As embodied subjects, we are aware of our bodies in distinctive ways.1 One source of this distinctiveness is that we have ways of finding out about our own bodies that we do not have about any other physical objects in the world. There are distinctive information channels that allow us directly to monitor both the body’s homeostatic states and its spatial properties. Some of these information channels are conscious and others unconscious. They all contribute, however, to a distinctive type of experience, namely the experience of oneself as an embodied agent. It is this distinctive type of experience that I call the phenomenology of bodily awareness. The phenomenology of bodily awareness has an important role to play in self-consciousness. It is, moreover, of critical importance in generating and controlling action.
Bodily awareness is a complex phenomenon that has received attention from a number of different theoretical and experimental approaches. It of course has intricate and highly developed physical underpinnings that are relatively well understood. Physiologists and neurophysiologists have devoted considerable attention to understanding the mechanisms of proprioception (awareness of limb position and bodily configuration) and somatosensation (bodily sensation).2 We have a good understanding of how bodily sensations originate in specialized receptors distributed across the surface of the skin and within the deep tissues. Some of these receptors are sensitive to skin and body temperature. Others are pain detectors (nociceptors). There are receptors specialized for mechanic stimuli of various kinds, such as pressure and vibration. Information about muscle stretch comes from muscle spindles. Other receptors monitor stresses and forces at the joints and in the tendons. Information from all of these receptors and nerve endings is carried by the spinal cord to the brain along three different pathways. One pathway carries information stemming from discriminative touch (which is a label for a complex set of tactile ways of finding out about the shape and texture of physical objects).3 Another carries information about pain and temperature. The third carries proprioceptive information. Each of these pathways ends up at a different brain area. The discriminative touch pathway travels to the cerebral cortex, while the proprioceptive pathway terminates in the cerebellum. The properties of these brain areas have been well studied. We know, for example, that tactile information is processed in the somatosensory cortex, which is located in the parietal lobe. The somatosensory cortex is somatotopically organized, with specific regions representing specific parts of the body. The cortical space assigned to information from each bodily region is a function of the fineness of tactile discrimination within that region (which is itself a function of the number of receptors there). Neuropsychologists, neuroimagers, and computational neuroscientists have made considerable progress in understanding how somatosensory and proprioceptive information is processed in the brain and how that processing can be disturbed by brain injury.
Explanations of the physiological underpinnings of bodily awareness can at best form only part of an understanding of the distinctiveness of the experience of embodiment—of what I earlier termed the phenomenology of bodily awareness. The gap between an understanding of the mechanisms underlying experience and the distinctive character of that experience has been much stressed in contemporary philosophy—many would think excessively so. Moving beyond physiology, bodily awareness has been approached from a number of perspectives. From the scientific point of view, much light has emerged from the study of patients with various forms of disorders of bodily awareness, such as deafferentation (where patients lose the ability to feel peripheral sensations) and autotopagnosia (where patients lose the ability to recognize and point to body parts). The verbal reports from patients suffering from neuropathies such as these can be very instructive in plotting the phenomenology of normal bodily awareness, precisely because of the insight they provide into what bodily awareness is like when certain central elements of normal bodily awareness are absent or distorted. So too can experimental exploration of the implications of abnormal bodily awareness for different types of motor behavior and deliberative action. As we will see further below, attention to the neuropathological data allows us to make considerable progress toward a taxonomy of the different components of bodily awareness.
From a philosophical point of view, there has been a resurgence of interest in the phenomenon of bodily awareness within the analytic tradition.4 In many ways this is an extension of analytical philosophers’ preoccupation with pain as a paradigmatic mental state. It is natural to compare and contrast the metaphysics and epistemology of pain with, say, the metaphysics and epistemology of bodily awareness. Many of the issues that arise mesh naturally with established concerns within the analytic tradition. So, for example, philosophers have explored whether the information about the body yielded by the various mechanisms of bodily awareness has the same type of privileged status that many theorists grant to the information about our own mental states that we derive from introspection. There are, moreover, long-standing debates about the role of bodily continuity in personal identity to which one might expect thinking about bodily awareness to be highly relevant. Many philosophers, beginning with Locke, have argued that psychological continuity (in the form of memories and other diachronic mental states) is what really matters for personal identity, so that bodily continuity is neither necessary nor sufficient for securing personal identity. The plausibility of this line of argument is likely to rest, at least partially, on the centrality that one accords to awareness of one’s own body in underwriting one’s sense of self. More generally, just as analytical philosophers have moved toward recognition that the mind is embedded within a social and physical context, and hence that we have to take social and environmental factors into account in thinking about the nature and content of mental states, so too are they coming to realize that we have to consider cognition and self-consciousness within an embodied context.
However, despite this resurgence of interest within the analytical tradition, the experiential dimension of bodily awareness has been most extensively explored within the phenomenological tradition.5 The most comprehensive treatment is to be found in part 1 of Merleau-Ponty’s Phenomenology of Perception (Merleau-Ponty 1962). One particularly interesting feature of Merleau-Ponty’s work in this area is how deeply informed it is by a detailed knowledge of current research in neuropsychology and neurophysiology at the time he was writing. Although the scientific study of bodily awareness has made huge advances since he was writing in the 1940s, the interface that he opened up between our understanding of experience, on the one hand, and our understanding of the mechanisms underlying that experience continues to be vitally important. And it remains the case that no subsequent author has explored this interface with anything like Merleau-Ponty’s depth and insight.
The problem that I will be addressing in this essay can be understood in terms of two of the different strands that we find in Merleau-Ponty’s rich exploration of the phenomenology of bodily awareness. From a phenomenological point of view, Merleau-Ponty explores in very insightful ways our distinctive ways of finding out about, and acting through, our bodies—the ways in which, as he puts it, “the body is the vehicle of being-in-the-world” (1962, 82). His project here is phenomenological in the nontechnical sense of the word. That is to say, he is concerned with characterizing agency and bodily awareness from the perspective of the experiencing subject. The distinction between first-person and third-person perspectives is useful at this point. Merleau-Ponty does not present matters quite in these terms, but it is one component of how he understands the distinction between the for-itself and the in-itself—a distinction that goes back at least as far as Hegel’s Phenomenology of Spirit, although Merleau-Ponty’s usage is no doubt more closely tied to Sartre’s use of the distinction in Being and Nothingness. From a first-person perspective, we experience the body qua for-itself, via the ways that it structures and gives meaning to our engagements with the physical world. From a third-person perspective, in contrast, we treat the body qua in-itself, as a complex of muscles, bones, and nerves that enters into causal interactions with other objects and that can in principle be studied and understood as one object among others, albeit a distinctive and highly complex object.6
The general contours of the distinction that Merleau-Ponty is making between the in-itself and the for-itself should be congenial to many theorists working within analytical philosophy of mind, although they would probably have some difficulty with how it is formulated. The basic idea that we cannot understand human agency in the same way that we understand causal interactions between nonanimate physical objects has been widely canvassed within the analytical tradition. One obvious point of contact is with theorists writing in the Wittgensteinian tradition, particularly those such as Anscombe, Taylor, and Kenny who argued that the reasons for which people act should not be understood in causal terms.7 But there are points of contact with theorists who accept that reasons can be causes. The distinction that some have tried to make between agent causation and event causation is somewhat in the spirit of Merleau-Ponty’s distinction.8 So too is the approach of Davidson’s anomalous monism, which is based on a sharp distinction between the law-governed domain of the physical and the norm-governed realm of the psychological.9
There are two aspects of Merleau-Ponty’s approach to agency and bodily experience, however, that analytical philosophers are likely to find unpalatable. The first is its susceptibility to an interpretation that draws strong metaphysical conclusions from the phenomenological distinction between the for-itself and the in-itself. Merleau-Ponty frequently writes as if the experienced body in some sense stands outside the physical world. He draws a distinction between the phenomenal body and the objective body that can be interpreted in a manner incompatible with any ontological position that, in the last analysis, treats the body as simply a highly developed biochemical object that stands apart from other objects in the world only in virtue of its complexity and organization.10 It is unclear, to this reader at least, where Merleau-Ponty draws the line between phenomenology and ontology in Phenomenology of Perception, but he often writes in a distinctly idealist vein, saying for example that “the constitution of our body as object” is a “crucial moment in the genesis of the objective world.” This dimension of his thinking might seem to place the experienced body outside the physical world in a way that is incompatible with even the weakest form of philosophical naturalism.
Even if we do not take Merleau-Ponty to be committed to such a drastic ontological position, and instead see him as primarily exploring a distinction between two ways of experiencing the body,11 he develops his views in a way that has significant repercussions for how we think about explanation—repercussions that philosophers of mind in the analytic tradition are unlikely to find congenial. As we will see in more detail in the next section, Merleau-Ponty is more than happy to draw the conclusion that the explanatory power of scientific investigation is severely constrained by the distinction between the phenomenal body and the objective body (which is a special case of his overall distinction between the phenomenal world and the objective world).12 The phenomenal body, Merleau-Ponty thinks, cannot be elucidated scientifically. Science can only inform us about the objective body.
The aim of this essay is to offer a way of doing justice to the phenomenological insights of Merleau-Ponty’s thinking about bodily awareness and its role in agency without following him in the limitations he places on the explanatory power of the scientific study of the body. I shall discuss one central feature of bodily awareness in a way that tries to respect the points that Merleau-Ponty stressed about the distinctive phenomenology of the experienced body. This feature is what Merleau-Ponty terms the “spatiality” of the body, which he discusses in the lengthy third chapter of part 1 of Phenomenology of Perception, “The Spatiality of One’s Own Body and Motility.” There are, I shall argue, some very fundamental differences between how we experience the spatiality of our own bodies and how we experience the spatiality of nonbodily physical objects. Bodily space, I shall argue, is represented in a fundamentally different way from the space within which we perceive and act on nonbodily physical objects. I will stress that we need to understand the spatiality of bodily awareness in terms of a non-Cartesian frame of reference, in contrast to the Cartesian frames of reference that structure our perception of, and interactions with, nonbodily physical objects. This basic distinction between two different types of frame of reference goes a long way, I shall suggest, toward accommodating what Merleau-Ponty correctly sees as the distinctiveness of the spatiality of our own bodies. And yet it is, of course, a distinction between two ways of representing space, rather than between two types of space—a distinction at the level of Fregean sense rather than Fregean reference. The fact that we experience our own bodies in terms of a non-Cartesian reference frame is perfectly compatible with our bodies being ontologically on a par with objects that we experience in terms of a Cartesian reference frame. Moreover, and this is the key methodological point, there is nothing about this distinction between two different frames of reference that stands in the way of our taking a third-person perspective on how bodily awareness feeds into and controls motor behavior and intentional action. I shall make good on this claim in the final section by illustrating how this approach to the experienced spatiality of somatic proprioception can be integrated with contemporary work on the psychology of motor control.
The general tenor of Merleau-Ponty’s thinking about our experience as embodied agents is given by his concise comment that “The outline of my body is a frontier which ordinary spatial relations to do not cross” (Merleau-Ponty 1962, 98). There is, he claims, a very fundamental discontinuity between the experienced spatiality of the physical world and the experienced spatiality of the body—more precisely, of the lived body, of the body as we might experience it from the inside. The body is not an object—or, more precisely, the lived body, the experienced body, cannot be understood as an object on a par with other objects in the external world.
In presenting Merleau-Ponty’s analysis of the phenomenology of bodily experience, I will focus on his discussion of the patient Schneider in the long chapter entitled “The Spatiality of One’s Own Body and Motility.” He is discussing a patient suffering from what he terms psychic blindness—the essence of the disorder being an inability to carry out what he (Merleau-Ponty) calls abstract movements, such as moving his arms and legs to order, naming and pointing to body parts, when his eyes are shut. He points out that there are certain movements that this patient is perfectly capable of making. Some of these are what we might call body-relative reactions. Here is an example.
A patient of the kind discussed above, when stung by a mosquito, does not need to look for the place where he has been stung. He finds it straight away, because for him there is no question of locating it in relation to axes of coordinates in objective space, but of reaching with his phenomenal hand a certain painful spot on his phenomenal body, and because between the hand as a scratching potentiality and the place stung as a spot to be scratched a directly experienced relationship is presented in the natural system of one’s own body. (Merleau-Ponty 1962, 105–106)
Here the distinction between the epistemological and the metaphysical strands in Merleau-Ponty’s thinking comes across very clearly. The epistemological point that he makes about the experience of localizing a sensation on the body seems exactly right. When one performs a simple body-relative action such as scratching a mosquito sting there is indeed no question of locating the sting on some sort of objective coordinate system, working out where one’s hand is on the same coordinate system, and then plotting a path between the two locations. The locations of both hand and sting are given in body-relative space (and I shall have more to say later about how this should be understood).
Merleau-Ponty uses these points about the phenomenology of bodily awareness to draw an explicit distinction between “objective space” and “the natural system of one’s own body.” The following passage is instructive:
The whole operation takes place in the domain of the phenomenal; it does not run through the objective world, and only the spectator, who lends his objective representation of the living body to the acting subject, can believe that the sting is perceived, that the hand moves in objective space, and consequently find it odd that the same subject can fail in experiments requiring him to point things out. (Merleau-Ponty 1962, 106)
These basic ideas then get generalized into a global distinction between the phenomenal body and the objective body. A few lines further on he writes: “It is never our objective body that we move, but our phenomenal body, and there is no mystery in that, since our body, as the potentiality of this or that part of the world, surges toward objects to be grasped and perceives them” (Merleau-Ponty 1962, 106). The distinction between the phenomenal body and the objective body plays an important role in Phenomenology of Perception. The phenomenal body is supposed to play a foundational role in the very constitution of the objective world. Here is a representative passage:
The body is not one more among external objects. It is neither tangible nor visible in so far as it is that which sees and touches. The body, therefore, is not one more among external objects, with the peculiarity of always being there. If it is permanent, the permanence is absolute and is the ground for the relative permanence of disappearing objects, real objects. The presence and absence of external objects are only variations within a field of primordial presence, a perceptual domain over which my body exercises power. Not only is the permanence of my body not a particular case of the permanence of external objects in the world, but the second cannot be understood except through the first: not only is the perspective of my body not a particular case of that of objects, but furthermore the presentation of objects in perspective cannot be understood except through the resistance of my body to all variations in perspective. (Merleau-Ponty 1962, 92)
I am not a Merleau-Ponty scholar, and I do not want to make any strong claims about what is going on here. What I would like to stress, however, is a conditional claim, namely, that if we accept Merleau-Ponty’s distinction between the phenomenal body and the objective body at face value, then it looks as if there will be very little scope for scientific study of the interesting and important aspects of bodily experience. Science, whether cognitive science, empirical psychology, or neurophysiology, can only inform us about the objective body. It can have nothing to say about the phenomenal body.
It would seem that Merleau-Ponty himself accepted this implication of the distinction between the phenomenal body and the objective body. That the distinction imposes limits on what we can learn from physiology and psychology is clearly stated in the first two chapters of part 1. In chapter l, “The Body as Object and Mechanistic Physiology,” Merleau-Ponty argues with some power that the physiological study of the body intrinsically involves an objectification of something that is fundamentally nonobjective. To study the physiology of the body is to treat the for-itself as an in-itself—to try to reduce the distinctive functioning of the body to mechanical causation of the type that governs interactions between nonbodily physical objects. As Merleau-Ponty brings out in discussing the phenomena of phantom limb and anosognosia (a patient’s refusal to accept the reality of his or her illness and deficits), the physiological treatment of neuropsychological disorders imposes upon us the burden of explaining how “the psychic determining factors and the physiological conditions gear into each other” (77)—an explanatory burden that he thinks it impossible to discharge. If we are to understand the phenomenology of bodily awareness, Merleau-Ponty concludes, there is little to be gained from studying the physiology of the body.
In chapter 2, “The Experience of the Body and Classical Psychology,” Merleau-Ponty takes a related but somewhat different tack. Although classical psychology, no less than classical physiology, is committed to treating the body as objective, he argues that it itself points us toward the inadequacy of the objectifying perspective. It is a little unclear what he means by “classical psychology,” but the points he wants to extract are clear enough. Within psychology we find descriptions of the body and of the role of the body in action that are, he thinks, simply incompatible with the idea that the body is just an object in the world among other objects. As far as this essay is concerned, one particularly interesting example Merleau-Ponty gives is the contrast drawn by “classical psychology” between ordinary perception of the movement of extrabodily physical objects and kinesthetic perception of bodily movement. Kinesthetic perception of bodily movement is global, he suggests, while our perception of the movement of ordinary objects is successive. Whereas we simply feel the body move, we perceive the movement of nonbodily objects by comparing their different positions at different times. The contrast is crude, according to Merleau-Ponty, but contains a germ of truth:
What they were expressing, badly it is true, by “kinaesthetic sensation” was the originality of the movements which I perform with my body: they directly anticipate the final situation, for my intention initiates a movement through space merely to obtain the objective initially given at the starting-point; there is as it were a germ of a movement which only secondarily develops into an objective movement. I move external objects with the aid of my body, which takes hold of them in one place and shifts them to another. But my body itself I move directly, I do not find it at one point of objective space and transfer it to another. I have no need to look for it, it is already with me. (1962, 94)
Nonetheless, Merleau-Ponty thinks, psychologists have failed to carry through their insights into the phenomenology of our experience of our own bodies. And what this means, of course, is that we can learn relatively little about the phenomenology of the body from empirical psychology—all we can learn, really, is the inadequacy of the objectifying approach of psychologists.
This pessimism about the possibility of learning from physiology and psychology might seem to be in tension with Merleau-Ponty’s interdisciplinary focus, and his constant appeal to neuropsychological case studies. However, the appearance is deceptive. What Merleau-Ponty is objecting to is the idea that we can understand the phenomenology of bodily awareness by studying the mechanisms that make the associated experiences possible—or, to put it in different terms, that we can study a first-person phenomenon through physiological and psychological mechanisms that are only susceptible to a third-person approach. But this is not in any sense incompatible with the thought that we can learn about the first-person phenomenon of bodily awareness by looking at the behavior of subjects in whom those mechanisms are not functioning properly—and indeed at how those subjects describe their experience of the world. It is mechanistic explanation of bodily awareness that Merleau-Ponty opposes, rather than scientific investigation per se.
Nonetheless, although Merleau-Ponty’s position is perfectly consistent, one can certainly wonder whether it is desirable. The price is high. It is difficult when reading Merleau-Ponty not to be convinced in very general terms that there must be some sort of distinction between two ways of thinking about the body—between those two approaches that he connects with the distinction between the for-itself and the in-itself. But should we follow him in concluding that there can be no dialogue between these two approaches; that there is nothing to be learned about the body qua for-itself by exploring the body qua in-itself? It is natural to wonder whether there might not be a way of doing justice to at least some of those features of bodily awareness that led Merleau-Ponty to make such a sharp distinction between the for-itself and the in-itself within a theoretical perspective that treats the body as ontologically on a par with nonbodily objects.
Let us look again at the crucial passage where Merleau-Ponty first begins to draw metaphysical conclusions from the phenomenology of bodily awareness.
The whole operation takes place in the domain of the phenomenal; it does not run through the objective world, and only the spectator, who lends his objective representation of the living body to the acting subject, can believe that the sting is perceived, that the hand moves in objective space, and consequently find it odd that the same subject can fail in experiments requiring him to point things out. (Merleau-Ponty 1962, 106)
The crucial claims here are both negative. The first is the denial that we should view Schneider’s awareness of his own body as awareness of an object (viz. the objective body), while the second is the denial that we should view Schneider’s reaching behavior as taking place in objective space. Only thus, Merleau-Ponty appears to be arguing, can we make sense of Schneider’s simultaneous ability to respond to stimuli on his own body and inability to point on command to locations on his own body.
This argument is not, as it stands, very persuasive. It seems plausible that a number of different information systems and neural circuits are involved in our awareness of our own bodies, and one would expect it to be occasionally the case that some of these systems and circuits are damaged while others are preserved. Indeed theorists concerned to distinguish various types of neural system frequently place considerable weight on the dissociations between different abilities and skills revealed by differential preservation in neuropathologies.13 From this perspective, the points that Merleau-Ponty notes are far better viewed as evidence for a distinction between two different ways of processing information about the body than as evidence for an ontological distinction between the objective body and the phenomenal body. To put the point in terms employed earlier, we can locate the distinction at the level of sense rather than the level of reference.
Of course, adopting this strategy only makes sense within the context of a general taxonomy of different types of bodily awareness—a taxonomy motivated by reflection on a wider range of cases and factors than those that are at issue here. In the remainder of this section, I will make some remarks in this direction before returning to Merleau-Ponty’s analysis of Schneider.
We can begin with a general distinction between high-level and low-level representations of the body. High-level representations of the body feed directly into central cognitive/affective processes, while low-level representations of the body feed directly into action. The distinction here is not between personal and subpersonal or between conscious and unconscious. Both low-level and high-level representations of the body function at the personal level and are usually conscious. The distinction will become clearer, however, with some examples.
Within the general category of high-level representations of the body, we can distinguish at least four types of representation or bodies of information:
Conceptual representations of the body (the set of beliefs we all have about the structure and nature of our body: how the body fits together, the functions of particular body parts, their approximate locations, and the sort of things that can go wrong with them).
Semantic knowledge of the names of body-parts (knowledge that interfaces with nonsemantic ways of identifying events in the body to allow us to give verbal reports of what is going on in our bodies).
Affective representations of the body (representations of the body associated with emotional responses to the body).
Homeostatic representations of the body (representations of the body relative to basic criteria of self-regulation and self-preservation).
Conceptual representations of the body are the least interesting, both philosophically and scientifically. There seems little reason to think that such conceptual representations will be any different in kind from the set of commonsense beliefs that we all have about the physical and social world. Homeostatic bodily representations present a number of interesting issues, but these are best considered in the context of the lower-level mechanisms that give rise to them. From the point of view of bodily awareness, it is more interesting to consider how the body is represented in the mechanisms that give rise to the experience of pain than it is to consider the judgment that one is in pain.
The remaining two types of higher-level representation are more interesting. There are identifiable pathologies specific to both semantic and affective representations of the body. The pathologies associated with affective representations of the body are familiar. Bulimia and anorexia are good examples—forms of emotional response based on distorted representations of the body. Patients with autotopagnosia have difficulty in naming body parts or pointing to body parts identified by name or by the application of some stimulus, either on their own bodies or on a schematic diagram of the body. The problems here are not purely semantic. Semantic representation of the body is not simply a matter of knowing the names of body parts. Although superficially similar deficits can be found in some aphasic patients (Semenza and Goodglass 1985), autotopagnosic patients do not have a localized word-category deficit. They lack a particular way of representing bodily locations, as we see from the fact that the problem carries across to pointing to body parts identified by the application of a stimulus.
Turning to lower-level representations of the body, here too we find a range of phenomena and associated information channels that need to be distinguished. The first is information about the structure and limits of the body. This type of body-relative information has a number of distinctive pathologies. The best-known example is the phenomenon of phantom limb found in many patients with amputated limbs, as well as some with amelia, the congenital absence of limbs (Melzack 1992). This first category of body-relative information performs two tasks. First, it is responsible for the felt location of sensations. Sensations are referred to specific body parts in virtue of a body of information about the structure of the body. Second, the same body of information informs the motor system about the body parts that are available to be employed in action.
This type of body-relative information should be distinguished from semantic representations of the body. In deafferented patients these types of information are dissociated in both directions. Deafferented patients have lost peripheral sensations in certain parts of the body. Jacques Paillard’s patient GL suffers from almost complete deafferentation from the mouth down, although she retains some sensitivity to thermal stimuli. If a thermal stimulus is delivered to a point on her arm that she is prevented from seeing, then, although she is unable to point to the location of the stimulus on her body, she is able to identify the location verbally and on a schematic body diagram. In my terms, she possesses semantic information without body-relative information. The dissociation also holds in the opposite direction. Another of Paillard’s patients had a parietal lesion that resulted in central deafferentation of the forearm. Although she could not verbally identify and report on a tactile stimulus delivered to her deafferented hand in a blindfolded condition, she was able to point to the location of the stimulus (Paillard et al. 1983).14 Here we have body-relative information without semantic information.
There is a second type of lower-level representation of the body. This is a moment-to-moment representation of the spatial position of the various parts of the body. This moment-to-moment representation of bodily position is essential for the initiation and control of action, and needs to be constantly updated by feedback from moving limbs. This representation has been called the short-term body-image by Brian O’Shaughnessy (O’Shaughnessy 1995), but the name is misleading, suggesting that there is a single way in which the disposition of body parts is represented, whereas in fact the spatial location of any given body part can be coded in three different and independent ways.
The first type of coding is relative to objects in the distal environment. Consider a simple action, like reaching one’s hand out for an object. The success of this action depends on an accurate computation of the trajectory from the initial position of the hand to the position of the relevant object.15 This requires the position of the hand and the position of the object to be computed relative to the same frame of reference. I shall call this object-relative spatial coding. It is most likely that object-relative spatial coding takes place on an egocentric frame of reference—that is to say, a frame of reference whose origin is some body part. The reason for calling this type of coding object-relative is that it deals primarily with the spatial relations between body parts and objects in the distal environment.
But many actions are directed toward the body rather than to objects independent of the body. Some of these actions are voluntary, as when I clasp my head in my hands in horror. Some are involuntary, as when I scratch an itch. Many more are somewhere between the two, as when I cross my legs or rub my eyes. Clearly, the possibility of any of these sorts of action rests on information about the location of the body parts in question relative to each other. We can call this sort of information body-internal spatial coding. It is information about the moment-by-moment position of body parts relative to each other.
Body-internal spatial coding is required, not just for body-directed action, but also for many types of action directed toward objects in the distal environment. Psychological studies of action often concentrate on very simple actions, such as grasping objects with one hand. But the vast majority of actions require the coordination of several body parts. When I play volleyball, for example, I need to know not just where each of my hands is relative to the ball as it comes over the net, but also where each of my hands is relative to the other hand. Both body-internal and object-relative spatial coding is required.
A third type of information about the moment-to-moment disposition of the body is just as important for the initiation and control of action as the first two. This is information about the orientation of the body as a whole in objective space, primarily involving information about the orientation of the body with respect to supporting surfaces and to the gravitational field. This information comes from the calibration of information from a number of sources. The three principal sources of orientational information are vision, the vestibular system in the inner ear, and the proprioceptive/kinesthetic system (at least two of which must be properly functioning for orientational information to be accurate). I shall call this orientational coding.
If the taxonomy I have offered is correct then there seem to be the following principal types of information about the body:
Let us return, then, to Schneider. What Merleau-Ponty finds so striking in Schneider is his inability to point on command to locations on his own body, when this is taken in the context of his residual abilities to respond to stimuli on his body by grasping and other body-directed behaviors. The following passage contains a very clear statement of the reasoning that leads him to the conclusion that we cannot understand the distinction between pointing and grasping in physiological terms.
If the grasping action or the concrete movement is guaranteed by some factual connection between each point on the skin and the motor muscles that guide the hand, it is difficult to see why the same nerve circuit communicating a scarcely different movement to the same muscles should not guarantee the gesture of Zeigen [pointing] as it does the movement of Greifen [grasping]. Between the mosquito which pricks the skin and the ruler which the doctor presses on the same spot, the physical difference is not great enough to explain why the grasping movement is possible, but the act of pointing impossible. The two ‘stimuli’ are really distinguishable only if we take into account their affective value or biological meaning, and the two responses cease to merge into one another only if we consider the Zeigen and the Greifen as two ways of relating to the object and two types of being in the world. But this is precisely what cannot be done once we have reduced the living body to the condition of an object. (1962, 123)
Without denying the insights that emerge from the existential analysis to which Merleau-Ponty subsequently turns, the argument here is far from persuasive. Merleau-Ponty may well be right that there is little physical difference between the mosquito bite and the touch of a ruler (although this is far from obvious), but this is the wrong place to look for an explanation of why one type of movement is possible but not the other. It seems far more plausible, particularly in the light of the taxonomy above, to seek an explanation in terms of the different representations of the body that the two types of movement respectively involve. So, for example, we might wonder whether Schneider’s difficulty in pointing is not, at least in part, best identified as a deficit in high-level bodily representations—as a problem in the mechanisms that underwrite explicit localization. It might also be the case that there are two fundamentally different forms of coding of moment-to-moment body parts in play in the two movements. Sensations such as mosquito bites are experienced within the boundaries of the body, in such a way that the movement of scratching the bite requires body-relative spatial coding, as opposed to the touch of a ruler which might be thought to require object-relative spatial coding. This would mean that pointing, unlike grasping, would require calibration of different forms of information about the location of body parts.16
The point here is not that we should interpret Schneider’s pathology in one or both of these ways. Rather, the claim is that careful distinctions between different types and levels of information about the body offer a greater number of potential resources for understanding what is going on in Schneider’s curious pattern of body-related motor behavior than Merleau-Ponty considers. In place of the simple distinction between the objective body and the phenomenal body, it makes sense to consider more complex distinctions between various ways of representing the objective body. Quite apart from avoiding metaphysical difficulties with Merleau-Ponty’s notion of the phenomenal body, this approach is likely to give a more nuanced way of tackling the phenomenology of bodily awareness.
It remains the case, however, that the finer-grained analysis of different types of information about the body proposed in the previous section does not yet do justice to Merleau-Ponty’s deeper motivation for the distinction between the phenomenal body and the objective body—namely his insistence that the spatiality of the body is fundamentally different from the spatiality of the objective world. I turn to this claim in this section, where I offer a way of thinking about how we represent bodily space that distinguishes it sharply from our representation of “body-external” space.
Almost all existing discussions of the spatiality of proprioception have presupposed that exteroceptive perception, proprioception, and the intentions controlling basic bodily actions must all have spatial contents coded on comparable frames of reference (where a frame of reference allows locations to be identified relative to axes centered on an object). This is an obvious assumption, since action requires integrating motor intentions and commands with perceptual information and proprioceptive information. Since the spatial locations of perceived objects and objects featuring in the contents of intentions are given relative to axes whose origin lies in the body—in an egocentric frame of reference—it is natural to suggest that the axes that determine particular proprioceptive frames of reference are centered on particular body parts, just as are the axes determining the frames of reference for perceptual content and basic intentions. The picture that emerges, therefore, is of a number of different representations of space, within each of which we find representations both of bodily and of nonbodily location. So, for example, we might imagine reaching behavior to be controlled by an egocentric frame of reference centered at some location on the hand—a frame of reference relative to which both bodily location (such as the mosquito bite on my arm) and nonbodily location (such as the cup on the table) can be identified.17
Despite its appealing economy, however, this account is ultimately unacceptable, because of a fundamental disanalogy between the bodily space of proprioception and the egocentric space of perception and action. In the case of vision or exteroceptive touch there is a perceptual field bounded in a way that determines a particular point as its origin. Since the visual field is essentially the solid angle of light picked up by the visual system the origin of the visual field is the apex of that solid angle. Similarly, the origin of the frame of reference for exploratory touch could be a point in the center of the palm of the relevant hand. But our awareness of our own bodies is not like this at all. It is not clear what possible reason there could be for offering one part of the body as the origin of the proprioceptive frame of reference.
There are certain spatial notions that are not applicable to somatic proprioception. For any two objects that are visually perceived, it makes obvious sense to ask both of the following questions:
The possibility of asking and answering these questions is closely bound up with the fact that visual perception has an origin-based frame of reference. Question (a) basically asks whether a line between the origin and one object would be longer or shorter than a corresponding line between the origin and the other object. Question (b) is just the question whether, if a line were drawn from the origin to the object that is furthest away, it would pass through the nearer object.
Neither question makes sense with respect to proprioception. One cannot ask whether this proprioceptively detected hand movement is farther away than this itch, nor whether this pain is in the same direction as that pain. What I am really asking when I ask which of two objects is farther away is which of the two objects is farther away from me, and a similar tacit self-reference is included when I ask whether two objects are in the same direction. But through somatic proprioception one learns about events taking place within the confines of the body, and there is no privileged part of the body that counts as me for the purpose of discussing the spatial relations they bear to each other.
To get a firmer grip on the distinctiveness of the frame of reference of bodily awareness one need only contrast the bodily experience of normal subjects with that of completely deafferented subjects, such as Jonathan Cole’s patient IW. The moment-to-moment information about their bodies that deafferented patients possess is almost exclusively derived from vision. Their awareness of their own body is continuous with their experience of the extrabodily world. They are aware of their bodies only from the same third-person person perspective that they have on nonbodily physical objects. The frame of reference for their bodily awareness does indeed have an origin—the eyes—and for this reason both of the two questions mentioned make perfect sense. But this is not at all the way in which we experience our bodies from a first-person perspective.
The conclusion to draw from this is that the spatial content of bodily awareness cannot be specified within a Cartesian frame of reference that takes the form of axes centered on an origin. But then how is it to be specified?
We can start from the basic thought that an account of the spatiality of bodily awareness must provide criteria for sameness of place. In the case of somatic proprioception this means criteria for sameness of bodily location. But there are several different types of criteria for sameness of bodily location. Consider the following two situations:
According to one set of criteria the pain is in the same bodily location in (i) and (ii)—that is to say, it is at a given point in my right ankle. According to another set of criteria, however, the pain is in different bodily locations in (i) and (ii), because my ankle has moved relative to other body parts. Let me term these A-location and B-location, respectively. Note, moreover, that B-location is independent of the actual location of the pain in “objective space.” The B-location of the pain in (ii) would be the same if I happened to be sitting in the same posture five feet to the left.
Both A-location and B-location need to be specified relative to a frame of reference. In thinking about this we need to bear in mind that the human body has both moveable and (relatively) immoveable body parts. On a large scale the human body can be viewed as an immoveable torso to which are appended moveable limbs—the head, arms, and legs. Within the moveable limbs there are small-scale body parts that can be directly moved in response to the will (such as the fingers, the toes, and the lower jaw) and others that cannot (such as the base of the skull). A joint is a body part that affords the possibility of moving a further body part, such as the neck, the elbow, or the ankle. In the human body, the relatively immoveable torso is linked by joints to five moveable limbs (the head, two legs, and two arms), each of which is further segmented by means of further joints. These joints provide the fixed points in terms of which the particular A-location and B-location of individual body parts at a time can be given.
A particular bodily A-location is given relative to the joints that bound the body part within which it is located. A particular point in the forearm is specified relative to the elbow and the wrist. It will be the point that lies on the surface of the skin at such-and-such a distance and direction from the wrist and such-and-such a distance and direction from the elbow. This mode of determining A-location secures the defining feature of A-location, which is that a given point within a given body part will have the same A-location irrespective of how the body as a whole moves, or of how the relevant body part moves relative to other body parts. The A-location of a given point within a given body part will remain constant in both those movements, because neither of those movements will bring about any changes in its distance and direction from the relevant joints.
The general model for identifying B-locations is as follows. A particular constant A-location is determined relative to the joints that bound the body part within which it falls. That A-location will either fall within the (relatively) immoveable torso or it will fall within a moveable limb. If it falls within the (relatively) immoveable torso then its B-location will also be fixed relative to the joints that bound the torso (neck, shoulders, and leg sockets)—that is to say, A-location and B-location will coincide. If, however, that A-location falls within a moveable limb, then its B-location will be fixed recursively relative to the joints that lie between it and the immoveable torso. The B-location will be specified in terms of the angles of the joints that lie between it and the immoveable torso. Some of these joint angles will be rotational (as with the elbow joint, for example). Others will be translational (as with the middle finger joint).
The way of specifying A-location and B-location seems to capture certain important elements in the phenomenology of bodily awareness.
To return, then, to Merleau-Ponty, my proposal is that a due recognition of the distinctive frame of reference relative to which proprioception and somatosensations are located can do justice to many of his insights about the phenomenology of bodily awareness. Moreover, and this is the important point when it comes to skepticism about Merleau-Ponty’s distinction between the objective body and the phenomenal body, the distinctiveness of bodily awareness is being accommodated at the level of sense rather than at the level of reference—that is, in terms of how we represent the body. There is no temptation to postulate a phenomenal body that stands apart from the objective body. The fact that we experience our own bodies in terms of a non-Cartesian reference frame does not in any sense rule out our bodies being ontologically on a par with objects that we experience in terms of Cartesian reference frame. Nor does it preclude our studying bodily experience from the third-person, scientific perspective. Indeed, as I shall try to bring out in the next section, properly understanding the reference frame governing bodily awareness offers a fruitful framework for the scientific study of bodily awareness.
The previous section offered a way of thinking about the spatiality of bodily awareness as fundamentally different from the spatial content of vision and other forms of exteroceptive awareness. One obvious question this raises is how proprioceptive content features in the control of action, given that action requires the contribution and integration of proprioceptive and exteroceptive awareness. What I will try to do in this final section is explain how the account I have offered of the spatial content of somatic proprioception fits in with some influential current thinking about motor control. This will go some way toward substantiating my earlier comments about the possibility of incorporating the distinctive phenomenology of bodily awareness within a “third-person” perspective on agency.
Any planned motor movement directed toward an extrabodily object requires two basic types of information:
The first question to ask is how this position information is coded. The study of trajectory errors and velocity profiles of hand movements suggests that the first type of information is coded in extrinsic coordinates in a frame of references centered on the hand (Ghez et al. 2000). Intended reaching movements are coded, roughly speaking, in terms of their goal and end point, rather than the means by which that end point is to be achieved. This coding involves hand-centered vectors, rather than the complex muscle forces and joint torques required for the action to be successfully carried out. One source of evidence for this is that hand movements directed at extrabodily targets have constant kinematic profiles, remaining straight and showing bell-shaped velocity curves with predictable acceleration at the beginning of the movement and deceleration as the target is approached (Morasso 1981). These kinematic profiles do not seem to be correlated with joint movements. There is considerable debate about whether the frame of reference on which target position is coded is egocentric or allocentric (Jeannerod 1997), but there is relatively little dispute that the coordinates are extrinsic rather than intrinsic (but see Uno et al. 1989 for the suggestion that the kinematic profiles observed by Morasso are consistent with the minimization of overall joint torque).
Turning to the second type of information, information about the starting position of limbs, the thrust of this essay has been that awareness of the body derived from somatic proprioception and somatosensation is not coded on either an object-centered or a body-centered frame of reference. The coordinates on which the location of body parts is coded are intrinsic rather than extrinsic. This leads us to an obvious second question: how is this type of bodily awareness involved in the control of action? The experienced spatiality of the body, as I have analyzed it, is closely bound up with awareness of the body’s possibilities for action. The body presents itself phenomenologically as segmented into body parts separated by joints because these are the natural units for movement. But what we need to know are the details of the contribution that somatic proprioception makes to the initiation and control of action.
If the spatial dimension of proprioception and somatosensation is as I have described it, somatic proprioception clearly cannot provide information about the position of the relevant limb that will be sufficient to fix the initial position of the movement vector. Somatic proprioception and somatosensation provide information about how limbs are distributed, but this information will not fix the starting position of the hand in a way that allows immediate computation of the movement vector required to reach the target. There is no immediate way of computing the trajectory between the location of a limb given in terms of A-location and B-location, and a target location given in terms of extrinsic coordinates. As Merleau-Ponty puts it in a passage cited earlier, “the outline of my body is a frontier which ordinary spatial relations do not cross” (1962, 98). To put things somewhat more prosaically, acting effectively on the world requires some sort of translation between two fundamentally different coordinate frames. The translation required for the calculation of the movement vector will involve integrating information derived from the various mechanisms of bodily awareness with visually derived information. This yields a testable prediction, namely, that subjects who are prevented from seeing their hands before making a reaching movement to a visible target should not be capable of making accurate movements. And this in fact is what experimenters have found (Ghez et al. 1995). The fact that information derived from bodily awareness is not sufficient to guide and control action is powerful evidence that there is no single spatial coordinate system that encompasses both bodily awareness and external perception.
Does the specific proposal that the coordinate system of bodily awareness should be understood in terms of A-location and B-location (as distinct from the general proposal that the spatiality of the body is fundamentally different from the spatiality of the perceived world) link up in any interesting ways with the empirical study of action?
Many researchers into motor control currently think that we need to distinguish the kinematics of motor control from the dynamics of motor control (Bizzi and Mussa-Ivaldi 2000). Movements are planned in purely kinematic terms, as a sequence of positions in peripersonal space that the hand will successively occupy during the performance of the movement. Clearly, however, the actual execution of the movement depends on these extrinsically specified feedforward motor commands being implemented by intrinsically specified muscle forces, joint angles, joint torque, and so forth. The transition from extrinsically specified coordinates to intrinsically specified coordinates comes when the nervous system computes the dynamical implementation of the kinematically specified goal. Various proposals have been made about how this computation is achieved.18 One traditional assumption is that this is a process of reverse engineering, so that the calculation of the muscle forces and joint angles required to implement the movement is achieved by working backward from the trajectory of the end point. There are obvious problems of computational tractability here. The problem does not have a unique solution, and in any case there are likely to be considerable difficulties in factoring in biomechanical factors due to fatigue and other variables. Accordingly it has been suggested that the translation into intrinsic coordinates does not depend on the solution of complex inverse-dynamic and inverse-computations but instead involves translating the targeted end point into a series of equilibrium positions (Feldman 1986). The basic thought here is that muscles and reflexes work as springs in ways that allow effector limbs to be treated as mass-spring systems that have adjustable equilibria. Motor planning, on this model, involves determining the equilibrium positions for the relevant effector limbs.
Whether the inverse-dynamic approach or the equilibrium approach is correct, it is precisely at this point that proprioceptively derived information about the distribution of body parts becomes crucial. The frame of reference of the intrinsic coordinates in which joint angles and equilibrium positions are coded seems much closer to the frame of reference of proprioceptive bodily awareness as I have characterized it than it is to the Cartesian frames of reference on which movement end points are coded. This provides a good explanation of why proprioceptive and somatosensory feedback is able to play such an integral part in the smooth performance and correction of actions, as indeed in the development of internal models of limb dynamics. The feedforward commands directed at the hand are recursively structured in much the same way as proprioceptive feedback from the hand and intervening body segments. Motor commands to the hand need to specify appropriate angles for the shoulder, the elbow, and the wrist. Proprioceptive feedback about the B-location of the hand will equally specify the relevant joint angles. Comparison is straightforward. The crucial role of bodily awareness in the initiation and control of action comes at the point of transition between kinematic plan and dynamic instruction, as well as later on in the execution of the movement. What makes it possible for somatic proprioception to perform this role is that the awareness of the body it provides is coded on a frame of reference that maps straightforwardly onto the internal model of limb dynamics that specifies the body’s potentialities for movement.
Let me draw the threads of the argument together. I began by considering two of the central themes in Merleau-Ponty’s discussion of the body in Phenomenology of Perception. The first theme stresses the distinctiveness of how we experience our own bodies, and in particular the phenomenological differences between our awareness of the spatiality of our own bodies and our awareness of the spatiality of the extrabodily physical world. This theme is predominantly phenomenological. The second theme has to do with the relation between the phenomenological investigation of bodily awareness and the scientific study of the body. This second theme emerges particularly in Merleau-Ponty’s development of the distinction between the objective body and the phenomenal body—between the body as a physical mass of bone, muscles, and nerves and the body as it is lived and experienced. As we saw, Merleau-Ponty develops this distinction in a way that places our first-person experience of our bodies and of our actions outside the domain of third-person physiology and scientific psychology.
The principal aim of this essay has been to try to accommodate the insights behind the first theme in Phenomenology of Perception in a way that keeps the body and bodily awareness “within the world,” and hence without following Merleau-Ponty in the conclusions he draws from the distinction between the phenomenal body and the objective body. It is true that our experience of our own bodies and of our own agency has a number of very distinctive features that sets it apart from our experience of nonbodily objects. But these differences can, I suggested, be illuminated by thinking about the physiological and psychological mechanisms and information sources that underlie them. As we saw in the context of Merleau-Ponty’s discussion of Schneider, careful distinctions between different types of body-relative information and different ways of representing the body show promise for dealing with the puzzles and problems that led Merleau-Ponty to the distinction between phenomenal body and objective body. More importantly, I proposed a way of thinking about the spatiality of bodily awareness that goes a considerable way to explaining the fundamental differences that Merleau-Ponty identified between our experience of our bodies and our experience of nonbodily objects. The key to these differences is that bodily locations are given on a non-Cartesian frame of reference. As brought out in the final section, this way of thinking about the phenomenology of bodily awareness has interesting and fruitful connections with current thinking about motor control.