To some degree, the body is one object among others, but intuitively our awareness of it is not like the awareness of any other object. Yet, to specify exactly what is meant by this is relatively difficult. We have an internal access to it through a range of sensory channels specifically tuned to the body, giving rise to a vast range of bodily sensations. We feel cold, hungry, or tensed. We feel our arms stretched and our respiration accelerating. We feel the itchiness of the scarf and the resistance of the table. Only in recent history have these sensations become an object of sustained experimental investigation in cognitive science. Recent findings have revealed three key features of bodily awareness. First, the brain exploits multiple sources of information that are integrated together into unified internal models of the body, or bodily representations. Second, bodily representations are shaped by their functional use, whether for conscious perception (body image) or for action (body schema). Third, we typically experience our body as our own (sense of ownership), but we can also embody external objects, including tools, and, conversely, experience a part of our body as alien. The crucial question then is: What are the sensory, agentive, and affective requirements for the sense of ownership?

History

William James (1890, p. 242) said, “Our own bodily position, attitude, condition, is one of the things of which some awareness, however inattentive, invariably accompanies the knowledge of whatever else we know.” Similarly, the phenomenologist Aaron Gurwitsch (1985, p. 31) claimed that “there is no moment in our conscious life when we are completely unaware of our bodily posture, of the fact that we are walking, standing, sitting, lying down.” Despite its pervasive presence, for a long time, bodily self-awareness was exclusively discussed in the phenomenological tradition (Husserl, 1913/1989; Gurwitsch, 1985; Merleau-Ponty, 1945). Only since 2000 has there been a boom of interest in analytic philosophy and cognitive science in this topic.

The most comprehensive treatment in the phenomenological literature can be found in Merleau-Ponty’s Phenomenology of Perception (1945)he defended a sensorimotor approach to bodily awareness. He drew the distinction between the objective body made of muscles, bones, and nerves and the lived body, that is, the body that we experience without paying attention to it. Merleau-Ponty claimed that the lived body cannot be an object that we perceive or represent, because it is what makes our awareness of objects possible by anchoring our perspective on the world. Instead, Merleau-Ponty argued that we need to understand the lived body in terms of action. In his words, the lived body consists in an “I can” (p. 137). By positing action at the core of bodily awareness, Merleau-Ponty initiated a long tradition in phenomenology in addition to inspiring the more recent enactive approach (Henri, 1965; De Preester, 2007; Gallagher, 2005; Hurley, 1998; Thompson, 2005).

Merleau-Ponty was strongly influenced by the neurology of his time, and in the study of neurological disorders, one can find the beginning of a science of bodily consciousness. In particular, in a foundational paper of more than a hundred pages, Head and Holmes (1911) introduced the notion of internal body models and used it to explain a variety of disorders. They offered a systematic description of the patients they encountered and proposed the existence of three types of body models. The postural schema encodes the relative position of body parts after each movement. The superficial schema is used to localize bodily sensations on the surface of the skin. Finally, there is a body image, which is consciously available.

After Head and Holmes, most neurologists agreed on the existence of a bodily representation (often called body schema, body image, or both at the same time, e.g., Schilder, 1935), but with no distinction among the various types. It was only in the 1980s that cognitive scientists began to take apart body schema and body image, although there was little agreement on the way to individuate and measure them (de Vignemont, 2010).

In philosophy, O’Shaughnessy (1980) highlighted the criterion of malleability: some bodily representations are short-term and constantly updated; others are more stable and possibly partly innate. By contrast, Gallagher (2005) highlighted a functional criterion: The body schema consists in sensorimotor models of the body that guide actions, whereas the body image groups all other representations about the body that are not used for action. In neuroscience, Paillard (1999) and Dijkerman and de Haan (2007) confirmed the functional distinction between body schema and body image on the basis of empirical dissociations in neurological patients. The analysis of clinical populations led Schwoebel and Coslett (2005) to refine the taxonomy and additionally highlight a criterion of representational format; they distinguished between three types of bodily representations, respectively with sensorimotor format (body schema), visuospatial format (body structural description), and conceptual format (body semantics).

Since 2000, however, most of the debate has been focused on another fundamental issue, the sense of bodily ownership and the origin of the awareness of one’s body as one’s own. This switch is due to the discovery of a new bodily illusion known as the rubber hand illusion (RHI; Botvinick & Cohen, 1998), which opened the door to the scientific investigation of bodily self-awareness and is now a standard experimental paradigm. In the classic setup, participants sit with their arm hidden behind a screen while fixating on a rubber hand aligned with their body. If the participants’ real hand and the rubber hand are stroked synchronously, many participants report that the rubber hand feels like part of their body. Furthermore, there are behavioral and physiological markers that suggest that participants treat the illusory hand as their own.

The RHI was the first of many illusions working along the same principle of visual capture of touch, including the out-of-body illusions with virtual-reality goggles (Lenggenhager et al., 2007; for a different version, Ehrsson, 2007) [see Virtual Reality]. The focus on experimental manipulation of bodily self-consciousness in both healthy and clinical subjects has been the hallmark of empirical research.

Core concepts

Body senses

The body senses are a range of specialized inner sensory systems that provide the raw material for bodily awareness. Touch carries information both about the external world and about the surface of the skin. Proprioception provides information about the position and movement of the segments of the body on the basis of receptors on joints, muscles, and tendons. Nociception carries information about dangerously intense mechanical, mechanothermal, thermal, and chemical stimuli. The vestibular system in the inner ear provides information about body balance, acceleration, and the pull of gravity. Finally, interoception provides information about a wide variety of bodily needs and physiological demands.

Bodily consciousness mostly results from multisensory interactions, combining all available information, internal and external, in order to improve the likelihood of detecting, localizing, and identifying bodily events and properties. In particular, vision is often more accurate for spatial properties (van Beers et al., 1999), and it can also supplement information that is not directly available through the body senses.

However, multisensory integration faces a computational problem: how to select only the sensory signals that are relevant to bind together? If the brain fails to correctly parse the signals, it can lead to errors such as the RHI: One experiences touch on the rubber hand because one sees it located there, but visual information about the rubber hand should not have been integrated with tactile information. A first answer to the parsing problem is to claim that the various inputs must be spatially and temporally congruent. However, several experiments show that the rules are more complex. When wearing optical prisms inducing visual distortions, one sees one’s hand at a location displaced to one side. The visuo-proprioceptive conflict is usually resolved in favor of vision: One feels one’s hand as if it were located at its displaced optical position (Welch & Warren, 1980). By contrast, patients who still experience the presence of their limb missing after amputation can report feeling their phantom hand at a location that is occupied by an object. The perceptual system in that case does not combine the visual information about the object with the proprioceptive information about the phantom hand, despite their spatial congruency (de Vignemont, 2014).

A further difficulty arises from differences among the senses. Spatial and temporal encoding of sensory information varies across the senses. Tactile signals sent from the toe take more time to reach the brain than visual information about the same body part, for instance. Furthermore, it is often assumed that vision is encoded in an eye-centered spatial frame of reference, whereas touch is encoded in a skin-centered frame. To solve these issues, recent accounts of bodily awareness favor a Bayesian model of multisensory integration [see Bayesian Models of Cognition] (e.g., Noel, 2018). At the core of the model lies the idea that the brain computes the most probable output given all available sensory and cognitive cues. Each stream of information is given a different weight depending on the task and the context (for instance, vision is better for the localization of body parts in horizontal and vertical planes, but proprioception is better for the sagittal plane).

Bodily representations

Multisensory integration results in bodily representations. Bodily representations are internal models of the body. As representations, they are meant to covary with the state of the body, but they can also misrepresent and become decoupled. There is likely a hierarchical set of bodily representations, with each stage gaining in complexity (Longo, 2017). The first level is at the level of the primary somatosensory area, which receives inputs from touch, pain, and proprioception. The cortical map takes the shape of a “homunculus” organized in a somatotopical manner: Different neural regions respond to different parts of the body, although they do not follow anatomical contiguity (e.g., the hand-specific area is next to the face-specific area; Penfield & Rasmussen, 1950). Furthermore, some body parts are overrepresented, such as the hand, whereas others are underrepresented, such as the torso.

The brain then builds up multisensory bodily representations at the level of the parietal cortex, which gain in accuracy and richness by comparison with the homunculus in primary somatosensory area (Medina & Coslett, 2016; Longo, 2017). Following the dual model of visual processing, which distinguishes between visuomotor transformation in the dorsal pathway and conscious perception in the ventral pathway (Milner & Goodale, 1995), it is currently accepted that there are at least two functionally distinct bodily representations: the body schema for action and the body image for bodily perception (Ataria et al., 2021). More precisely, the body schema can be defined as the representation that carries precise information about bodily parameters required to act. It is encoded in a sensorimotor format that is directly exploitable by the motor system to guide action planning and control. The body image, on the other hand, is a purely descriptive perceptual representation. It carries information about more properties, although possibly with less precise and fine-grained content. This is true of perception in general: Perceptual content is richer and, thus, more informative than sensorimotor content.

The distinction between body schema/image has been confirmed by dissociations in both healthy and neurological populations. For instance, in the RHI, participants perceptually mislocate their hand, but their movements can be accurate when they reach for their hand or use it, thus remaining immune to the illusion (Kammers et al., 2009). Likewise, patients with numbsense (deficit of tactile awareness) are able to point to where they are touched, although they do not even feel that they have been touched (Paillard, 1999; see also Anema et al., 2009, for further dissociations). The distinction between the two body models can explain successful actions (thanks to a preserved body schema) in presence of distorted bodily experiences (due to a disturbed body image). However, pure deficits of either body schema or body image are relatively rare. Even for the RHI, some studies found that action can be sensitive to the illusion under certain conditions, though not in others (e.g., Kammers et al., 2010). More generally, in many instances, the body that one perceives coincides with the body that one acts. The brain generally tends to avoid inconsistency as much as possible, as shown in many cases of multisensory integration. Arguably, conflicts between body schema and body image are also to be avoided. If so, one can speculate that the two bodily representations can recalibrate each other and interact to maximize consistency, either when the discrepancy is beyond a certain threshold or when it lasts for a certain time. However, the rules that govern these interactions remain largely unknown.

Embodiment and ownership

Thanks to the notion of bodily representations, one can specify what is required for an object to become embodied (or incorporated). Embodiment is usually defined computationally: An object is embodied if it is encoded in a bodily representation and, thus, processed in the same manner as the parts of one’s body. For instance, there is clear evidence that tools can be embodied at the sensorimotor level as shown by the fact that tool use significantly modifies the kinematics of participants’ movements. In one study, after repeatedly using a long mechanical grabber, participants performed movements as if their arm were elongated, even for action they did not do with the tool (Cardinali et al., 2009). However, the embodiment of the external object is only partial: A tool is not processed as a limb at all levels. Typically, a grabber can be damaged without its user reacting as if a part of their own body is damaged, and one generally does not take it to be a part of one’s body even if at some level it is processed in the same way (for an exception, see the Toolish Illusion, Cardinali et al., 2021).

One must thus distinguish the computational notion of embodiment with the phenomenological notion of ownership. The sense of bodily ownership refers to the awareness of one’s body as one’s own. Within the sense of ownership, one can distinguish between the feeling of ownership (one experiences one’s hand as one’s own) and the judgment of ownership (one believes one’s hand to be one’s own). When appropriately grounded on body senses, ownership judgments are said to be immune to error through misidentification relative to the first-person (Evans, 1982): I cannot rationally doubt that these are my own fingers that are typing when I feel my fingers typing but not when I merely see them. The phenomenology of ownership, however, is relatively elusive, and it is mainly in neurological or psychiatric disorders that it becomes more salient. This is the case in somatoparaphrenia, which is caused by a lesion in the right parietal lobe and is often associated with motor and somatosensory deficits as well as spatial neglect. Patients with somatoparaphrenia deny that one of their limbs (often their left hand) belongs to them, and they ascribe it to another individual. Likewise, in the psychiatric syndrome of depersonalization, which can happen in depression, in schizophrenia, or after trauma, patients experience abnormal bodily properties and feel detached from their body, as if it did not belong to them or as if it had disappeared. Along with the RHI, the study of disownership syndromes can shed light on the grounds of ownership.

Questions, controversies, and new developments

Three main theories of the sense of ownership have been put forward. First, it has been argued that multisensory integration is the key to ownership (e.g., Blanke, 2012). The RHI can be induced only if the conditions for visuo-tactile integration are met. However, even if multisensory integration is important for experiencing ownership toward external objects, it is not sufficient: Tool use involves integrating vision, proprioception, and touch, but it does not come with a feeling of ownership. The second main approach to ownership relates to agency (e.g., Baier & Karnath, 2008; Ma & Hommel, 2015). These theories claim that the body that one experiences as one’s own is the body that is under direct control. Although most versions of the RHI are purely perceptual, it is also possible to induce it motorically by seeing the rubber hand moving while moving one’s own hand in a congruent manner (e.g., Sanchez-Vives et al., 2010). This view also faces difficulties explaining tool use. Nor can it account for the fact that many paralyzed individuals still experience their limbs as their own or that the RHI can be induced purely perceptually in the complete absence of agentive dimension (Longo et al., 2008). A third approach characterizes ownership in affective terms. The main measure of the RHI is to threaten the rubber hand and measure the physiological response. It has been found that the more participants react, the higher participants report feeling the rubber hand as their own (Ehrsson et al., 2007). On the other hand, patients with disownership syndromes display no affective response when they see the disowned limb under threat (Romano et al., 2014). On the basis of such findings, it has been argued that the sense of ownership has to represent the body necessary for survival. It constitutes a form of affective awareness of one’s body (de Vignemont, 2023).

Recent discussions have introduced a further dimension into bodily consciousness by considering the impact of beliefs and expectancies in the way one experiences one’s body. Bayesianism offers a fruitful model of bodily awareness but also raises questions about the impact of cognitive factors on perception. Are bodily experiences informationally encapsulated or are they also sensitive to the influence of cognitive states such as thoughts, desires, and expectations? Since Pylyshyn’s (1999) foundational paper on cognitive penetration of perception, the question of penetrability has been at the center of numerous debates. Yet those mostly focus on vision. In a highly debated paper (Lush et al., 2020), it was claimed that the RHI “requires” the contribution of top-down factors, although the evidence in favor of this claim was unclear (Ehrsson et al., 2022). Participants were first told what they should feel for each condition (synchronous and asynchronous), and only then did they receive the visuo-tactile stimulation. It was found that their expectations had no effect on the various dimensions of the illusion. It thus remains to be shown to what extent the experience of one’s body is influenced by one’s knowledge and expectations. There is a vast literature on pain modulation (e.g., Benedetti et al., 2020; Shevlin & Friesen, 2021), but little has been done yet on other dimensions of bodily awareness (e.g., Apelian et al., 2023).

Broader connections

One may ask to what extent the mechanisms described above apply to nonmammals and at what age they develop in human infants. Interestingly, it has been found that animals such as crayfish use interceptive strategy in chasing their prey, involving movements that are preplanned on the basis of the model of their own body dynamics and the expected target motion (Sillar et al., 2016). Such an ability can be taken as evidence that they have an internal model of their body equivalent to the body schema in primates. Further findings reveal that cleaner fish are able to recognize themselves in a mirror (Kohda et al., 2023), which can indicate another model, this time closer to the body image. In human infants, first evidence of mirror self-recognition can be found at 3 months old (Rochat, 1998), but newborns can already detect visuo-tactile synchrony when related to their own bodies (Filippetti et al., 2013).

Finally, questions about bodily ownership find practical application in work on protheses. On the basis of the RHI and tool use, it might seem that bodily representations are highly malleable. Yet patients who have had a limb amputated can experience phantom limbs for decades (Melzack, 1990). Their bodily representation does not adjust to the physical change despite constant visual reminders of the absence of the limb and the need to modify the way they act. Furthermore, when they wear prosthesis, they often have difficulty appropriating their artificial limb, describing it as like an object merely attached to their body. There is considerable research on the sensory dimension of artificial limbs (e.g., Bensmaia et al., 2020; Graczyk et al., 2019). So far, it mainly involves providing information about contact between the device and objects, either by providing tactile stimulation on a displaced skin surface or by directly activating the neural pathways originally supporting the sensory function. This research is motivated by the assumption that somatosensory feedback could improve motor control and the experience that users have of their prosthesis. For instance, two amputees described that their artificial limb felt more natural and their control more intuitive and less attention-demanding when their prosthesis was sensory-enabled through neural interface than when it was not. Interestingly, when the phase trial of sensory restoration stopped, one patient noted that his prosthesis “does not feel like me—went back to being an attachment” (Graczyk et al., 2019, p. 8). However, evidence on the actual role of somatosensory feedback remains anecdotal and has relatively low predictive power (Zbinden et al., 2022), suggesting the need for further study.

To conclude, whether one is interested in basic actions or in the more complex notion of the self, one cannot bypass the body. This may be true even for further abilities. For those who defend embodied cognition, the body plays a pervasive role in cognitive architecture, from emotions to social cognition and decision-making (Gallagher, 2005). However, in order to assess the importance of embodiment for these abilities, it is important to first understand how the brain encodes the body.

Bodily Sensations