2010) or exteroceptive (i.e. arising from outside the body). These contributors to embodiment have been considered distinct as they are processed via different cortical pathways (see Craig, 2002 and Craig, 2003 for reviews). But following evidence of substantial crosstalk between these sensory channels (Critchley and Harrison, 2013), recent studies have begun to examine the influence

74 74 of interoceptive and exteroceptive body signals simultaneously using body illusions (e.g. Suzuki et al., 2013), with results being described under a single framework of predictive coding (Seth, 2013; Apps and Tsakiris, 2014; Fotopoulou, 2012; Clark, 2013). According to the predictive coding account, for the brain to make sense of a multitude of interoceptive and exteroceptive signals that arise from one s own body and interaction with the environment, it must attempt to infer the causes of sensory signals (Clark, 2013; Friston, 2009; Hohwy, 2013). To do this, current sensory signals are compared to existing models of causes, consisting of hierarchically organised predictions about how sensory signals tend to occur and change over time. When bottom-up signals and top-down predictions are consistent, current sensory signals are explained away under a specific model of the current cause of sensory input. A lack of consistency produces a prediction error which, using Bayesian optimal inference (for the mathematical details of Bayesian optimal inference see Friston, 2010), the brain attempts to minimise through either active or perceptual inference. During active inference, prediction errors are minimised through behaviour that generates new sensory input, whereas during perceptual inference prediction errors are minimised by updating predictions to match current sensory signals. Through the minimisation of prediction errors these mechanisms are proposed to maintain a stable sense of embodiment (Seth, 2013; Apps and Tsakiris, 2014; Fotopoulou, 2012; Clark, 2013). The predictive coding account has been used to explain changes to embodiment that occur during multisensory body illusions such as the Rubber Hand Illusion (RHI; Apps and Tsakiris, 2014). The synchronous touch used to elicit the RHI matches a top-down expectation about the body, yet generates a prediction error regarding what the body is (i.e. the recognizably fake RH is not expected to be part of the body). In explaining the cause of sensory input, a perceptual inference is made and the topdown model of embodiment is updated to include the rubber hand as a part of the body (Apps and Tsakiris, 2014). During this process unimodal prediction errors are also minimised through perceptual inference by incorporating the sensory features of the rubber hand into the model of embodiment. For example, when there is a separation between the location of the rubber hand and the location of the participant s hand, participants minimise the discrepancy by perceiving their hand to

75 75 be closer to the location of the rubber hand after the illusion. Based on the predictive coding account the subjective and sensory outcomes of body illusions, such as the proprioceptive outcome of the RHI (Botvinick and Cohen, 1998) and the visual outcome of the enfacement illusion, indicate that perceptual inference has occurred. Collectively, these effects are a demonstration of how bottom-up sensory signals (seen and felt touch) are compared to top-down multisensory predictions, producing prediction errors at multiple levels of the processing hierarchy relevant for bodily perception, including high level subjective self-awareness and low level unisensory changes to perceived self-location and self-appearance. As discussed in chapter one, susceptibility to the RHI is influenced by the accuracy of interoceptive perception, such that individuals with low interoceptive sensitivity are more susceptible to the subjective and perceptual effects of the illusion. This negative relationship suggests that individuals weight the contribution of body signals according to their reliability (Tsakiris et al., 2011; Tajadura-Jimenez and Tsakiris, 2013). Within the predictive coding account, prediction errors arising from current sensory input are associated with precisions which determine their influence on subsequent hierarchical processing, for example, modulating the extent that prediction errors are resolved through perceptual inference (Friston et al., 2010). When interoceptive prediction errors have low estimates of precision, the contribution of exteroceptive signals has a disproportionate contribution to embodiment. It has been suggested that this imbalance may contribute to disordered body perception in a number of clinical disorders such as anorexia (Seth, 2013). A more balanced contribution of interoceptive and exteroceptive body signals may be achieved by increasing the precision of interoceptive prediction errors during hierarchical inference. The precision of prediction errors can be optimised by attention (Feldman and Friston, 2010), specifically, by increasing attention to interoceptive body signals during meditative practices such as mindfulness (Sel, 2014). Mindfulness training encourages participants to become more aware of their present situation and to disengage from habitual cognitive appraisals (Farb, Anderson and Segal, 2012). It has been shown to increase subjective interoceptive awareness and cortical responses of interoceptive attention (Mehling et al., 2013). Moreover, it has proven to be an effective treatment for a number of clinical disorders, particularly those characterised by deficits of self-perception like anorexia

76 76 (for reviews, see Baer, 2003; Chiesa and Serritti, 2011; Wanden-Berghe Sanz-Valero and Wanden-Berghe, 2010). The mechanisms underlying the therapeutic benefits of mindfulness are currently underspecified. One possibility is that increasing interoceptive attention may increase the precision of interoceptive prediction errors, altering the contribution of interoceptive signals during multisensory integration. This hypothesis is supported by recent neuroimaging findings. Eight weeks of mindfulness training produced changes to the insular cortex. Specifically, participants showed increased functional connectivity between the posterior insula and the anterior insula (Farb, Segal and Anderson., 2013). The posterior insula has been identified as a brain region for primary interoceptive processing (Flynn, Benson and Ardila 1999) and the anterior insula integrates interoceptive and exteroceptive signals regulating the direction of attention between internal and external sensory signals (Critchley, 2005; Craig, 2009; Mutschler et al., 2009). Increased connectivity between these regions suggests that mindfulness training increases the contribution of interoceptive signals during the process of integration (Farb et al., 2013). Increasing the contribution of interoceptive signals during integration may alter the other body signals such as vision, touch and proprioception, potentially improving the accuracy of body perception. Initial stages of mindfulness training aim to improve awareness of the body using Body Scan Meditation (BSM). During BSM participants are instructed to focus on various parts of their body and notice, without judgement or interpretation, the sensations that are located there (Kabat-Zinn, 1990; Segal et al., 2002). Three aspects of BSM training are likely to influence the way that interoceptive signals contribute to the multisensory process of embodiment. Firstly, during BSM practitioners direct their attention to body sensations leading to increased interoceptive awareness (Mehling et al., 2013). Secondly, by deliberately switching attention to internal sensations at different body locations, practitioners improve their ability to selectively attend to interoceptive signals (see Cahn and Polich, 2006, Chiesa, Calati and Serretti, 2011; Lutz, Slagter, Dunne and Davidson, 2008 for reviews). And finally, by openly monitoring their body sensations without judgement or interpretation, the influence of top-down expectations on the interpretation of interoceptive signals can be reduced so that they can be adaptively modified (van den Hurk, Janssen, Giommi, Barendregt and Gielen, 2010). By increasing attention to bottom-up sensory signals, rather than explaining away

77 77 sensory signals according to existing causal models, these changes are thought to help individuals learn about their own body and the world (Farb et al., 2015). The present study aims to assess whether BSM can improve the interpretation of interoceptive and exteroceptive signals during body perception. If BSM improves the precision of interoceptive prediction errors, then the contribution of other body signals such as vision, touch and proprioception during body perception should be altered after a period of BSM training. Though previous research has shown that BSM training alters interoceptive awareness, it is unclear precisely how it affects the perception of other body signals as few studies have addressed this directly. Some previous findings suggest that meditative practices can improve the accuracy of a number of sensory modalities relevant to body perception. For example, increased tactile sensitivity has been observed in experienced tai chi practitioners (Kerr et al., 2008), improved proprioception has been observed following mindfulness-based cognitive therapy for individuals with depression (Michalak, Troje and Heidenreich et al., 2011) and improved visual discrimination has been observed following mindfulness based stress reduction (MacCoon, MacLean, Davidson, Saron and Lutz et al., 2014). To examine changes to the perception of multiple body signals, the accuracy of interoceptive perception was assessed in a heartbeat detection task and susceptibility to the RHI was measured using multiple sensory outcomes including measures of vision, proprioception and touch (using the method outlined in study two). We predicted that BSM training would improve the reliability of interoceptive prediction errors and reduce susceptibility to the subjective and perceptual effects of the RHI Method Participants An advertisement was placed on the University of Manchester experimental participation scheme website, inviting participants to take part in a multimodal investigation of body perception. Meditation and mindfulness were not mentioned in the advert to avoid biasing recruitment towards individuals who were interested in meditation. 47 psychology undergraduates were recruited in total (male: female = 6: 41; mean age= 20 years, age range =18-35 years, handedness: right-handed: left-

78 78 handed = 43: 4). The majority of participants had not heard of the RHI before the study (68%). One participant had taken part in the RHI before the study. Participants were required to have normal or corrected-to-normal vision, with no tactile or proprioceptive deficits, or substantial meditative or mindful practice. This was classified as >6 months of weekly practice of mindfulness, meditation, yoga or tai chi. Written informed consent was obtained from each participant. The research was approved by the University of Manchester Ethics Committee Design Participants attended two experimental sessions conducted one week apart, with six days of training in between. In session one, participants initially completed the IRI questionnaire and the heartbeat detection task. Participants then completed the RHI three times. Participants completed a baseline sensory task before each period of the RHI and this task was repeated as soon as the RHI was finished. Three different sensory tasks were completed in total to assess changes to vision, touch and proprioception. The presentation order of the tasks was counterbalanced and participants rested for 10 minutes before starting the next sensory task. Following the final sensory task, participants completed a subjective questionnaire. Further details of the materials and procedures for the RHI and sensory tasks are described in study two A. After the subjective and sensory measures for session one were complete, an Eprime program was used to pseudo-randomly allocate participants to either the experimental group or the control group at the end of the first session, based on their participant number. The program was pre-programmed by a colleague (as recommended by Shultz and Grimes, 2002). This ensured that the experimenter was unaware of the participants group and counterbalancing allocation until the end of the first session. Finally, participants listened to a 15 minute audio guide. Participants in the experimental group listened to a BSM guide (BSM group) and participants in the control group listened to a human anatomy guide (AC group). Session two followed a similar procedure except the audio guide was completed at the start of the session and a short interview was conducted at the end of the session.

79 79 Participants completed the same order of sensory tasks in each session. Fig 3.1 shows the study design and procedure Materials Heartbeat Detection task Interoceptive sensitivity was assessed using a mental tracking task (Schandry, 1981). Participants were instructed to focus on the internal sensation of their heartbeat and count the number of heartbeats that occurred within a specified length of time. Four trials of different durations (25, 35, 45, 100 seconds) were conducted; the order of presentation was counterbalanced across participants. Participants were unaware of the trial durations and were not given feedback on their performance. Trials were timed using a computer program. Participants were instructed to start counting when the experimenter said go and stop counting when the beep occurred. During this period the experimenter took the participants pulse to count the actual number of heartbeats that occurred. Interoceptive sensitivity was calculated as 1/4 Σ [1 - ( recorded heartbeats counted heartbeats /recorded heartbeats)]. Higher values indicate greater interoceptive sensitivity. Questionnaire measures RHI Questionnaire (see table 3.1): 15 item questionnaire for assessing subjective experience of embodiment in the RHI. Items 1-7 were taken from the original Botvinick and Cohen (1998) study. Items 1-3 were summed to give an overall score for the illusion strength (standard illusion score) and items 4-7 were summed to give an overall control score. Items 8-15 developed from the qualitative study in chapter two were included. Responses to these items are not analysed in this study but are assessed in study five. The order of presentation for these items was randomised in each questionnaire. Responses to all items were collected on a 7-point Likert scale ranging from -3 (strongly disagree) to +3 (strongly agree). An example of this questionnaire is presented in Appendix B.

80 Study five items Control Standard illusion 80 Table 3-1: RHI Questionnaire items 1...the rubber hand was my hand* the touch I felt was caused by the touch I saw on the rubber 2 hand* 3 I felt the touch in the location that I saw the touch it seemed as though my real hand was drifting towards the 4 rubber hand 5 it seemed as though I might have more than one right arm it seemed as though the touch I felt came from somewhere in 6 between my hand and the rubber hand it seemed as though (visually) the rubber hand was drifting 7 towards my hand 8...I could move the rubber hand*... I could not separate the experience of my own hand from my 9 experience of the rubber hand* there was some form of association between myself and the 10 rubber hand* there was some form of connection between myself and the 11 rubber hand* 12...the rubber hand was an object that had become joined to me* 13...I became more aware of the attributes of the rubber hand* 14...the touch I saw and the touch I felt were the same event* 15...the touch I saw caused the touch that I felt* Note: items begin with It seemed as though. *items from the Revised Embodiment Change Questionnaire

81 81 Figure 3-1 Study design and procedure The Interpersonal Reactivity Index (IRI; Davis 1980): 28 item questionnaire for assessing multiple dimensions of trait empathy using four separate scales including perspective taking (7 items, e.g. I try to look at everybody's side of a disagreement before I make a decision. ), empathic concern (7 items, e.g. I often have tender, concerned feelings for people less fortunate than me. ), personal distress (7 items,

82 82 e.g. I sometimes feel helpless when I am in the middle of a very emotional situation. ) and fantasy Scale (7 items, e.g. I daydream and fantasize, with some regularity, about things that might happen to me. ). Responses were collected on a 5-point likert scale ranging from does not describe me well to describes me very well. The scale has been found to be reliable in past research (Davis, 1980). Responses to this questionnaire are not presented in this study but are assessed in study five. An example of this questionnaire is presented in Appendix C. Training materials and procedure Participants in the BSM group listened to two versions of a fifteen minute body scan meditation exercise edited from a longer script (used previously by MacIver, Lloyd, Kelly, Roberts, and Nurmikko, 2008 and Mirams, Poliakoff, Brown and Lloyd, 2013). Each recording began with a two minute introduction to purposes of body scan meditation. Participants were then instructed to focus their attention on different parts of their body, starting with the sensations of breathing in the chest, nostrils and throat. In version one, participants were instructed to focus their attention on their left and right legs, feet and toes, pelvis, hip bones, and abdomen, lower and upper back and chest. In version two participants were instructed to focus their attention on their neck and shoulders, left and right arms, hands and fingers, the head and face, eyes and mouth. Throughout each recording participants were instructed to notice any sensations in each of the body parts and to observe what was happening with their thoughts and body, moment by moment, without judgement or criticism. They were also periodically reminded to re-direct their attention back to the present moment, if their thoughts had begun to wander. At the end of each recording, participants were instructed to focus once again on sensations of breathing. Participants were not explicitly instructed to focus on their heartbeat in either version. Participants in the BSM group listened to version one at the end of day one and version two on day seven. During the training period, participants practiced body scan meditation once a day by listening to an audio guide. To help maintain task engagement participants were instructed to alternate between the two versions so that they did not listen to the same guide on two consecutive days.

83 83 Participants in the AC group listened to a 15 minute excerpts from Basic Human Anatomy: The Beauty of Form and Function by Professor John K. Young (Recorded Books, LLC, 2008). Audio recordings of stories or educational texts have been used as control interventions in a number of other meditation studies (e.g., MacCoon, MacLean, Davidson, Saron and Lutz et al., 2014; Erisman and Roemer, 2010). Participants listened to 15 minutes of the introductory section of the audiobook at the end of day one, followed by seven 15 minute clips selected from different parts of the audiobook. Clips were selected so that participants listened to information about many different areas of the body including the head and face, arms, hands, spine, chest, hips, legs and feet. Unlike participants in the BSM group, participants in the AC group were not instructed to attend to body sensations. Before listening to an audio guide, participants in each testing session were instructed to pay attention as much as possible and to close their eyes throughout. Participants listened to the audio guide through headphones to minimise distraction noises. At the end of the session on day one participants were given CD recordings of the audio guide that they had been assigned. They were instructed to listen to a guide on each of the days until the second session on day 7 and that they should try to recreate similar listening conditions by sitting in a private place where they are unlikely to be disturbed, listen through headphones, close their eyes and pay attention as much as possible. To encourage task adherence, participants were informed that they would be asked some general questions about the audio guide but that this was not a test, nor would their performance be assessed. At the end of the testing session participants completed an interview debrief about their experiences listening to the audio guide. Participants were asked to describe how and when they listened to each audio guide and how many they had listened to in total. Participants were further asked what their experience was like and whether they had noticed any changes to their body experience during or after listening to an audio guide. The debrief interview was audio recorded and transcribed verbatim but this interview data is not presented in this thesis.

84 Results 41 participants returned for session two. Details for each of the two groups can be seen in table 3.2. In session two, the sensory data from two participants was lost due to technical error, and the visual data from a third participant could not be collected following their withdrawal from the study due to illness. All sensory data was normally distributed. Data was assessed for the presence of extreme values >2.5 standard deviations from the mean. This identified a number of univariate outliers which were removed from the relevant statistical tests as follows, session 1: visual (N=1), proprioceptive (N=2), session 2: visual (N=2), proprioceptive (N=1). Questionnaire data was non-normal and analysed using non-parametric tests. On average each group reported listening to seven audio guides during the training interval. As participants listened to an audio guide at the end of session one and at the beginning of session two, each participant verifiably listened to a minimum of two audio guides RHI effects in session one and two Two Wilcoxon signed rank tests were conducted to assess whether participants reported the subjective experience typical of the RHI. The results showed that illusion items were affirmed more than control items in session 1, Z=-5.826, p<.001, r=.85 (N=47), and session 2, Z=-5.52, p<.001, r=.86 (N=41). Six one sample t-tests were conducted to assess whether the RHI elicited changes to visual, tactile and proprioceptive judgements. After the RHI in session one participants judged blended images with a higher proportion of rubber hand as being like their own hand, judged the position of their index finger to be closer to the location of the rubber hand, but did not show a change to tactile sensitivity. After the RHI in session two participants judged blended images with a higher proportion of rubber hand as being like their own hand, judged the position of their index finger to be closer to the location of the rubber hand and showed a decrease to tactile sensitivity. Details of these statistical tests are shown in table 3.3.

85 85 Table 3-2: Demographic information for participants in session 2 for each group (BSM vs. AC) separately and combined N Age Handedness Gender Mean (Range) Left Right Male Female BSM (18-35) AC (18-21) Total (18-35) Table 3-3: One sample t-tests assessing changes to visual, proprioceptive and tactile sensory judgements in sessions 1 and 2 Sensory judgement Visual (% rubber hand) Mean change (SD) df t-statistic p-value d 8.40 (8.68) <.001*** 1.92 Session 1 Proprioceptive (mm) (38.54) <.001*** 1.09 Tactile (output volume) (350.76) Visual (% rubber hand) 3.70 (7.80) ** 0.94 Session 2 Proprioceptive (mm) (26.45) * 0.88 Tactile (output volume) (325.25) * 0.70 Note. Significance levels are marked as follows: p<.05*, p<.01**, p<.001***

86 Effects of training on body perception Heartbeat detection accuracy Two independent samples t-tests were conducted to assess the effect of training type on HBD accuracy using a Bonferroni corrected criterion of p=.025. In session 1 HBD accuracy was not significantly different in the BSM (mean=.61, SD=.2) and AC group (mean=.71, SD=.15), t(45)=1.923, p=.061, d=0.57. But following training, increases in HBD accuracy between sessions one and two were significantly larger in the BSM group (mean=.09, SD=.12) than the AC group (mean=.00, SD=.12), t(39)= , p=.028, d= Means and standard deviations are shown in table 3.4. These results suggest that participants who completed BSM could detect their heartbeat more accurately following the training interval. The AC and BSM groups were not significantly different at baseline, however, examination of the mean scores and standard deviations shown in table 3.4 suggests that some participants in the BSM group had lower HBD accuracy at baseline, in comparison to the sample in general. Subjective experience of the RHI Two Mann-Whitney U tests were conducted to assess the effect of training type on subjective experience of the RHI using a Bonferroni corrected criterion of p=.025. In session one, illusion scores were not significantly different in the BSM and AC group, Z=-.076, p=.94, r=.01. Following training, changes to illusion scores between sessions one and two were not significantly different in the BSM and AC groups, Z= -.764, p=.445, r=.12. Means and standard deviations are shown in table 3.4. These results suggest that participants in the AC and BSM group reported a similar experience of the RHI in session one and this outcome of the RHI was not influenced by the type of training completed.

87 87 Table 3-4: Means and standard deviations for the illusion score and HBD accuracy for each session (one vs. two) and each training group (AC vs. BSM) Illusion score Session 1 Session 2 Change AC 6.08 (1.84) 5.43 (3.31) (3.14) BSM 6.32 (2.87) 4.7 (3.34) (2.37) HBD AC.71 (.15).72 (.15).00 (.12) BSM.61 (.20).67 (.17).09 (.12) Immediate perceptual effects of the RHI Three 2x2 mixed ANOVAs were conducted to assess the effects of time (session 1 vs. session 2) and training type (BSM vs. control) on perceptual judgements in visual, proprioceptive and tactile tasks. Means and standard deviations are displayed in table 3.5. In the visual task, there was a significant main effect of time, F(1, 33)=11.971, p=.002, eta 2 =.266, but the main effect of training type, F(1, 33)=.009, p=.924, eta 2 =.00, and time x training interaction, F(1, 33)=1.37, p=.25, eta 2 =.04, were not significant. There were no significant main effects or interactions in the proprioceptive task, time: F(1, 34)=1.286, p=.265, eta 2 =0.036, training type: F(1, 34)=1.589, p=.216, eta 2 = 0.045, interaction: F(1, 34)=.895, p=.351, eta 2 =0.026; or the tactile task, time: F(1, 37)=1.156, p=.289. eta 2 =.03, training type: F(1, 37)=0.137, p=.713, eta 2 =.004, interaction: F(1, 37)=0.002, p=.963, eta 2 =.00. This indicates that susceptibility to the immediate visual effect of the RHI decreased after the training interval for both the AC and BSM group. That is, the increase in the proportion of rubber hand in images judged as being like the participants hand observed in session two was smaller than it had been in session one. But there were no effects of time or training on proprioceptive or tactile judgements. In general, the type of training did not influence susceptibility to the immediate effects of the RHI.

88 88 Table 3-5: Means and standard deviations of visual, proprioceptive and tactile change scores in sessions 1 and 2 for the BSM and AC groups Sensory judgement Session 1 Session 2 Visual (% rubber hand) AC 9.07 (7.38) 2.91 (6.74) BSM 7.29 (8.64) 4.25 (9.15) Proprioceptive (mm) AC (42.30) (29.85) BSM (36.63) (24.06) Tactile (output volume) AC (286.72) (323.59) BSM (385.90) (325.46) Long-term effects of the RHI To assess long-term effects of the RHI, pre-illusion scores obtained in session one and two were subtracted, means and standard deviations are displayed in table 3.6. Two one sample t-tests were conducted to assess whether there was a long-term visual effect in the BSM and AC groups, and a further independent samples t-test was conducted to assess whether the long-term visual effect was different in each group. These analyses used a Bonferroni corrected criterion of p= The results showed a significant long-term visual effect in the AC group, t(18)=3.088, p=.006, d=1.38, but not in the BSM group, t(15)=-0.603, p=.556, d= Under the strict Bonferroni criterion for significance, the independent t-test was not significant, t(33)=2.214, p=.034, d=0.75. However, the data does suggest a strong trend consistent with the initial one sample t-tests, showing a long-term effect in the AC group and a reduced long-term effect in the BSM group (mean= 6.6, SD= 9.32, mean= -2.07, SD= respectively). The results suggest that the control group showed a long-term visual effect and the BSM group showed a reduced long-term visual effect. The lack of a significant difference between the two groups, may reflect the extent that participants in the BSM group adhered to the training schedule. This possibility was explored further by evaluating relationship between training effects (i.e. the long-term visual effect and changes to HBD accuracy).

89 89 Two one sample t-tests were conducted to assess whether there was a long-term proprioceptive effect in the BSM and AC groups. The results showed no long-term proprioceptive effects in either the BSM group, t(18)=-0.405, p=.69, d=-0.18, or the AC group, t(17)=-0.02, p=.984, d=0.01. Two one sample t-tests were conducted to assess whether there was a long-term tactile effects in the BSM and AC groups. The results showed no long-term tactile effects in the BSM group, t(19)=-0.151, p=.882, d=0.22, or the AC group, t(19)=-1.129, p=.273, d=0.37. Collectively, these results show that BSM training did not alter susceptibility to the subjective or immediate perceptual effects of the RHI; however, it did lead to an increase in HBD accuracy and a reduction to the long-term visual effect of the RHI. Table 3-6: Means and standard deviations of long-term effects in visual, proprioceptive and tactile sensory modalities for the AC and BSM groups Visual Proprioceptive Tactile AC 6.60 (9.32) (59.32) (475.28) BSM (13.74) (45.56) (498.93) Relationships between BSM training effects Two Pearson s r correlations were conducted to assess the relationship between changes to HBD accuracy and the long-term visual effect of the RHI in each group. Data was assessed for extreme values using inspection of Mahalanobis Distance scores (Mahalanobis, 1936). This analysis identified two multivariate outliers with a Mahalanobis D² p<.001, one from each group, which were removed from the relevant tests. A significant moderate strength negative relationship between the two variables was observed in the BSM group, r=-.515, p=.05 (r 2 =26.52), but there was no significant relationship in the AC group, r=-.185, p=.462 (r 2 = 3.42). These results show that in the BSM group, improvements in HBD accuracy were associated with a reduction in the long-term visual effect of the RHI. Figure 3.2 displays the relationship between changes to HBD accuracy and the long-term visual effect of the RHI.

90 90 Figure 3-2: Scatterplots of the long-term visual effect of the RHI and changes to HBD accuracy. These scatterplots show that in the BSM group there was a negative correlation between improvements in HBD accuracy and the long-term visual effect of the RHI (r 2 =26.52), but no relationship between these variables in the AC group Discussion The present study aims to assess whether BSM can alter the contribution of interoceptive and exteroceptive signals during body perception. The results showed that BSM training improved heartbeat detection accuracy and reduced the long-term visual effect of the RHI. Changes to heartbeat detection accuracy and the long-term visual effect were negatively correlated, such that individuals with the largest improvements in heartbeat detection accuracy showed the smallest long-term visual effect. This effect was, however, modality specific as training effects were evident in the visual modality only while there were no changes to tactile or proprioceptive modalities. In addition to the primary finding that BSM alters how interoceptive and visual primary sensory signals contribute to embodiment, this study reports two further novel research findings. In particular, this is the first study to show a change to heartbeat detection accuracy due to mindfulness and the first to show a durable perceptual change following a multisensory body illusion.

91 91 After the RHI participants judged blended images with a higher proportion of rubber hand as being like their own hand. This result replicates the findings in study two and previous studies showing changes to the visual aspect of embodiment immediately after the enfacement illusion (Lewis, Gowen, Jones and Poliakoff, in prep; Mazzurega et al., Sforza et al., Tajadura-Jimenez et al., 2012; Tsakiris, 2008). Novel to this study is the finding that visual changes can persist for up to one week. That the participant s ability to distinguish the appearance of their own hand from the appearance of the rubber hand was reduced, even in the absence of illusory embodiment, shows a durable change to the internal representation of body appearance. It is unlikely that a visual effect was caused by the type of training completed in the control group because the audio extracts were purely factual with no activities of a sensory nature. Instead it seems that by embodying the rubber hand in session one, expectations regarding typical body appearance had been updated to include the features of the rubber hand, making it more difficult to distinguish the appearance of each hand at the beginning of session two. In contrast, no durable effects were present in the tactile or proprioceptive modalities supporting a tacit understanding that RHI effects are disrupted once the participant can move and use their hand again (e.g. McKenzie and Newport, 2015). Whereas tactile and proprioceptive changes recalibrate quickly after the illusion, visual changes have a more durable effect on body perception. That sensory changes in individual modalities operate over different timescales provides further evidence that the RHI influences two conceptual aspects of body perception, the body image and the body schema (Holmes and Spence, 2007). The body schema is characterised as a rapid, online system maintaining consistency between sensory components of body perception, whereas the body image is characterised as a more stable system of perceptions, attitudes and beliefs about the bodily self as an individual, situated within its environment and social context (Gallagher, 1986; 2012). The properties of tactile and proprioceptive changes could be considered consistent with the body schema. They occur rapidly and arise from internal consistency between sensory components of body perception. For example, a proprioceptive change occurs because there is a discrepancy between visual and proprioceptive signals both indicating hand location. In contrast, the properties of visual changes could be considered consistent with the body image. They are more stable over time and arise

92 92 from intra-sensory comparison between prior expectations about hand appearance and current visual signals, rather than consistency between sensory components. For visual changes the source of prediction error likely arises from stored representations of one s own body and bodies of others considered to have a similar appearance. A number of researchers argue that the representation of the self is intersubjectively constructed (Gallagher, 2012). The influence of social factors upon body perception has been examined in a number of studies. For example, the effects of the enfacement illusion are stronger when the participant has met and likes the person whose face is used in the illusion, and experiencing the enfacement illusion changes implicit social biases associated with an in-group (Maister et al., 2013; 2015; Bulfalari et al., 2014). Visual effects of body illusions are specifically relevant to social variables such as in-group/out-group identity. Examination of the long term visual effect of the RHI may be useful for investigating how in-group/out-group identity remains stable over time. In the BSM group the longevity of the visual effect was reduced following the training interval. This reduction cannot be attributed to individual differences in susceptibility to the visual effect as both groups showed similar responses to the RHI in session one. The reduction is most likely related to changes to interoceptive awareness during BSM. Participants in the BSM group showed a significant increase in heartbeat detection accuracy, indicating that the accuracy of interoceptive perception had improved. BSM has been shown to improve awareness of interoceptive signals such as respiration (Daubenmier, Sze, Kerr, Kemeny and Mehling, 2014), but a number of studies investigating changes to heartbeat detection accuracy have not observed an improvement (Melloni et al., 2013; Parkin et al., 2014). These studies examined heartbeat detection performance over durations <50 seconds whereas in the present study performance was examined in additional 100 second trials, similar to previous research examining individual differences in embodiment (Tajadura-Jiménez and Tsakiris, 2014; Tsakiris, Tajadura- Jiménez and Costantini, 2011). As such, the improvements in heartbeat detection accuracy reported here reflect the ability to maintain attention on interoceptive signals over longer durations. This could account for these seemingly conflicting results, and would be consistent with previous research showing that mindfulness increases cortical responses during interoceptive attention (Mehling et al., 2013).

93 93 In the BSM group, improvements in heartbeat detection accuracy were negatively correlated with the magnitude of the long-term visual effect but this relationship was absent in the control group. The presence of this relationship in the BSM group suggests that BSM can alter top-down expectations regarding the reliability of interoceptive and visual body signals. Increasing attention to interoceptive signals may have increased the reliability of interoceptive prediction errors, altering the relative contribution of exteroceptive signals to body perception. This finding partially replicates previous studies showing a negative relationship between heartbeat detection accuracy and susceptibility to the immediate perceptual effects of body illusions (Tsakiris et al., 2011; Tajadura-Jiménez and Tsakiris, 2014). However, in this study this relationship was only evident in the novel long-term visual effect and not the immediate visual effect of the RHI. Reduced susceptibility to the immediate visual effect of the RHI was observed in both the BSM group and the control group and as such, cannot be directly attributed to the effect of mindfulness training. One possibility is that susceptibility to the visual effect generally reduces due to repeated experiences of the RHI, that is, independently of changes to interoceptive awareness. Alternatively, reduced susceptibility in each group may have two separate causes. Consistent with previous research, the BSM group s increased interoceptive awareness may lead them to be less susceptible to the visual effect of the RHI. Whereas reduced susceptibility in the control group may occur because the durable visual effect from the RHI in the first session elevates the pre-illusion score in the second session. By starting with a higher baseline there is less potential to measure further changes to visual judgements before the ceiling level of the task is reached, i.e. participants judge images that are 100% rubber hand as being like their own hand. A further study could assess whether a reduction in susceptibility occurs in the control group when a rubber hand of a different appearance is used in the second session of the RHI. That BSM predominantly influenced the long-term visual effect of the RHI suggests that BSM training altered the way that participants attended to visual body signals during the training interval. BSM training directs attention to internal sensations but also encourages participants to engage in open monitoring of these sensations. This practice allows participants to disengage from habitual interpretations of sensations, instead learning how these signals tend to occur so that they can be adaptively

94 94 modified (Pagnoni and Porro, 2014). This can increase the granularity of perceptual experience such that new, more specific interpretations of sensory input can be developed (Farb, 2015). Participants in the BSM group may have been better able to interpret the visual effect of the RHI in session one against the context of their ongoing body perception during the training interval. By attending to their body signals, they may have learned that their RHI experience was just a temporary and illusory change to body appearance. Whereas participants in the control group who did not deliberately engage in open monitoring of body signals, interpreted their RHI experience according to existing expectations, that embodiment of the rubber hand requires durable perceptual inferences about the appearance of their hand. Being able to develop new interpretations of sensory signals that are grounded in bottom-up sensory integration is beneficial because perceptual experience can become more consistent with actual sensory input or more veridical. Previous research has shown that BSM training can improve the veridicality of perceptual experience by reducing illusory perceptions (Grossman, 2004; Mirams et al., 2013). In the somatic signal detection task, participants judge the occurrence of a threshold vibrotactile stimulus whilst a concurrent LED flash occurs intermittently,. The presence of the LED flash can lead participants to report an illusory tactile perception (Lloyd, Mason, Brown and Poliakoff, 2008). Brief body scan meditation reduces the number of illusory tactile perceptions during the task, suggesting that participants are better at identifying the origins of sensory signals leading to more veridical perception (Mirams et al., 2013). Although brief BSM training did not reduce susceptibility to the immediate perceptual and subjective effects of the RHI, it did improve the accuracy of body perception over time as participants could detect their heartbeat more accurately and their durable representation of own body appearance was more accurate. The development of more accurate body perception may help to account for some of the beneficial effects of mindfulness practices, particularly for the treatment of clinical disorders characterised by disrupted self-perception such as anorexia (Rodríguez, Cowdrey and Park, 2014). Individuals with anorexia perceive their body shape and size to be very different to its objective form. Previous studies have shown that they interpret body signals differently to controls, having reduced interoceptive sensitivity (Kaye, Fudge, and Paulus, 2009; Park, Dunn and Barnard et al., 2012;

95 95 Pollatos et al., 2008) and increased susceptibility to the effects of the RHI (Keizer, Smeets, Postma, Elburg and Dijkerman, 2014). The beneficial effects of mindfulness in treating anorexia may be because the way that interoceptive and visual body signals contribute to their sense of embodiment is optimised. Examining changes to heartbeat detection and the long-term visual effect of the RHI during mindfulness based cognitive therapy may reveal how disruptions to embodiment occur and how they can be altered. The presence of a correlation between the reduction in the long-term visual effect and increases in heartbeat detection accuracy provides further evidence to support the idea that awareness of internal sensations influences how external body signals become integrated into body representations (e.g. Seth, 2015). However, determining the extent that the present results are consistent with previous studies is somewhat problematic. Previous studies have examined the relationship between heartbeat detection accuracy and susceptibility to the immediate effects of the RHI (Tsakiris et al., 2011; Tajadura-Jimenez and Tsakiris, 2013). Whilst there is evidence to support this association in the present study, this cannot be unambiguously interpreted for the reasons discussed above. The primary finding of the present study is that increases in heartbeat detection accuracy are associated with a reduced long-term visual effect of the RHI. Whether reductions to the long-term visual effect should be considered indicative of increased or decreased malleability of body representations is unclear. The reduction could be could be considered indicative of reduced malleability, such that the visual effect of the RHI on self-perception was reduced. Under this interpretation, the direction of the present results would be consistent with previous findings. Alternatively, reductions to the long-term visual effect could be interpreted as indicative of increased malleability, such that participants incorporated additional body information from their own body potentially diluting or over-writing the visual effect of the RHI. Under this interpretation, the direction of the present results would conflict with previous findings. As it is unclear how malleability should be interpreted in this context, this paper has argued that the concept of malleability may not be sufficient to explain the relationship observed. However, an alternative possibility is that the results do not fit neatly into the current explanatory framework because the results are spurious. This cannot be directly assessed because the present

96 96 study is the first of its kind to examine changes to a visual measure of embodiment following an intervention. Further research is required to explore these possibilities, specifically, to confirm the presence of a long-term visual effect of the RHI and to investigate precisely how the long-term visual effect becomes reduced during an intervention. Here, BSM was used to alter the contribution of interoceptive and exteroceptive body signals to embodiment. Following BSM training, participants could maintain interoceptive attention to detect their heartbeat more accurately. Though increased interoceptive awareness did not alter susceptibility to the immediate perceptual effect of the RHI, it reduced a previously unreported long-term visual effect of the RHI. The presence of a negative correlation between improvements in interoceptive awareness and a perceptual effect of the RHI suggests that attending to interoceptive sensations can improve the precision of interoceptive precision errors, increasing their contribution during the integration of body signals. Increasing the contribution of interoceptive signals may have improved the way that visual body information was interpreted. That this relationship is specific to a long-term effect indicates that BSM improved the granularity of body perception such that new, more veridical interpretations of body signals were learned. 3.2 Study four: the effect of anatomical dissection training on embodiment Introduction Empathy refers to the ability to share the feelings of others (Singer and Lamm, 2009). It consists of a number of dissociable components which describe different empathic capacities: (1) affective sharing to become affectively aroused by the valence and intensity of others emotions (2) empathic understanding to become consciously aware of the emotional state of another, (3) empathic concern to be motivated to care for another s emotional state and well-being, and (4) cognitive empathy take another s perspective and imagining what they are thinking and feeling (Decety and Jackson, 2004; Goubert, Craig and Buysse, et al., 2009; Singer and Lamm, 2009; Derntl et al., 2010; Decety, 2011; Decety and Cowell, 2014). It is widely agreed that empathy is an integral skill for doctors and other medical

97 97 practitioners. Doctors use social cues for two main reasons. Firstly, they use expression, intonation and body language firstly, to infer and transiently experience the patient s emotions, helping them to understand the patient s emotions (Hirsch, 2007). Secondly, to communicate their concern for the patient s situation to build rapport, encouraging patients to give more complete histories (Halpern, 2012). Skilful empathic communication results in a number of clinical benefits such as improved patient satisfaction (Blatt, LeLaucheur, Galinksky, Simmens and Greenberg et al., 2010), increased adherence to treatment and more favourable health outcomes (Hojat et al., 2011; Rieiss, Kelley, Bailey, Dunn and Phillips et al., 2012; Rakel et al., 2009; Kelley, Kraft-Todd, Schapira, Kossowsky and Riess et al., 2014). High empathy also confers benefits to doctors leading to increased health, wellbeing, effectiveness and professional satisfaction (Gleichgerrcht and Decety, 2012; Huntington and Kuhn, 2003; Di Blasi, Harkness, Ernst, Georgiou and Kleinjnen et al., 2001). However, in some ways empathy is incompatible with certain skills necessary for being a good doctor (Spiro, 1993). In the clinical setting, doctors must be able to separate their emotions and overcome automatic reactions, such as flinching or recoiling at another s pain, which are counterproductive to medical procedures, (Smajdor, Stöckl and Salter, 2010; Decety and Lamm, 2009). As such, doctors require a specialised form of empathy in which they can understand others internal states whilst maintaining a degree of detachment, this is often referred to as clinical empathy or detached concern (Halpern, 2012; Neumann, Bensing, Mercer, Ernstmann, Ommen, and Pfaffa, 2009). Clinical empathy is, understandably, a difficult balance to achieve (Helmich, Bolhuis, Laan, Dornan and Koopmans, 2013). Later in their careers, many doctors report experiencing compassion fatigue, a form of severe emotional exhaustion, and emotional detachment from others (Gleichgerrcht and Decety, 2014; 2013). These experiences have been identified as the cost of caring for others in pain (Figley, 1982; 2012). A substantial literature shows that despite the importance of empathy in medical practice, medical training often leads to a reduction in empathy, particularly following the first year of training and first year of clinical practice (for a review see Neumann et al., 2011; Youssef et al., 2014; although for an alternative perspective see Pedersen, 2009; Colliver, Conlee, Verhulst and Dorsey, 2010; Costa, Magalhães

98 98 and Costa, 2013; Roff, 2015). Doctors who score highly on personal distress measures are particularly susceptible to reductions in empathy, as well as Alexithymia (ie. an inability to identify and describe emotions), burnout and secondary traumatic distress (Gleichgerrcht and Decety, 2013). Given these negative effects for medical practitioners and patients it is important to understand how medical training influences various aspects of empathy as well as the mediating influence of personal distress. In previous research, reductions in empathy during medical training have been associated with high workload, stress and lack of peer support (Neumann et al., 2011). Though these issues are no doubt influential, the skills learned during medical training may themselves require changes to the neural processes that allow feelings to be shared. At a neural level, empathy involves processing another s subjective state as if it was one s own, triggering empathic concern and empathy (Decety, Michalska and Akitsuki, et al., 2008)). A number of functional neuroimaging studies have shown striking similarities in the neural circuits involved in the processing of both the first-hand experience of pain, and by the sight of other individuals in pain. These studies have consistently shown that the perception of pain in others elicits the activation of the brain areas supporting the processing of the affective and motivational dimensions of one s own experiences of pain (Jackson, Rainville and Decety, 2006). Medical training appears to alter these neural processes. When watching pain stimuli such as a needle puncturing skin, experienced acupuncturists give lower ratings of subjective pain and unpleasantness and have a reduced neural empathic pain response in comparison to untrained controls (Cheng, Lin, Liu, Hsu, Lim, Hung and Decety, 2007). However, these neural changes are not associated with changes to empathy as there were no differences in trait empathy between the experimental and control groups of these studies (Cheng et al., 2007; Fan and Han, 2008; Han, Fan and Mao et al., 2008). This may indicate a specialisation of selfother perception in experienced medical practitioners allowing them to empathise with another s state without incorporating it into their self-perception to the same degree as those without medical training. EEG studies suggest that the specialisation of self-other perception in medical practitioners is supported by early sensory processing. The temporal dynamics of perception of pain in others consists of two responses: (1) an early emotional sharing

99 99 component (frontal N110); and (2) a late cognitive evaluation (centro-parietal P3), a subsequent ERP study of the same effect demonstrated that neural processing differences between acupuncturists and controls occurs during early sensory processing (Fan and Han, 2008; Han et al., 2008). The sensory processes of selfother perception are often measured using multisensory body illusions such as the rubber hand illusion (RHI; Botvinick and Cohen, 1998). As discussed in chapter one susceptibility to the effects of the RHI varies between individuals. Individuals who are more susceptible to the RHI are also high in empathy suggesting that flexibility of self-other perception is useful for understanding the states of others (Asai et al., 2011). This flexibility may be a desirable trait for doctors to possess. However, susceptibility to the RHI tends to function in opposition to self-awareness, such that individuals with low awareness of their own heartbeat experience stronger effects of the RHI (Tsakiris, 2011; Tajadura-Jiménez et al., 2012). Individuals with high heartbeat detection accuracy appear to be more aware of their internal states in general, as they provide higher emotional intensity ratings to emotional stimuli (Wiens, Mezzacappa and Katkin, 2000; Dunn et al., 2010; Pollatos, Kirsch and Schandry, et al., 2005). Ideally for clinical empathy, doctors should have high flexibility of self-other perception, allowing them to experience the feelings of others, as well as high awareness of their own internal experience, so that they can clearly distinguish their own internal states. That these aspects of self-other perception typically function in opposition may help to explain why clinical empathy is a complex skill to acquire. Investigating how these components of self-other perception change during medical training may help to specify how clinical empathy develops and further identify how self-other processes that are maladaptive to clinical practice manifest. One aspect of medical training that is pertinent for changes to self-other perception is learning about gross anatomy. Gross anatomy generally commences at the beginning of the first year of medical training (Dickinson, Lancaster, Winfield, Reece and Colthorpe, 1997). It involves dissecting human cadavers to reveal various components of the body so that their organisation and function can be studied. This aspect of the course is widely considered, by students, educators and medical practitioners, to be an essential part of medical training not only for the anatomical knowledge that can be acquired but also for the development of professionalism (Böckers, Jerg-Bretzke,

100 100 Lamp, Brinkmann, Traue and Böckers, 2010; Arráez-Aybar, Sánchez-Montesinos, Mirapeix, Mompeo-Corredera and Sanudo-Tejero, 2010). Professionalism is often described as the sense of detachment that allows a student to objectively perform medical procedures without distraction or emotional response. Initially, this aspect of the course is somewhat distressing, with 4-6% of students describing difficulties adjusting which are expressed in nightmares, sleeplessness, impaired learning abilities and even post-traumatic stress like symptoms (Snelling, Sahai and Ellis, et al., 2003, Finkelstein and Mathers, 1990, Druce and Johonson, 1994; Dinsmore, Daugherty and Zeitz, et al., 2001). Despite initial issues students quickly adapt to the situation (Böckers et al., 2010). Part of this process of adaptation requires individuals to perform behaviours that would harm a living body without experiencing personal distress. A number of strategies are employed to reduce personal distress by dehumanising or de-personalising the cadaver, for example covering personal characteristics such as faces and hands (Shalev and Nathan, 1985; Finkelstein and Mathers, 1990). But ultimately medical students must, consciously or unconsciously, inhibit automatic empathic responses (Gustavson, 1988; Druce and Johonson, 1994; Lempp, 2005; Dickinson et al., 1997) and develop internal strategies for regulating their personal distress. The present study aims to assess how anatomical dissection training influences selfawareness and self-other perception in medical students who are high and low in trait personal distress. To assess this, changes to RHI susceptibility and heartbeat detection accuracy following the initial three months of medical training are assessed. Personal distress could potentially be reduced in a number of ways, some of which may subsequently have a negative effect on empathic ability. For example, personal distress may be lowered by reducing the flexibility of self-other perception such that external body information is less likely to be incorporated into selfexperience. In this case, participants are expected to show smaller immediate effects of the RHI on subjective and sensory outcome measures following anatomical dissection training. Reducing the flexibility of self-other perception may be a beneficial strategy for reducing personal distress in the dissection context but, generalised to the context of clinical practice may reduce empathy during patient interaction. Alternatively, personal distress may be reduced by inhibiting internal awareness of internal physiological states, such that attentional resources are

101 101 externally rather than internally focussed. In this case, participants are expected to show less accurate heartbeat detection performance and greater flexibility of selfother perception following anatomical dissection training. In the RHI, greater flexibility may be observed as larger immediate effects of the illusion and reduced long-term visual effect. Inhibiting awareness of internal physiological states may be a beneficial strategy for reducing personal distress in the dissection context but subsequently impair perception of one s own emotional state and the ability to infer the states of others Method Participants An advertisement was placed on the University of Manchester Medical Student Communities Hub and Experimental Participation Scheme website, inviting undergraduate students studying medicine and life sciences to take part in a multimodal investigation of body perception. Participants received 30 for their participation and were informed that the study was not eligible for any course credit. 56 undergraduates studying either medicine (N= 26, male: female = 14: 12, righthanded: left-handed = 20: 6, mean age=19 years) or life sciences (N=30, male: female = 13: 17, right-handed: left-handed = 28: 2, mean age=19 years) were recruited in total. None of the participants had experienced the RHI prior to the study. 52 participants returned for session 2. Participants were required to have normal or corrected-to-normal vision, with no tactile or proprioceptive deficits. Written informed consent was obtained from each participant. The research was approved by the University of Manchester Ethics Committee Design Participants attended two experimental sessions conducted three to four months apart (during this time participants completed 13 weeks of undergraduate classes). In this time, medical students completed 10 anatomy practical classes. In half of the practical classes, students studied body prosections (i.e. pre-dissected body parts) and in the other half participants dissected a human cadaver. Session one was

102 102 conducted in weeks 1-3 of the first semester. In session one, participants initially completed the Interpersonal Reactivity index (IRI; an example of this questionnaire is presented in Appendix C) which measures trait empathy on four scales (Empathic Concern, Personal Distress, Fantasy and Perspective Taking) and a heartbeat detection task. They then completed the RHI three times, each time participants completed a baseline sensory task before the RHI, and this task was repeated as soon as the RHI was finished. Three different sensory tasks were completed in total to assess changes to vision, touch and proprioception. The presentation order of the tasks was counterbalanced and participants rested for 10 minutes before starting the next sensory task. Following the final sensory task, participants completed a subjective questionnaire (An example of this questionnaire is presented in Appendix B). See study two A in Chapter two for further details regarding the procedure and materials for RHI measures, and Study three of this chapter for further details regarding the IRI, subjective questionnaire and heartbeat detection task. At the end of the session, participants also completed an uncanny valley task (Poliakoff, Beach, Best, Howard and Gowen, 2013) in which they rated a series of photographic images of hands on a laptop (not via the HMD). The hands in these images ranged from unrealistic (e.g. mechanical hands) to realistic (e.g. prosthetic and real hands). Participants rated images for eeriness or human-likeness (in a counterbalanced order) using a 9-point scale from not at all to extremely. Eerie was defined as: mysterious, strange, or unexpected as to send a chill up the spine and human-likeness as having human form or attributes. The next image was presented 2 seconds after the participant made their rating. Data from this task is not presented in this thesis. Session two followed the same procedure as session one except a short interview was conducted at the end of the session. Participants completed the same order of sensory tasks in each session. Figure 3.3 shows the study design and procedure.

103 103 Figure 3-3: Study design and procedure Training interval No training was administered as part of the study. During the training interval participants in each group completed their undergraduate studies as usual. In the first semester of the medicine BSc, students complete a module of anatomical dissection. Teaching involves both the use of prosections (i.e. pre-dissected models of individual body parts) and dissection of a cadaver. During the training interval, medical trainees completed five weeks of cadaver dissection covering the upper limbs, breast, vertebral column, thoracic wall, respiratory tract and lungs. Session

104 104 one was conducted at the beginning of the course prior to cadaver dissection. In the first semester of the life sciences BSc students complete a number of modules describing biological functions. These modules are taught in a lecture format with a number of practical classes in which students become familiar with equipment such as microscopes, but do not involve dissection of any living organism. Debrief At the end of the study participants completed a short debrief interview in which they were asked the following questions: Have you been absent from any of your classes? Did you participate in any anatomical dissection training prior to the course? Has the course altered how you think about the human body? Has the course altered how you think about your own body? What aspects of the course have been beneficial? Which aspects of the course have been challenging? How have you coped with challenging aspects of the course? Medical students were additionally asked to describe their experiences of anatomical dissection class and how their experience changed over time. Interview data was audio recorded and transcribed verbatim but is not presented in this thesis Results None of the participants reported absences from their lectures or practical classes. All data was normally distributed. Inspections for extreme values greater than 2.5 standard deviations from the mean, identified two outliers in data from the visual outcome measure of the RHI, who were excluded from the relevant analyses Empathy measures Two 2x2 (training type: medicine vs. life sciences, time point: before training vs. after training) ANOVAs were conducted to assess whether the levels of empathic concern and personal distress were different in medical students and life sciences students before and after the training period. For personal distress, there was a main effect of training type, F(1, 50)= 5.28, p=.026, but no main effect of time point, F(1,

105 105 50)= 0.75, p=.391, or interaction, F(1, 50)= 0.325, p=.571. For empathic concern, there was no main effect of training type, F(1, 50)= 2.772, p=.102, or time point, F(1, 50)= 0.08, p=.778, and no interaction, F(1, 50)= 0.149, p=.701. The results showed that medical students had significantly lower personal distress than life sciences students (8.16 vs ) but levels of personal distress did not change over time and were not influenced by training type. There were no differences in empathic concern between medical and life science students and empathic concern was not influenced by time or training. Table 3.7 shows the means and standard deviations for empathic concern and personal distress subscales of the IRI for medical and life science students in sessions one and two. Table 3-7: Comparison of means and standard deviations of Personal Distress and Empathic Concern subscales between student type (medical vs. life sciences) and session (one vs. two) Personal Distress Empathic Concern Session 1 Session 2 Session 1 Session 2 Medical 8.52 (3.38) 7.8 (4.4) 20.8 (3.44) (4.19) Life Sciences (5.64) (5.89) (5.33) (5.06) To investigate the effect of personal distress on embodiment for each student type, participants were split into high and low personal distress groups. The median personal distress score in session one was 8.5, participants scoring 9 were assigned to the high personal distress group (PD-high; N: medical students = 11, life sciences students = 17) and participants scoring <9 were assigned to the low personal distress group (PD-low; N: medical students = 15, life sciences students = 13).

106 Effects of high and low Personal Distress and student type on embodiment Subjective measure of embodiment A 2x2 repeated measures ANOVA (session: one vs. two, item type: illusion vs. control) was conducted to assess whether participants reported the typical subjective experience of the RHI. The results showed a main effect of item type, F(1, 51)= , p<.001, but no main effect of session, F(1, 51)= 0.631, p=.431, or interaction between item type and session, F(1, 51)= 1.031, p=.315. Overall illusion items were rated more positively than control items in both sessions (means = 1.57 vs ). A further 2x2x2 mixed ANOVA was conducted to assess whether type of training and personal distress influenced illusion scores in each session (training type: medicine vs. life sciences, personal distress: high vs. low, session: one vs two). The results showed no main effects of session, F(1, 48)= 0.032, p=.859, training type, F(1, 48)= 0.002, p=.859, or personal distress, F(1, 48)= 0.025, p=. 874, no two way interactions between personal distress and training type, F(1, 48)= 3.018, p=.089, training type and session, F(1, 48)= 0.18, p=.673, or session and personal distress, F(1, 48)= 0.096, p=.758, and no three way interaction between session, personal distress and training type, F(1, 48)= 0.014, p=.907. These results show that participants reported the typical subjective experience of the RHI but this did not change over time and was not influenced by course activities or level of personal distress. Means and standard deviations are displayed in table 3.8. Sensory outcome measures A 2x2x4 mixed AVOVA (personal distress: high vs. low, training type: medicine vs. life sciences, time point: 1, 2, 3, 4) was conducted to assess the effect of personal distress and training type on the visual measure of body perception across time points (before and after the RHI in session one and two). The results showed a significant main effect of time point, F(2.244, )= , p<.001, but no main effects of training type, F(1, 46)= 0.004, p=.949, or personal distress, F(1, 46)=0.226, p=.637, or two way interactions between time point and training type, F(2.244, )= 2.465, p=.084, training type and personal distress, F(1, 46)= 0.039, p=.844, or time and personal distress, F(2.244, )= 1.208, p=.306. But a significant three way interaction was observed between personal distress, training

107 107 type and time point, F(2.244, )= 6.207, p=.002. Figure 3.4 displays mean data showing the influence of personal distress and training type on the visual measure of body perception at each time point. Table 3-8: Comparison of the means and standard deviations of questionnaire scores (illusion vs. control) for each training type (medicine vs. life sciences) and level of personal distress (high vs. low) in each session (one vs. two) Session 1 Session 2 Illusion Control Illusion Control Medicine High 1.74 (1.08) (1.21) 1.85 (0.92) (1.02) Low 1.39 (1.26) -1.5 (1.06) 1.36 (1.15) (1.06) Total 1.57 (1.16) (1.12) 1.61 (1.04) (1.05) Life Sciences High 1.38 (0.83) (1.2) 1.31 (1.45) (1.26) Low 1.91 (1.04) (1.12) 1.79 (0.78) (0.49) Total 1.59 (0.94) (1.16) 1.51 (1.22) (1.01) Total 1.58 (1.04) -1.4 (1.13) 1.56 (1.13) (1.05) The presence of a three way interaction indicates that the proportion of rubber hand in images judged to be like own hand at each time point was influenced by level of personal distress and training type. To investigate the three way interaction further, additional analyses were conducted to assess the effect of personal distress and training type on susceptibility to the immediate visual effect of the RHI in session one and two, and the long-term visual effect of the RHI. To investigate the effect of personal distress and training type on the immediate visual effect of the RHI, an additional value was calculated by subtracting preillusion scores from post illusion scores (time 2 time 1) in session one and two. This measures the change in the proportion of rubber hand in images judged to be

108 108 like the participants own hand immediately after the illusion. A 2x2x2 mixed ANOVA was conducted to assess the effect of personal distress and training type on the immediate visual effect of the RHI. The results showed significant main effects of personal distress, F(1, 46)= 4.16, p=.047, and training type, F(1, 46)= 9.588, p=.003, but no main effect of time, F(1, 46)= 1.302, p=.26. There were no significant two way interactions between training type and time point, F(1, 46)= 0.029, p=.866, training type and personal distress, F(1, 46)= 0.029, p=.865, or personal distress and time point, F(1, 46)=0.048, p=.828. The three way interaction between personal distress, training type and time point was also non-significant, F(1, 46)= 0.834, p=.366. These results show that the immediate visual effect of the RHI was significantly larger for medical students than life sciences students (12.93 vs. 5.4). It was also significantly larger for the high personal distress group than the low personal distress group (11.7 vs. 6.63), however, the magnitude of the immediate visual effect did not change over time in any of the groups. Table 3.9 shows the means and standard deviations of the immediate visual effect of the RHI in sessions one and two for each training group and level of personal distress. To examine the effect of personal distress and training type on the long-term visual effect of the RHI, an additional value was calculated by subtracting scores at time two from scores at time three. This measures the reduction in the proportion of rubber hand in images judged to be like own hand following the training interval. A 2x2 between groups ANOVA was conducted to assess whether the long-term effect of the RHI was influenced by personal distress (high vs. low) and training type (medicine vs. life sciences). The results showed a significant main effect of training type, F(1, 46)=4.767, p=.034, but no main effect of personal distress, F(1, 46)= 2.448, p=.125, and a significant two way interaction between training type and personal distress, F(1, 46)=6.031, p=.018.

109 109 Figure 3-4 Bar chart comparing mean % of rubber hand in images judged to be like own hand at each time point for each training type (Medicine vs. Life Sciences) and level of Personal Distress (PD; High vs. Low). For all participants mean proportions increase after the RHI in session one (time points 1 and 2) and two (time points 3 and 4). Students studying Medicine who are high in PD show greater changes across each time point, showing greater increases between time points 1 and 2 and between time points 2 and 3, than the other groups. To investigate this interaction four independent t-tests were conducted with a Bonferroni corrected criterion of p= The difference between the long-term effect in medical students who were high and low in personal distress approached significance but did not reach the Bonferroni corrected criterion, t(23)=2.337, p=.029, means= (25.16) vs (16.1), or between life science students with high and low personal distress, t(23)=-0.913, p=.371, means= 0.91 (11.8) vs (11.28). For individuals with low personal distress, the long-term visual effect was

110 110 not influenced by type of training, t(19)=0.215, p=.832, means= (16.1) vs (11.28), however, for individuals with high personal distress, the long-term visual effect was significantly higher in life science students than medicine students, t(16.252)= , p=.008, means = 0.91 (11.8) vs (25.16). Table 3-9: Comparison of the means and standard deviations of immediate (sessions one and two) and long-term visual effects of the RHI for each training type (medicine vs. life sciences) and level of personal distress (high vs. low) Immediate Session 1 Session 2 Long-term Medicine High (14.88) (16.29) (25.16) Low 10.6 (13.95) (10.68) (16.1) Total (14.58) (13.62) (23.2) Life Sciences High 8.28 (7.42) 7.62 (8.38) 0.91 (11.57) Low 4.85 (9.06) 0.84 (7.63) (11.28) Total 7.05 (8.04) 5.18 (8.62) (11.57) Total (12.21) 8.09 (11.66) (19.11) These results suggest that medical training reduces the long-term visual effect of the RHI for individuals who are high in personal distress but not for individuals who are low in personal distress, whereas life sciences training does not influence the longterm visual effect. Table 3.9 shows the means and standard deviations of the longterm visual effect for each training group and level of personal distress. A 2x2x4 mixed AVOVA (personal distress: high vs. low, training type: medicine vs. life sciences, time point: 1 vs. 2 vs. 3 vs. 4) was conducted to assess the effect of personal distress and training type on the proprioceptive measure of body perception

111 111 at each time point (before and after the RHI in session one and two). The results showed a significant main effect of time F(2.117, )= , p=.001, but non-significant main effects of group, F(1, 48)= 1.033, p=.315, and personal distress, F(1, 48)=0.035, p=.852, two way interactions between training type and time point, F(2.117, )=1.708, p=.185, personal distress and time point, F(2.117, )= 1.197, p=.308, training type and personal distress, F(1, 48)= 2.559, p=.116, and a non-significant three way interaction between personal distress, training type and time point, F(2.117, )= 1.362, p=.261. Pairwise comparisons examining the main effect of time showed that proprioceptive judgements were significantly closer to the location of the rubber hand at time two than time one, p<.001, means = (58.95) vs (57.18), at time four than time three, p<.001, means = (55.6) vs (55.25), and at time two than time three, p<.001, means (58.95) vs (55.25). However, there were no significant differences between judgements at time one and time three, p=.162, means = (57.18) vs (55.25), at time one and time four, p=1.0, means = (57.18) vs (55.6), and time two and time four, p=.17, means = (58.95) vs (55.6). These results show that participants perceived their hand to be closer to the location of the rubber hand after the RHI in session one and two, but there was no long-term effect or any differences between high and low personal distress groups or between life sciences and medicine students. A 2x2x4 mixed AVOVA (personal distress: high vs. low, training type: medicine vs. life sciences, time point: 1 vs. 2 vs. 3 vs. 4) was conducted to assess the effect of personal distress and training type on the tactile threshold at each time point (before and after the RHI in session one and two). The results showed significant main effects of time, F(1.859, )= , p<.001, and training type, F(1,48)= 4.15, p=.047, but no main effects of personal distress, F(1, 48)= 1.715, p=.197, no two way interactions between time point and group, F(1.859, )= 0.825, p=.434, time point and personal distress, F(1.859, )= 0.554, p=.564, or training type and personal distress, F(1, 48)=0.039, p=.844, and no three way interaction between time point, personal distress or training type, F(1.859, )= 0.318, p=.713. Pairwise comparisons examining the main effect of time showed that the increase in tactile sensitivity between time one and two, means = (600.99) vs (558.07), and between time three and four, means = (523.72) vs

112 112 (526.99) was not significant (both p=1.00). However the decrease in tactile sensitivity between time two and three, means = (558.07) vs (523.72), between time one and three, means = (600.99) vs (523.72), time one and four, means = (600.99) vs (526.99) and between time two and four, means = (558.07) vs (526.99) were significant (all p<.001). Pairwise comparisons examining the main effect of training type showed that overall medical students had significantly greater tactile sensitivity than life sciences students, means = ( vs ). These results show that participants did not have an increase in tactile sensitivity following the RHI in either session one or two, however, all participants showed a decrease in tactile sensitivity between sessions one and two independent of training type and personal distress group. Heartbeat detection accuracy A 2x2x2 mixed AVOVA (personal distress: high vs. low, training type: medicine vs. life sciences, time point: 1 vs. 2) was conducted to assess the effect of personal distress and training type on heartbeat detection accuracy at each time point (session one and two). The results showed non-significant main effects of time F(1, 48)= 0.679, p=.414, training type, F(1, 48)= 0.83, p=.367, personal distress, F(1, 48)= 3.356, p=.073, a significant two way interaction between time point and personal distress, F(1, 48)= 4.419, p=.041, non-significant two way interactions between time point and group, F(1, 48)= 1.483, p=.229, and training type and personal distress, F(1, 48)= 0.443, p=.509, and a significant three way interaction between time point, personal distress and training type, F(1, 48)= 4.066, p=.049. To investigate this interaction further a HBD change score was calculated by subtracting HBD accuracy in session one from HBD accuracy in session two and four independent t-tests were conducted using this score and a Bonferroni corrected criterion of p= The results showed that for life science students there was no difference between HBD changes scores in the high and low personal distress groups, t(25)=0.065, p=.949, means=.04 (.1) vs..04 (.16), while medical students who were high in personal distress showed significantly lower HDB change scores than medical students who were low in personal distress, t(23)=2.278, p=.012,

113 113 means= -.09 (.11) vs..07 (.17). For individuals who were low in personal distress, type of training did not influence changes to HBD accuracy, t(21)= 0.437, p=.667, means =.07 (.17) vs..04 (.16), but for individuals who were high in personal distress, medical training led to significantly lower HBD accuracy following the training interval than life sciences training, t(27)= , p=.004, means= -.09 (.11) vs..04 (.11). These results show that unlike life sciences students and low personal distress medical students, medical students who are high in personal distress show a reduction in HBD accuracy following the training interval. Figure 3.5 shows the means and standard errors for HBD change scores in each training group and at each level of personal distress. Figure 3-5: Bar chart displaying means and standard errors of HBD accuracy change score for Medicine and Life sciences students with high and low trait personal distress. Whereas most participants showed a slight increase in HBD accuracy in session two, medical students who were high in trait personal distress showed a decrease in HBD accuracy following the training interval.

114 114 To assess the relationship between changes in HBD accuracy and the long-term visual effect, two Pearson s r correlations were conducted. The results showed that there was a positive relationship between the variables for medical students, r(25)=.482, p=.015 but no significant relationship was observed for life science students, r(25)=.069, p=.743. To investigate the relationship observed in the medical student group further, two separate correlations were conducted for participants who were high in personal distress and low in personal distress. The results showed that there was no significant relationship between changes in HBD accuracy and the long-term visual effect in either high personal distress, r(11)=.467, p=.092, or low personal distress, r(14)=.156, p=.648), medical students. These results suggest that for medical students, reductions in the visual effect following the training interval are associated with reductions in HBD accuracy, but this relationship is not specific to the high personal distress group. In contrast, there is no relationship between changes in HBD accuracy and the long-term visual effect for life science students Discussion This study aimed to assess changes to self-awareness and self-other perception following 3 months of medical training in individuals who are high and low in trait personal distress. The results showed that in general medical students and high personal distress individuals were more susceptible to the immediate visual effects of the RHI such that after the RHI they judged images with a higher proportion of rubber hand as being like their own hand than life science students and students with low trait personal distress. Susceptibility to the immediate visual effect did not change after the training interval indicating that it was not affected by course activities completed by either training group. The majority of students in both groups showed a long-term visual effect of the RHI such that the proportion of rubber hand in images judged to be like own hand measured after the RHI in session one did not significantly decline following the training interval. Only medical students who were high in personal distress showed a significant decline in the long-term visual effect of the RHI. These participants also showed a significant decrease in heartbeat detection accuracy following the training interval. Although significant subjective

115 115 and proprioceptive effects of the RHI were measured in both sessions, these measures were not influenced by type of training or level of personal distress. Individuals who were high in personal distress were more susceptible to the immediate visual effect of the RHI suggesting that their self-other perception is more flexible, that is, they are more likely to incorporate visual body information from other bodies, into their self-experience, specifically in this case the rubber hand. Susceptibility to the RHI is positively related to measures of empathy such that individuals reporting higher subjective ownership of the rubber hand and showing larger proprioceptive changes following the illusion also have higher scores on the empathic concern subscale of the IRI, although no relationships with the personal distress subscale were observed (Asai et al., 2011). In this study, high and low personal distress did not influence subjective and proprioceptive effects of the RHI, however, the present results extend previous research findings by demonstrating that individuals with higher scores on the personal distress subscale of the IRI were more susceptible to the immediate visual effect of the RHI. This may indicate that certain body signals have a specific role for different subtypes of empathy, however, further studies would be needed to assess this. The present study compared groups who were high and low in trait personal distress. This group design may have obscured relationships between trait empathy and subjective and proprioceptive outcome measures of the RHI. Future studies using a correlational design could assess these relationships further. Medical students were also more susceptible to the immediate visual effect. Though the results show that this sample of medical students was not higher in empathic concern than life sciences students, their chosen profession is in its nature caring, indicating that they have made life choices guided by empathic concern. High flexibility of self-other perception may be supportive of their capacity for empathy allowing them to incorporate visual information from other bodies, such as posture, expression and pallor, into their representation of self so that they can be understood using personal experience. A number of studies have shown that the perception of pain in others relies, at least partly, on the activation of a mental representation of pain in the self (for a review see Jackson et al., 2006). Similarly for the perception of touch, behaviour and facial expressions, observation and experience activates similar network of brain regions (Keysers et al., 2004; Jackson and Decety, 2004; Carr,

116 116 Lacoboni, Dubeau, Mazziotta and Lenzi, 2003). This neural sharing of others experiences is an important aspect of the cognitive hierarchy supporting empathy, at least on a neural level, visual information displaying or indicating another s internal state is incorporated into self-experience. That these medical students had a higher tendency to incorporate the visual features of the rubber hand suggests that even before medical training commences, they display the flexibility of self-other perception that would allow them to vicariously experience the states of others. Although the flexibility of self-other perception may be beneficial for empathy, it can come at a cost if vicarious experiences are not adequately distinguishable from self-experience. In this case, individuals may experience vicarious responses in inappropriate contexts not only producing personal distress but also directing attention away from external task demands. This is a particular issue in the context of anatomical dissection where observing mutilation of the human body could produce distress even though the cadaver does not feel pain (Evens and Fitzgibbon, 1992). A number of studies of medical students reactions to anatomical dissection show that many students find anatomical dissection distressing at first, with up to a third of students reporting unpleasant physical reactions such as nausea, fainting and disgust (Gustavson, 1988; Penney, 1983, 1985; Horne, Tiller, Eizenberg, Tashevska and Biddle, 1990, Evans and Fitzgibbon, 1992). These visceral responses may indicate that students have internalised the cadaver body and experienced a vicarious empathy response. Some evidence for this may be taken from the reduction of the long-term visual effect observed in medical students with high personal distress. This reduction was not observed in life science students with high personal distress and so indicates that it is medical training that reduces the longevity of the visual effect. It may be that these individuals, given their higher flexibility of selfother perception, are more likely to incorporate the visual features of the cadaver during anatomical dissection. This may cause interference with the long-term effect of the RHI as two hand appearances are incorporated rather than just one. Such incorporation of visual features may be more likely in the dissection context than in social interaction, in general due to the attentional demands of the situation. The attentional demands of anatomical dissections share some similarities with the RHI context. During both situations participants are instructed to focus attention upon a body object and as such require endogenous attention, further (i.e. self-

117 117 focussed), in both situations unusual and surprising perceptual events involving the body object occur that are likely to capture attention exogenously (i.e. externally captured). In the RHI, the surprising event is that tactile sensations occur when the rubber hand is touched. In the dissection context many unusual perceptual events occur including the revealing of what is inside the body and the absence of any pain reaction in response to painful stimuli. These surprising sensory events may increase the likelihood that individuals with high personal distress will incorporate visual information about others bodies into their self-perception. This would account for the reduced long-term visual effect of the RHI. Despite many students experiencing negative responses to anatomical dissection initially, this is rapidly reduced within the first 6 weeks of training (Evens and Fitzgibbon, 1992). Medical students report deliberately focussing on the task of dissection to distract themselves from the context (McGarvey, Farrell, Conroy, Kandiah and Monkhouse, et al., 2001) and supressing emotions to cope with the dissection context (Madill and Latchford, 2005). In the present study, medical students who were high in personal distress showed a significant reduction in heartbeat detection accuracy. It is unlikely that this change in heartbeat detection accuracy is due to stress. Medical students do report high levels of stress due to the demanding nature of their studies (Dyrbye, Thomas and Shanafelt, 2005), and stress may be expected to change awareness of one s physiological state, however, a number of studies have demonstrated that stress actually increases heartbeat detection accuracy when assessed through Schandry s internal tracking method (Schulz, Lass-Hennemann, Sütterlin, Schächinger and Vögele, 2013; Durlik, Cardini and Tsakiris, et al., 2014; Herbert, Pollatos, Flor, Enck and Schandry, et al., 2010; and for a review see Schulz and Vögele, 2015). The reduction in heartbeat detection accuracy observed in this study may reflect a deliberate strategy to focus attention away from internal body sensations. This may be beneficial within the anatomical dissection context as shifting focus away from one s internal state would free cognitive resources to be applied to the external task demands of learning. However, beyond the dissecting room, reducing awareness of internal body sensations may have a negative impact on both physician well-being and quality of care. Reduced awareness of internal body sensations is associated with lower awareness of one s own emotional state (Wiens, Mezzacappa and Katkin, 2000; Dunn et al., 2010;

118 118 Pollatos et al., 2005) and increased flexibility of self-other perception (Tsakiris, 2011; Tajadura-Jiménez et al., 2012). Combined, these changes to self and self-other perception may both increase incorporation of visual body information from others and reduce the ability to distinguish one s own emotion state from the states of others. This may lead leading doctors to internalise the states of their patients without the ability to distinguish whose state belongs to whom. Such processes may help to explain the seemingly paradoxical findings that experienced doctors who are high in personal distress are lower in empathic concern, yet more likely to report compassion fatigue and burnout (Gleichgerrcht and Decety, 2014). The changes to self-other perception observed in medical students high in trait personal distress may be prevented or reduced using mindfulness training such as body scan meditation. Study three showed that following body scan meditation, participants showed increased awareness of their heartbeat and improved ability to distinguish the appearance of their own hand from the appearance of the rubber hand in the RHI. A number of studies have shown that mindfulness training reduces stress, psychological distress, personal distress and burnout whilst increasing empathic concern and well-being in medical students (Dobkin and Hutchinson, 2013; Harwani, Motz, Graves, Amri, Harazduk and Haramati, 2014). The changes to selfother perception observed in this study may help to explain the beneficial effects of mindfulness interventions for medical students. During mindfulness, practitioners are instructed to focus upon their internal body sensations without judgement or interpretation (Kabat-Zinn et al., 1992). This process not only increases awareness of internal sensations, it also encourages practitioners to disengage from automatic topdown interpretations of current sensory information so that adaptive interpretations of body signals can be learned (Segal et al., 2012). Completing this practice during medical training may reduce personal distress by encouraging students to reinterpret their internal body sensations within a new sensory context rather than simply inhibiting their awareness of them. The effects of anatomical dissection appeared to be restricted to medical students who were high in personal distress. Medical students who were low in personal distress did not show any changes over time on measures of the RHI or heartbeat detection. Although this suggests that low personal distress may be a protective factor against the development of maladaptive coping strategies, it does not elucidate

119 119 how this protection is manifest in the perception of self and others. Previous studies demonstrate that self-other processes in experienced medical practitioners do change because they show a reduced neural response during the perception of vicarious pain when compared to controls (Cheng et al., 2007; Fan and Han, 2008; Han et al., 2008). The present results did not identify changes to sensory processes that could precipitate such a change to self-other perception, however, examination of the heartbeat detection changes scores suggests a potential sensory change worthy of further investigation. Heartbeat detection accuracy showed a small increase in the life sciences group which is most likely a practice effect due to repeating the task. Although low distress medical students also showed an increase in heartbeat detection accuracy this was slightly larger than the change observed in the life sciences group. Increased sensitivity to internal body sensations in medical students with low personal distress may help them to maintain high flexibility of self-other perception whilst still distinguishing their own internal state. Although this trend is visible in the present results, it may require substantial statistical power to detect this increase in heartbeat detection accuracy over and above the increase in the control group due to practice. Correlation analyses were conducted to assess whether changes to heartbeat detection accuracy and the long-term visual effect of the RHI were related. Previous studies have indicated that an individual s awareness of internal sensations influences the extent that external body information is incorporated into body representations immediately following the RHI (Tsakiris et al., 2011; Tajadura- Jimenez and Tsakiris, 2013). More relevant to the present study, study three showed that there was a negative relationship between heartbeat detection accuracy and reductions in the long-term visual effect of the RHI following body scan meditation training. In contrast, the results of the present study showed a positive correlation between changes to heartbeat detection accuracy and the long-term visual effect of the RHI in the medical student group only. Such that, medical students showing reduced heartbeat detection accuracy also showed a reduced long-term visual effect of the RHI. This provides further evidence to support the idea that awareness of internal sensations influences how external body signals become integrated into body representations (e.g. Seth, 2015), as well as partially supporting the findings of study three, that perception training activities alter heartbeat detection accuracy and visual

120 120 body representations in a co-ordinated way. However, the direction of this relationship is in the opposite direction to the relationship observed in study three. Thus far it has been argued that the pattern of results observed in each study reflect the specific nature of the training employed (this is discussed further in chapter five). Alternatively, it could be argued that the findings from at least one of these studies may be spurious. Spurious findings would be more likely in the small samples recruited for these studies and particularly in the present study, which examined subgroups within the sample. Additionally, the use of small samples in the subgroup analyses may also lack the statistical power required to detect true effects. This issue is particularly relevant to the subgroup correlations examining the relationship between changes to heartbeat detection accuracy and the long-term visual effect in medical students who are high and low in personal distress. Though the variables were correlated in the medical group as a whole, the correlations in the high and low personal distress groups were not significant. The lack of a significant correlation could potentially be explained by the use of subgroups. The split sample approach employed would remove useful variance from the analysis, as well as reduce statistical power by reducing the already limited sample size. In the absence of previous research investigating changes to embodiment following body perception training, it is difficult to identify when effects may be spurious or undetected through comparison to existing literature. These issues are not unexpected given the exploratory nature of the present research; however, further research using larger samples is required to assess the replicability of these effects before firm conclusions can be drawn. The present results identify some changes to self and self-other perception during medical training that may precipitate empathy decline in medical students. This link remains hypothetical, however, because no changes to personal distress or empathic concern were observed following the training interval. The training interval of three months is relatively short in comparison to previous studies that have shown a decrease in empathy (Neumann et al., 2009). These studies tend to show a decline in empathy after completion of the first year and again after the third year when clinical practice begins. In addition, the measure of empathy used in this study assesses trait

121 121 empathy rather than experiences of empathy at a specific time point. Changes to trait empathy may require longer than three months and so changes to empathy during medical training may be better assessed using an alternative questionnaire or a longer training interval. From the present results and comparison to previous literature, it can be hypothesized that the reduced awareness of internal sensations and flexibility of self-other perception within the anatomical dissection context would have a negative effect upon physician empathy and well-being. However, the stability of these changes is unknown and could potentially permit medical students with high personal distress to adapt to the dissection context in beneficial ways currently unspecified. Examining changes over the course of medical training may reveal a more complex pattern of sensory changes that relate to the development of clinical empathy. It would be particularly useful to examine how sensory changes relate to the second decline in empathy observed following the first year of clinical practice (Hojat, 2009), when changes to self-other perception may have a more obvious impact upon medical students capacity to empathise with patients and course assessment. This study aimed to assess changes to self-other perception following anatomical dissection training in medical students who were high and low in trait personal distress. The results showed that medical students, in general, were more susceptible to the immediate visual effects of the RHI suggesting a greater flexibility of selfother processes. This flexibility may be a factor influencing whether individuals choose caring professions such as medicine. Self-other perception did not appear to be affected by anatomical dissection training for medical students who were low in personal distress. But medical students who were high in trait personal distress showed a reduced long-term visual effect of the RHI and reduced HBD accuracy. These results suggest that medical students who are high in personal distress may react to aversive aspects of anatomical dissection by inhibiting awareness of their own physiological state, further increasing the extent that they incorporate visual information from others bodies. The combination of high self-other flexibility and low self-awareness may predispose doctors who are high in personal distress to compassion fatigue, empathy decline, burnout and Alexithymia. Further longitudinal studies are required to assess how these aspects of self-other perception are related to

122 122 changes in empathy throughout medical training and clinical practice, and whether they are influenced by mindfulness interventions. 3.3 Summary of studies three and four In studies three and four, an augmented reality approach developed in Chapter two was used to examine multiple sensory outcomes of the RHI before and after internal and external forms of body perception training. The results of these studies identified modality specific changes. Proprioceptive and subjective measures of the RHI did not change over time and were not influenced by either form of body perception training. In contrast, the visual outcome measure identified changes to embodiment due to both forms of body perception training. Specifically, in both studies the visual outcome of the RHI had a durable effect on embodiment such that participants in the control groups still judged the appearance of the rubber hand as being similar to the appearance of their own hand. However, participants who completed internal and external forms of body perception training showed reduced longevity of the visual effect. This suggests that participants had incorporated visual body information during the training interval, which as a form of interference, reduced the longevity of the visual effect of the RHI. Reductions in the longevity of the visual effect were associated with changes to interoceptive sensitivity but these changes were different for each form of body perception training. Participants completing internal forms of body perception training showed an increase in interoceptive sensitivity indicating that their body perception may have become more veridical. Whereas for participants who showed reduced longevity of the visual effect following external body perception training, showed a decrease in interoceptive sensitivity indicating that their body perception may have become less veridical as they incorporate body information from other bodies. The presence of modality specific changes to embodiment in these studies demonstrates the utility of measuring multiple sensory outcome of the RHI. By comparing across sensory outcome, this approach allows the unique contributions of individual modalities to be inferred.

123 123 4 Chapter 4: Development of the Embodiment Change Questionnaire Study one characterised the subjective experience of embodiment in the RHI when visual and tactile discrepancies were present in the illusion context. The results of study one suggested that when a tactile discrepancy was present participants reported changes to their body experience that could be characterised as body ownership or body extension. These changes to embodiment were closely related to each other and the perception of causality between the seen and felt touch elicited the RHI. In this chapter the qualitative observations from study one were used to inform an embodiment changes questionnaire that could be used to measure subjective embodiment in a number of contextual variations of the RHI. Study five evaluates the psychometric properties of the embodiment change questionnaire using confirmatory factor analysis and an analysis of convergent validity with conceptually related measures. 4.1 Study 5: A psychometric analysis of the Embodiment Change Questionnaire Though embodiment is considered to be a trait-like construct (Bekrater-Bodmann et al., 2012; Spence, 2015) a number of researchers have noted that its operationalization is often poorly defined (Longo et al., 2008; Spence, 2015). Changes to subjective embodiment are generally assessed using the questionnaire developed by Botvinick and Cohen (1998). Some of its items describe the typical experience of the RHI, while control items describe experiences not typically associated with the RHI. The measure includes a number of items describing various aspects of body experience in the RHI such as global self-perception (e.g. the rubber hand is my hand), modality specific sensory changes to self-perception and perception of the rubber hand (e.g. my hand is in the location of the rubber hand, the rubber hand resembled the appearance of my own hand), Perceived Causality between the seen and felt touch (e.g. the touch I felt was caused by the touch to the rubber hand) and a number of control statements that describe body experiences that are not typical of the RHI (e.g....i had two right hands). By aggregating response scores to a subset of these items the construct of embodiment can be measured as an illusion score which is then used in statistical analysis. The majority of studies

124 124 aggregate the following three items: the rubber hand was my hand, the touch I felt was caused by the touch to the rubber hand and my hand was in the location of the rubber hand and so the illusion score, from now on referred to as the standard illusion score, is a combination of global self-perception, Perceived Causality and modality specific sensory changes to self-perception (e.g. Preston, 2013; Kaplan et al., 2014; Botvinick and Cohen, 1998; Schütz-Bosbach, Tausche and Weiss, 2009). The selection of items used in the standard illusion score has been justified by the observation that these are items most commonly endorsed by participants after the illusion (Peled, Pressman, Geva and Modai, 2003; Ehrsson et al.,, Rosén, Stockselius, Ragnö and Köhler, 2008; Paton et al.,, Hohwy and Enticott, 2012; Kaplan et al., 2014). However, a conceptual rationale describing why these items are related to embodiment as a trait is lacking. In the absence of a conceptual rationale there is considerable variation between studies in the way embodiment is measured. While the majority of studies use the standard illusion score, many other studies use varied combinations of items. These combinations often include additional modality specific sensory changes to self-perception such as the skin on my hand was turning rubbery, control items such as I had more than one right hand (Asai et al., 2011) and items which are unique to a single study (e.g. the touching of the rubber hand felt just like an actual touch ; (Bekrater-Bodmann et al. 2012). This variation in measurement reduces clarity regarding what embodiment is, producing low interpretability and generalisability of research findings. For example, previous studies have shown that the subjective experience of the RHI is stable over time (Bekrater-Bodmann et al. 2012), and correlates positively with sensory measures of the RHI (Botvinick and Cohen, 1998) and the empathic concern measure of empathy (Asai et al., 2011). These studies each use different combinations of questionnaire items in their analyses so it is unclear whether the results are related to trait embodiment or the specific aspects of RHI experience described in aggregated items. Therefore, it is unclear how the selection of items in each study has influenced the relationships measured or whether a single measure of embodiment can replicate all of these previous findings. These issues are compounded when comparing across studies using different types of body illusion where features of the illusion context produce a different phenomenology to the RHI. For example, a change to self-

125 125 location is a reliable aspect of the RHI phenomenology which is not reported in the enfacement illusion (Tajadura-Jiménez et al., 2012) or when the RHI context includes a tactile discrepancy (as described in study two). Though body illusions are all used to study embodiment, they often use items that are unique to the sensory features of a particular illusion. Greater methodological consistency between studies of embodiment could be achieved if a conceptual rationale describing the essential features of embodiment, independent of the contextual influences of body illusions, was defined and psychometrically assessed. This issue was addressed in a previous study which used principle components analysis (PCA) to examine the underlying components of a larger 27 item questionnaire measure of the RHI (Longo, et al., 2008). This approach identified four components described as: embodiment of the rubber hand, loss of own hand, movement, and affect. A further analysis of the embodiment of the rubber hand component identified three subcomponents of: ownership, agency, and location. This study confirmed that the subjective experience of the RHI can be measured, however, the components identified are conceptually problematic in two respects. Firstly, the analysis separated conceptually related components of body ownership. Embodiment of the rubber hand is widely considered to replace embodiment of one s own hand and so the loss of one s own hand is a necessary aspect of feeling ownership over the rubber hand (De Preester and Tsakiris, 2009). The association between embodiment of the rubber hand and loss of one s own hand was also evident in the qualitative analysis conducted in study one. The analysis identified two changes to embodiment that can arise in the RHI. The first was body ownership, where participants described feelings of ownership and agency over the rubber hand. The second was body extension, where participants described an association with the rubber hand that was not like body ownership because they could still perceive their hand as a separate object to the rubber hand. The key distinction between these two changes to embodiment is the extent to which participants are aware of their own hand as distinct from the rubber hand. Secondly, location was identified as a subcomponent of embodiment of the rubber hand, but the IPA analysis indicated that this aspect is more closely related to the context of the RHI than the experience of embodiment. When there is no separation between the location of the rubber hand and the participant s hand, participants no longer describe changes to self-location.

126 126 Conceptual issues commonly arise from PCA because the analysis is an exploratory approach that maximises the independence of components. As the authors state, the items that are included in the analysis determine the number or type of components that are identified (Longo et al., 2008). The analysis cannot determine the need for additional components that may not be included in the item pool.. Further, the inclusion of items that correlate with the construct of interest but are not conceptually related to it, can alter the components that are identified. The PCA may have produced conceptually problematic components because items describing body extension were not included in the analysis. An alternative approach to investigating subjective embodiment is to specify a factor structure a priori and create a questionnaire measure that represents each factor with a minimum of three items describing distinct aspects of the factor (Brown, 2006). Following the collection of questionnaire data, the factor structure can then be compared to observed item correlations using confirmatory factor analysis (CFA; Brown, 2006). CFA offers a number of advantages over exploratory methods such as PCA. As the model is specified a priori, it is possible to statistically evaluate the validity of existing conceptual distinctions; further, CFA provides substantial information about fit between the hypothesised model and observed correlations so that model re-specifications can be identified. Finally, when a model fits the observed data well, factor scores can be generated and their validity and reliability assessed further. These characteristics of CFA make it a powerful tool for the refinement of questionnaire measures and underlying factor structures. The IPA results were summarised as a correlated three-factor model of Embodiment Change including factors of Body Ownership, Body Extension and Perceived Causality. Each factor was represented in an Embodiment Change Questionnaire (ECQ) using at least three items that were semantically related but had distinct meanings. The Body Ownership factor was represented with three items describing feelings of ownership and agency over the rubber hand (as identified in the PCA), as well as reduced awareness of one s own hand. The Body Extension factor was represented with four items describing feelings of association and connection with the rubber hand as an object joined to the body that has phenomenologically salient sensory attributes. The Perceived Causality factor described an expected association between the seen and felt touch such that they were causally related as the same

127 127 event. Items describing modality specific sensory changes to self-perception were not included in the questionnaire because the IPA showed that the tendency for participants to use such descriptions was influenced by the sensory discrepancies present in the RHI context and so they may not be an essential feature of embodiment. The present study aims to evaluate the psychometric properties of the ECQ and three-factor model of Embodiment Change. To achieve a thorough psychometric assessment, the ECQ and three-factor model of Embodiment Change was examined in three analyses. Analysis one provided a preliminary assessment of the acceptability of the three-factor model and identification of model respecifications. Following model re-specifications, analysis two assessed the fit of the three-factor model in separate sample. Finally, analysis three assessed the factor scores of the model and the standard illusion score in terms of their convergent validity with sensory measures of the RHI and trait empathy and test-retest reliability. 4.2 Analysis one Method Participants 100 participants (female: male =50: 50, mean age 25 years, age range years, right-handed: left-handed = 89:11) were recruited for the analysis. Data for 35 participants is taken from study two A. An additional 65 participants were recruited for this analysis; of these participants, 19 were recruited through the School of Psychological Sciences study participation scheme and 46 were recruited at a university open day. 29 participants had experienced the RHI prior to the study. All participants had normal or corrected to normal vison and were free from tactile and proprioceptive deficits in their right hand. Written informed consent was obtained from each participant. Research was approved by the University of Manchester Ethics Committee.

128 Factor structure A correlated three-factor model of Embodiment Change was specified with factors: Body Ownership, Body Extension, and Perceived Causality. Complete model specification including hypothesised item loadings and factor correlations are shown in figure 4.1. Marker indicators of each of the three factors were: the rubber hand was my hand, there was some form of association between myself and the rubber hand and the touch I saw caused the touch I felt, respectively. The measurement model contained no double loading indicators and all measurement error was presumed to be uncorrelated. The factors were permitted to be correlated based on qualitative analysis in study one showing that Body Ownership and Body Extension are closely related experiences that are both elicited due to the perception of causality between seen and felt touch Procedure and data analysis Participants completed the RHI questionnaire (an example of this questionnaire is presented in Appendix A) following a single RHI. Further details about the questionnaire and procedure are described in study two A (session one). A subset of these items forms the Embodiment Change Questionnaire (ECQ), which is presented in figure 4.1. These items were used in the present analysis. Data analysis was conducted using R (R Core Team, 2012) and the latent variable analysis package Lavaan (Rosseel, Oberski and Byrnes, 2011) which has been shown to generate the same results as other CFA software packages (Narayanan, 2012). Given that the data was ordinal and likely to be non-normal in distribution, maximum likelihood estimation was deemed inappropriate (DiStefano and Hess, 2005; Lubke and Muthén, 2004). Thus, a polychoric correlation matrix using the mean and variance adjusted weighted least square estimator (WLMSV; Flora and Curran, 2004) was calculated. This estimator has been shown to be robust to violations of

129 129 Figure 4-1: Path diagram of three-factor model of Embodiment Change tested using confirmatory factor analysis. Note. All items begin It seemed as though. normality (e.g., Dumenci and Achenbach, 2008; Li, 2015) and to provide reliable parameter estimates and accurate test statistics even in small samples (e.g. N=50; Moshagen and Musch, 2014). Goodness of fit was evaluated using the root mean square error of approximation (RMSEA) and its 90% confidence interval (90% CI; cf. MacCallum, Browne, and Sugawara, 1996), the standardized root mean square residual (SRMR), and the comparative fit index (CFI). Acceptable model fit was defined by the following criteria: RMSEA (<0.08, 90% CI <0.08), SRMR (<0.05), CFI (>0.90), and TLI (>0.90) as suggested by Hu and Bentler (1999). Multiple indices were used because they provide different information about model fit (i.e. absolute fit, fit adjusting for model parsimony, fit relative to a null model); used together, these indices provide a

130 130 more conservative and reliable evaluation of the solution (Brown, 2006). Standardized factor loadings >.32 were considered to be salient (Tabachnick and Fidell, 2001), values higher than 0.71 (accounting for 50% of variance or more) are considered excellent, values around 0.45 (20%) fair, and values below 0.32 (10% of variance) poor (Comrey and Lee, 1992) Results Input variables were assessed for normality, missing data, and multivariate outliers. All input variables were significantly non-normal (K-S all p<.001) validating the selection of the WLSMV estimator. Item 10 I expected to feel a touch when I saw the experimenters finger approaching showed extreme skew (98% of response scores between +2 and +3) and was excluded from the CFA. Following this exclusion the model was over-identified with df=24). Inspection of polychoric correlations shown in table 4.1 showed that with the exception of item 3 the experience of my hand was less vivid than normal, all input variables were significantly correlated with small to moderate strength relationships ( ). Item 3 was significantly correlated with five input variables, but the strength of relationships was small ( ). No correlations exceeded the cut-off for multicollinearity (all <.85; Keith, 2006), thus were deemed appropriate for inclusion in the CFA. The data set was complete so no strategy for missing values was implemented. Inspection of Mahalanobis Distance values revealed two outliers (9df, >27.877, p<.001). Responses from these participants were different to the rest of the sample because they were the only participants to score every item as -3 and so they were excluded from the CFA (N=98). The sample size is consistent with recommendations for CFA (i.e. 10 participants per item included (Joreskog and Sorbom, 1989). The three-factor model was on the borderline of acceptable fit with most indices around the proposed cut-offs, χ 2 (24) =29.091; p=.217; RMSEA =.047; 90% CI =.00,.099; SRMR =.048; CFI =.951. Table 4.2 displays the standardized loadings for each factor. Standardized estimates of factor loading showed that all items had significant positive relationships to their proposed factor. The majority of loadings were fair to excellent in size (.45 to.91) with the exception of item 3, my

131 131 experience of my own hand was less vivid than normal which was poor (.23) and did not meet the specified criterion for utility (>.32; Tabachnick and Fidell, 2001). Evaluation of localized areas of strain in this solution indicated that the sample correlation matrix was adequately reproduced by the model implied by the correlation matrix (all modification indices <4; Brown, 2006), and inspection of standardized expected parameter change did not indicate that the model could be improved by including constrained parameters. All factors were strongly correlated, with the correlation between ownership and extension factors (.89) and between ownership and causality (.86) exceeding the recommended cut-off for discriminative validity (>.85; Tabachnick and Fidell, 2001; Brown, 2006), suggesting poor discriminant validity of these dimensions and the possibility that a more parsimonious solution could be obtained Discussion Analysis one aimed to assess the acceptability of the three-factor model of Embodiment Change for the ECQ. The results showed that the three-factor model had borderline acceptable fit to the observed data, suggesting that a three-factor model of Embodiment Change is tenable, but could be improved. The results identified ways in which the model may be incorrectly specified. Firstly, the factor of Body Ownership showed poor discriminant validity as it was highly correlated with the other factors of Body extension and Perceived Causality. Though factors were expected to be correlated, the results indicate that better fit may be achieved with a one-factor model. Secondly, the results showed that item three was problematic in a number of respects. It had low correlations with other items in the ECQ, it had a negative mean score and its factor loading did not reach the criterion for utility. Such poor performance indicates that this item should be removed or replaced. The poor performance of item three is likely to have contributed to the high correlation between the factors of Body Ownership and Body Extension, as the distinguishing feature of these experiences is the extent to which participants perceive their hand as separate from the rubber hand. Item three was selected to measure this distinction because it was the most semantically similar item that had

132 132 been used in previous research. Item three may have performed poorly in the analysis because its meaning is somewhat ambiguous. Table 4-1: Polychoric correlation matrix and mean values of each item in the Embodiment Change Questionnaire Item number Mean SD When participants feel ownership over the rubber hand they have a less vivid experience of their own hand as a separate object, but a vivid experience of the rubber hand as their own hand. In this context the phrasing the experience of my own hand was less vivid than normal does not adequately describe the experience of Body Ownership. Rephrasing the item as I could no longer separate the experience of my own hand from my experience of the rubber hand would be a more valid representation of the phenomenology of Body Ownership as distinct from

133 133 Body Extension (described in study one). Inclusion of this item in a revised ECQ may improve discriminant validity of Body Ownership and Body Extension factors in the three-factor model of Embodiment Change. Table 4-2: Structure of factor loadings and correlations of the Embodiment Change Questionnaire: CFA of the three-factor model Item: it seemed as though Body Ownership Body Extension Perceived Causality 1 The rubber hand was my hand.74** 2 I could move the rubber hand.64** 3 The experience of my hand was less vivid than normal.23* 4 There was some form of association between myself and the rubber hand.75** 5 There was some form of connection between myself and the rubber hand.89** 6 The rubber hand was an object that had become joined to me.64** 7 I became more aware of the attributes of the rubber hand.45** 8 The touch I saw and the touch I felt were the same event.71** 9 The touch I felt was caused by the touch I saw on the rubber hand.91** Body Ownership Body Ownership Factor correlations Body Perceived Extension Causality Body Extension.89** Perceive Causality.86**.79** Note. All items start with "It seemed as though ". The CFA was conducted with robust weighted least squares estimation (WLSMV), p<.05*, p<.001**

134 134 To address issues of discriminant validity and the poor performance of item three, the ECQ was revised and two alternative models of Embodiment Change proposed. The two models tested were a three-factor model as described in analysis one and a one-factor model in which all items load onto a single factor of Embodiment Change. Model re-specifications based on a single data set can lead to over-fitting, where models are changed to represent the unique qualities of a sample rather than the construct of interest. Following model re-specification it is recommended to conduct an additional analysis in a new sample. Analysis two aims to assess the fit of the three-factor model of Embodiment Change in a new sample following revisions to the ECQ and to compare model fit to a one-factor model of Embodiment Change. 4.3 Analysis two Method Participants 103 participants recruited for session one in study three (N=47) and study four (N=56) participated in the analysis (female: male =70: 33, mean age = 19 years, age range years, right-handed: left-handed = 91:12). One participant had experienced the RHI prior to the study. All participants had normal or corrected to normal vison and were free from tactile and proprioceptive deficits in their right hand. Written informed consent was obtained from each participant. The research was approved by the University of Manchester Ethics Committee Factor structure The correlated three-factor model of Embodiment Change specified in chapter two was assessed again in this analysis. All factor loadings were identical, except the phrasing of item three was changed. Model specification is shown in figure 4.2 with the rephrased item in bold. The three-factor model was compared to a one-factor model in which all items loaded onto a factor of Embodiment Change. Comparison of non-nested models using AIC and BIC fit indices is not possible with the WLSMV estimator for ordinal data as these indices use the log likelihood value computed using ML estimation. To allow comparison between the one factor and

135 135 three factor models, the one factor model was nested within the three factor model by fixing factor covariance in the three factor model to 1. The one and three factor models were overidentified with df=27 and df= 24 respectively. Figure 4-2: Path diagram of the updated three-factor model of Embodiment Change tested using confirmatory factor analysis. Note. All items begin It seemed as though. Item 10 from first analysis excluded. Item 3 (highlighted in bold) replaced with less ambiguous item I could not separate the experience of my own hand from my experience of the rubber hand Procedure and data analysis Participants completed the RHI questionnaire (an example of this questionnaire is presented in Appendix B) following three repetitions of the RHI and sensory tasks. Further details about the questionnaire and procedure are described in study three of Chapter three. A subset of these items forms the Revised Embodiment Change Questionnaire (ECQ-R), which is displayed in figure 4.2. The questionnaire was