Chapter 5 Waiting for the self

Jesse Prinz

1. HUME’S THESIS

In one of his most famous passages, Hume writes:
This remark admits of various interpretations: metaphysical and phenomenal, strong and weak. Here I want to defend a strong phenomenal version of Hume’s view. I want to argue that there is no phenomenal quality corresponding to the subject of experience – no phenomenal I.
This topic has been much discussed by philosophers, and, like Hume, most tend to rely on intuitions, introspective reports, and contestable transcendental arguments. My review will begin with Descartes, but I will be more interested in assessing lines of research that treat the phenomenal I as a topic for empirical inquiry. There have been a number of recent attempts to identify a subject of experience using the techniques of contemporary brain science. I will consider some of the more promising proposals and argue that they come up short in important ways.
Let me begin with some clarifications. Hume offers his introspective report in a discussion of personal identity, which is traditionally regarded as a metaphysical issue. Thus, one can interpret him as addressing the question of what it is to be a self and how a single self can endure changes over time. Hume’s so-called “bundle theory” identifies the self with collections of perceptions, and, elsewhere, Hume despairs of ever finding an adequate theory of endurance if this is right (Hume 1739, 635f.). My concern here is not metaphysical. Perhaps selves exist, and perhaps they endure. My concern is with the phenomenal thesis that Hume advances in the course of addressing the metaphysical question.
I take the phenomenal thesis to be this: among the various phenomenal qualities that make up an experience, there is none that can be characterized as an experience of the self or subject in addition to the qualities found in perceived features of the world, sensations, and emotions. Though we might convey the content of our experience using a sentence of the form, “I am experiencing X,” the actual qualities that make up the experience can be exhaustively surveyed by enumerating the qualities that constitute the experience of X. There is no remainder corresponding to the word “I” in the subject position of the sentence.
This interpretation remains ambiguous on a crucial point, however. The claim that there are no I-qualia in addition to the qualities of perception, sensation, and emotion might be understood in two ways. On a weak reading, there are qualia corresponding to the I, but these are nothing above and beyond the qualities of perception, sensation, and emotion. That is to say, I-qualia exist but are reducible to other kinds of qualia. I feel myself as a bundle of perceptions. When I say, “I experience X,” there is an experience of the I, but it is to be found in the experience of X. Call this the reductive reading of Hume’s thesis. On a stronger reading, Hume’s thesis says that there are no I-qualia, whether reducible or not. On this reading, we cannot save the phenomenal I by equating it with the qualia that correspond to something that the I would be described as experiencing. Call this the eliminativist reading of Hume’s thesis. I will be defending a version of this stronger interpretation.
Those who believe in the phenomenal I, whether reductive or not, claim otherwise. They claim that we can find an I in experience. Reductionists claim that the I is there in Hume’s bundle, and non-reductionists say that the I is something above and beyond the bundle. Some believers say the I comes and goes and others say it is always there in every experience. I reject both claims.
To be clear, the eliminativist about the phenomenal I need not be an eliminativist about the I in a metaphysical sense. The eliminativist is not necessarily what Strawson calls a no-subject theorist, nor a defender of Wittgenstein and Anscombe’s semantic thesis that “I” is not a referring expression. Perhaps “I” refers. Nor is the eliminativist committed to the view that we cannot find anything in experience on the basis of which an I can be inferred. Perhaps we are indirectly or inferentially aware of an I. Indeed, there may be an implicit I in every experience. In the concluding section, I will discuss several ways in which this might be true.

2. NON-REDUCTIVE THEORIES OF THE PHENOMENAL SELF

2.1 The Cartesian I

Philosophical thinking about self-consciousness got a major boost when Descartes formulated his cogito. In saying “I think” in the first person, and declaring that this is indubitable, Descartes implies that there is an I, which is directly accessed in consciousness. He then went on to speculate on the nature of this I, noting that it can be accessed with greater certainty than any thing in the physical world and, implying too, that it exists independently from any existing thought. I am a thing that thinks, he concluded, a res cogitans.
Descartes’ account is a paradigm case of a non-reductive theory of the phenomenal self because it implies that the self is present in experience but not reducible to anything else. Unfortunately, however, Descartes does not say much to back up this claim, and he has been accused, most famously by Lichtenberg (1765–99, 190), of smuggling in the I. It may be indubitable that there is thought, quips Lichtenberg, but not that there is a thinker. The I is not given in the experience of thought; we just experience whatever it is that we are thinking about.
A Cartesian might try to avoid this reliance on grammar by appeal to introspection. Perhaps we find an I in experience whenever we think. This strategy is unpromising. To see why, consider the kind of thoughts that lead up to Descartes’s cogito. Consider the thought we might express as, “I doubt that the table in front of me really exists.” In consciousness, this thought might present itself as a combination of verbal imagery (saying “this table may be a dream”), visual imagery (seeing the table and perhaps imagining it vanish suddenly), and emotions (a feeling of uncertainty or withered confidence). Now the question for Descartes is: can we find a phenomenal I in this conscious episode? If so, it certainly is not obvious. It seems I can have all these experiences without any experience as of an I.
One might think the I appears when introspection turns inward, and I recognize my own doubt. I move from doubting the reality of the table to judging that I have such a doubt. But this shift can be explained without bringing in any phenomenal I. To recognize my own attitude toward the table, I need only have a capacity to notice and label my feeling of doubt. Imagine that this is achieved the way ordinary perceptual object recognition is achieved: I match a current experience against stored, labeled records of prior states. When I turn inward, I am able to classify my feelings of uncertainty as such. In so doing, I do not seem to require any kind of conscious experience of a self. I just need to experience an emotional state. Indeed, phenomenologically, the experience of doubting that a table is real and the experience of recognizing such a doubt in myself may be extremely similar. In both cases, there is an experience of an emotion, some verbal narrative, and a table. An inward focus on the emotion may lead to an intensification of that feeling, which is what happens when we focus on any perceived feature, and the verbal narrative might change from “the table may not be real” to “I am having doubts about the table’s reality.” But these changes, including the addition of the word “I” in my subvocal report, do not bring in an experience of a subject; just the doubt itself, a particular feeling, comes into sharper view.
Concerns about the limits of introspection are not limited to Descartes. We find contemporary thinkers falling into the same trap. One example is Kriegel (2005) who claims that it is introspectively apparent that every phenomenal experience contains an element of me-ness. It is usually not the focus of experience, he concedes, but it is there in the periphery, in the way one might be aware of a canvas while looking at what a painting depicts. He uses this introspective intuition to ground a self-referential theory of consciousness according to which consciousness arises when mental states represent themselves. Self-referential states have the requisite duality, according to Kriegel: they represent features of the world, and they represent representing, which imparts a sense of me-ness (notice that this theory is reductive, unlike Descartes’).
There are some problems with Kriegel’s account, which I will no more than flag here. One issue concerns the biological plausibility of the claim that perceptual states represent themselves. Neurons are said to represent perceptual features in virtue of their response-profiles: edge detectors fire when edges are present. There is nothing about these neurons in virtue of which we should say they also represent themselves: they do not cause themselves to fire, and when perceptual cells fire unconsciously, they do not change their dynamics in any way that could be plausibly interpreted as a loss of a self-representing function. Another concerns the claim that self-reference would impart a sense of me-ness. Self-reference is cheap. The word “word” refers to itself; this sentence refers to itself; a concave mirror represents itself infinitely. None of these things can be said to have a sense of me-ness as a result. It is not clear why a perceptual state that also represents itself would give rise to a phenomenal I. Finally, and most importantly, Kriegel’s theory rests on an introspective datum that might be challenged. When I introspect, I do not find the me-ness that Kriegel describes. I find only the world. Which one of us is introspecting more accurately? Who knows! Clashing reports of this kind are notoriously difficult to resolve. That is precisely why introspectionist psychology failed.
In summary, the long and distinguished tradition of arguing for a phenomenal I by appeal to intuition may be a blind alley. Of course, Hume’s argument rests on introspection as well. He says he finds no I when he looks inward. But Hume is at a slight advantage. He does find other things that everyone can agree on: perceptual, sensory, and emotional qualia. Given the obviousness of these, one might wonder why the phenomenal I is not obvious to everyone. That asymmetry gets his position off the ground. But, it does not settle the case, because there is always a chance that Hume and his followers are not introspecting well. So we need to look for other kinds of evidence.
In making these remarks, I do not mean to be suggesting that we should do without introspection entirely. Studies of consciousness often depend heavily on first-person reports. Rather the claim is that introspection on its own may not suffice. Those who claim to find an I in experience would do well to find some non-introspective convergent evidence. Perhaps a substantive theory of what the phenomenal I consists in, backed up by non-introspective evidence for whatever the theory postulates, can help the Humeans see that there is an I in experience, after all.

2.2 Self, lost and found

Descartes supposes that the I is always present. There is no experience without an experiencer. For some, this does not ring true phenomenologically. There is a common phenomenon that we refer to as “losing yourself” in experience. Sometimes we become so absorbed in an activity that we seem to lose awareness of everything else, including, it is said, one’s own self. Taken at face value, this phenomenon would be hard to square with the Cartesian perspective on the topic. When we lose our selves, we do not stop thinking, as a Cartesian might suppose, and we do not lose conscious unity. More importantly, the phenomenon suggests that the consciously experienced self can come and go. If that is right, there is a clear strategy for trying to find out what the conscious self consists in: we can compare what happens when we lose ourselves to what happens when the self reappears.
This is exactly the strategy that has been pursued by a group of researchers in contemporary neuroscience. Goldberg et al. (2006) used fMRI to test brain activation while subjects listened to brief musical recordings or looked at pictures. In one condition, they had to say whether the pictures depicted animals and whether the recordings contained a trumpet. They were given ample time to reflect on their choices. In another condition, they had the same instructions, but the recordings and pictures were presented very quickly. The authors reasoned that when subjects are forced to answer quickly they would tend to lose themselves in the task. Finally, there was a condition in which subjects were asked to decide whether they liked recordings and pictures, a task that requires introspection, and is thus maximally self-involving. They found that the introspection condition showed greater activation in the superior frontal gyrus (SFG) than the other two conditions, and SFG was actually suppressed in the rapid task, suggesting that self-awareness was actively being suppressed. In another pair of studies, they also found that SFG was highly active when people read a word list and decided whether each was true of themselves, as compared to when people read a word list and decided whether each word is a noun or a verb. The word list included a wide range of objects and activities, which are not especially emotional in nature, unlike the picture preference task (e.g. study, run, coffee, and bus). Here again, SFG seems to correlate with the degree of self-involvement. One of the authors of this work was also involved in a further study that showed minimal SFG activation (along with other frontal structures) during passive watching of a movie, suggesting that viewers lose themselves in the film. All this evidence suggests that SFG is active in self-related tasks, and inactive when people lose themselves. The authors conclude that it is a neural correlate of self-awareness.
Interestingly, the authors also imply that SFG is not reducible to any other aspect of experience. It is an area associated with perception (it is active with both sounds and pictures) or emotion (it is active when we reflect on emotional preferences but also on the applicability of any words pertaining to the self). This might be taken to suggest that it is the neural correlate of a pure sense of self. In this respect the work differs from research that will be summarized below, which attempts to reduce self experience to some kind of bodily experience.
To assess this research, it is important to have some perspective on SFG. We cannot conclude that this is the neural correlate of the self without knowing what other roles it might play in cognition. The answer that emerges from other areas of cognitive neuroscience is that SFG plays a role in working memory, especially when working memory tasks are highly demanding (du Boisgueheneuc et al., 2006). That is interesting for two reasons. First, Goldberg et al. predict that SFG should diminish with task difficulty, because we tend to lose ourselves when cognitive load increases. But the research on working memory shows just the opposite. Second, the fact that SFG is involved in working memory can explain the Goldberg results without assuming that it is a correlate of self-awareness. In order to decide whether a picture or a song is appealing one has to focus attention on the song as well as one’s reactions to the song. Holding two things in mind at once requires more use of working memory than mere perceptual classification. And searching memory to see whether words apply to one’s self requires more working memory effort than deciding whether they are nouns or verbs. In both cases, the authors are comparing a relatively passive bottom-up classification task to one that requires reflection. In their high-speed task, reflection would slow down performance, so working memory is inhibited. And, when we watch movies, we often absorb the content without reflecting. Thus, all the results are consistent with the view that SFG is a structure whose function is to play a role in demanding forms of working memory – the kind that are characteristic of reflection. There is little pressure to say it is the neural seat of the self.
Goldberg et al. might object that this deflationary reading is ad hoc. After all, they explicitly ask subjects to introspect by giving them questions about their preferences and traits. Isn’t it obvious that the tasks require self-awareness? I do not think so. Or, rather, I do not think the tasks require any special phenomenal character that corresponds to the experience of oneself as a subject. The picture/music preference task involves emotional responses to stimuli. These emotions exist within the minds of the experimental participants, of course, but there is a difference between feeling an emotion and feeling a phenomenal I. If you introspect and discover delight in a piece of music, your experience contains delight and the music. You might describe it by saying “I like this music,” but you could equally well say, “there is a delight response happening along with the music,” without using a first-person pronoun. Likewise, when you decide whether a word applies to you (“Do I run?” “Do I take a bus?”), you may call up memories of your morning jog or your morning commute, but there need be no experience of a self in these. The recollection may be relived from a particular point of view (running from this perspective), but points of view need have nothing especially self-like. Notice that a movie camera presents the world from a point of view. To assume that such memories involve a phenomenal I begs the question against those who think the recalled activities do not involve a phenomenal I. When I board a bus, I experience the bus, but not a subject of that experience.
These lines of reply might be summarized by saying that the Goldberg et al. study can at best be regarded as an investigation of the self as object, rather than the self as subject. In their tasks, we report things about ourselves, but, in so doing, we are treating the self as just another thing in the world with certain describable features. We are not experiencing ourselves acting as the subject of thought or experience – this not elusive self as I. They do not establish that some thoughts have a qualitative component that occupies the position that the word “I” occupies in self-ascriptions, such as “I like this music.” As Lichtenberg might say, they establish the liking, not the liker.

3. REDUCTIVE THEORIES OF THE PHENOMENAL SELF

3.1 Feelings

Damasio came into widespread recognition because of his (1994) work on emotions. Here he resuscitated a theory that originates with William James and a Danish physician, Carl Lange. On the James–Lange theory emotions are not the things that cause us to laugh, or cry, or flee, or fight. Rather, they are perceptions of patterned bodily changes that include such characteristic responses. When an emotionally significant stimulus is encountered, it triggers a bodily change that prepares us for a behavioral response, and then the mind perceives this change, and our experience of that bodily perception is the emotion. Damasio argues that contemporary research in psychology and neuroscience (including his own field, neurology) supports the James–Lange theory, and I have argued at great length elsewhere that he is right (Prinz 2004). But, Damasio also thinks that this story can help provide an account of self-consciousness, and here we part company.
Damasio’s (1999) story goes like this. Emotions are short-lived events that arise in response to specific elicitors, but emotions are not the only mental events in which we perceive changes in our bodies. In reality, we are perceiving our bodies all the time, and emotions are just one special case of this broader phenomenon. Damasio (1994) calls such perceptions “background feelings” and, despite the term, he thinks these can go on unconsciously. In his (1999) book, he says we can think of these background feelings as a “proto-self” for reasons that will become clear in a moment. Background feelings are not constant over time, but rather change incessantly as objects in our environment impinge on our senses. When we see something or hear something or smell something, the body reacts. Damasio speculates that the nervous system must keep track of these changes; it must monitor how the world is affecting the body at all times. To do this, there must be “second-order maps” – representations of the first-order bodily representations as they undergo changes. Damasio thinks these second-order representations are conscious, and he thinks they constitute a basic form of self-consciousness shared by humans and other animals. To experience the feelings constituted by fluctuations in your body is, according to Damasio, to experience yourself. Damasio calls this the “core-self” and he distinguishes it from the “autobiographical self” which involves the reliving of episodic memories. The autobiographical self is not found in all animals, and can be absent in people with anterograde amnesia. Its status as a form of self-consciousness depends, in part, on the fact that when we relive memories, we experience the core-self acting in events from the past. The core-self is essential for self-experience. As I understand him, Damasio thinks the core-self is necessary and sufficient for having an experience of one’s self as a subject.
The bodily feeling proposal has some intuitive plausibility. First of all, the body is part of the person, so an experience of the body, unlike an experience of some object out there in the environment, is an experience of the self, in some sense. Second, changes in the body reflect a person’s subjective responses to the world, including emotional responses, as well as the myriad unnamed responses we have to everything we encounter. Everything makes us feel some way, and tracking those feelings is, to that extent, tracking subjectivity. Third, there may be repeatable or predictable bodily responses that correspond to enduring aspects of the self such as values and personality traits, which are part of personal identity. The way my body reacts reflects something about who I am as a person, and I can recognize those reactions as such. I can also notice that, on some days, I do not “feel myself” because my reactions are different than they normally are. Here again, we find something very self-like in the experience of the body.
Given all this, it is very tempting to adopt the Damasio–James approach and identify the phenomenal I with the experience of bodily feelings, associated with second-order bodily maps in the brain. But I think this temptation should be resisted. I think the experience of our bodily feelings is neither necessary nor sufficient for a phenomenal experience of one’s self as a subject.
Let us begin with necessity. The intuition that bodily feelings contribute to self-awareness is strongest when we imagine encounters with physical objects that matter to us. If a loud noise startles me, for instance, I will feel a strong sudden response in my body. Attention is drawn inward as I feel my body reacting to the sound. But there are many other cases where the experience of the body is less pronounced without any loss in the impression that I am the subject of my experiences. Consider intellectual exercises such as reflecting on philosophy, answering crossword puzzle questions, or performing calculations in your head. Such activities might engage the body but, intuitively, they need not. To take one simple case, imagine counting backwards from 100. It seems very plausible that, while doing this, one periodically loses conscious experience of the body for brief moments. Indeed, there is evidence that tasks involving working memory tend to suppress activity in brain structures associated with emotion and bodily experience (Yun et al. 2010). But, when that happens, the task seems no less agentic, no less self-involving. I am not claiming that there is some phenomenal I present in counting backwards, which is preserved when we stop experiencing the body. I think there is no experience of an I in either case. My present point is that Damasio’s theory predicts that a phenomenal I should disappear when body awareness does, and this seems implausible. Body awareness does not seem necessary at all.
Nor does body awareness seem sufficient for a sense of self. To see this, consider the phenomenon of emotional contagion (Hatfield et al. 1993). Emotions are catching. If I see you grimace in anger, I will undergo the corresponding bodily change, and feel a flash of anger too. It turns out that this capacity may play a crucial role in emotion attribution. My ability to know what you are feeling depends to some degree on my ability to catch your emotions. We use our own bodily switchboard to attribute feelings to others. This is confirmed in many neuroimaging studies, showing that the same brain structures are used to both experience emotions and to recognize them in others, and injuries to these structures result in deficits of both experience and attribution (Adolphs 2002). Now, imagine a case where I recognize an emotion in you. Perhaps I see that you are mad. The science suggests that I do this by first mirroring your bodily expression in my own body and then feeling that change in me. Still, there is no temptation for me to say, I am mad. I might not be. I feel madness, but I do not feel like it is mine. The madness I feel is, in some sense, yours. I feel through myself; but it does not feel like myself that I am feeling. There is no subjective ownership of the felt state. This suggests to me that we can feel patterns of change in our own bodies without that feeling constituting a phenomenal I. This may sound paradoxical (how can I feel my body as you?), but perplexity subsides when you recall that a bodily feeling is just an inner state that registers basic physiological facts: a heart is racing, muscles are tensed, breathing is constrained, and so on. It seems plausible that one could become aware of those facts without thereby experiencing them as facts about me. For example, if my brain could get wired to your heart, I could feel your heartbeat, just as I do when I put a stethoscope to your chest.
Finally, it should be pointed out that, if bodily feelings were the basis of self-consciousness we should find a correlation between these feelings and the degree to which one experiences a sense of self. But that does not seem to be the case. The sense of self neither increases or decreases with felt bodily changes. For example, consider very intense emotions. These do not seem to engender a greater phenomenal sense of self than mild emotions do. Terror after hearing an intruder enter your house does not feel more self-like than does mild delight while strolling on the beach. Indeed, with intense emotions focus is often more outward than inward: the terror makes you forget yourself for a moment and focus intensely on the sounds coming from the intruder. Conversely, a sense of self can increase without a change in the intensity of bodily experience. Emotions, again, can illustrate. Consider a familiar contrast between shame and guilt. With guilt we focus on some offending act, and, with shame, we focus on the self. When ashamed, one feels like a bad person. Both emotions can be equally intense, and hence the bodily feelings are equally vivid, but one seems to involve a greater degree of self-awareness. I am not suggesting that there is a phenomenal I in either of these emotions (I am skeptical about the phenomenal I), but if there is anything at all to the idea that experiences can relate to the self, shame seems to be a better case than guilt. The difference may involve the thoughts and actions that come to mind when shame is experienced: there is a desire to conceal one’s self, a feeling of being impure, and a self-conceptualization as a bad person. These specific contents, not the degree of bodily involvement, add up to a pattern of behavior and judgment that can be characterized as self-focused. The intuition that some mental states involve a greater degree of self-awareness seems to track the degree to which those states lead us to think about ourselves, not anything about their bodily character.

3.2 Ownership

One difficulty with identifying the phenomenal I with feelings and emotions is that there seems to be nothing paradoxical about an unfeeling self. Another difficulty is that the mere presence of feelings does not yet decide between self and other: we need a story about what makes feelings count as mine. Both of these concerns are addressed by another approach to the phenomenal I, which has been gaining momentum in cognitive science. It is the view that the experience of a self can be reduced to a feeling of body ownership. The phenomenal I emerges when we feel our own bodies and feel them as belonging to ourselves.
This view bears a resemblance to the view advocated by James and Damasio, because it places emphasis on the body, but it shifts from inner feelings and emotions, which are associated primarily with visceral changes, and focuses on bodily position and location. Here, the most relevant sensory qualities are proprioception and kinesthesia, which help us determine the configuration of our limbs and torso in space. By shifting away from the viscera, the view avoids an implausible commitment to the view that emotions and feelings are essential to the experience of selfhood. Perception of body position can be present even when experience of emotions and feelings is not.
That said, a mere feeling of a body in space is not sufficient for a sense of selfhood. Just as you can feel emotions without feeling them as yours, you can feel a body without feeling it as yours, even if it happens to be your own body. This may seem paradoxical, but the possibility is brought out clearly in some pathological cases. Consider, for example, alien hand syndrome in which a person comes to believe that one of his or her own hands belongs to someone else. For example, Moro et al. (2004) describe patients with right brain injuries who suffer from a lack of feeling in their left arms (a syndrome often co-morbid with unilateral visual neglect); when these patients cross their arms, some feeling is restored, but they continue to insist that the affected limb does not belong to them. Healthy people may also be capable of experiencing bodily sensations without feeling a sense of self. Indeed, this may be commonplace in social cognition. Just as we use our own emotion systems when perceiving emotions in others, we use our own capacity for bodily sensation when perceiving the bodies of others. For example, Keysers et al. (2004) found activation in somatosensory cortex when individuals watched another person’s leg being touched. Thus, any theory that identifies the phenomenal I with perception of bodily position needs a further element: an account of how the body feels like it is mine. This is often called a sense of ownership.
Body ownership has been intensively studied in recent cognitive neuroscience. One popular strategy is to investigate cases where the ordinary sense of ownership breaks down. Consider out-of-body experiences, in which one seems to occupy a ghostly imaginary body that hovers in the air or stands beside one’s real body. This may sound dubiously mystical, but actually such experiences are relatively common. Moreover, they can be induced in the laboratory, and they can also arise with high prevalence after certain brain injuries. Blanke and Metzinger (2009) review this literature, and they argue that research on out-of-body experiences can help us identify brain structures associated with body ownership. Normally, people feel as if they own their real bodies, but in these exotic cases, ownership shifts to a new imagined body. Blanke and Metzinger reason that the underlying mechanism may be the neural correlate of felt body ownership, which they regard as a minimal form of the phenomenal self. (More elaborate forms of phenomenal selfhood arise when this minimal self is integrated into a global workspace.) To find the mechanism, we can look for the brain area that is active when people have out-of-body experiences and that leads to such experiences when damaged. Here, all signs point to the right temporoparietal junction (de Ridder et al. 2007). This, Blanke and Metzinger surmise, is the neural correlate of the phenomenal I.
There is a problem with this inference, however, and it has to do with what we know about the right temporoparietal junction. This region is also known to be active in mental-state attribution (Saxe and Powell 2006), and when subjects imagine another person performing an action (Ruby and Decety 2001). Such findings suggest that the region has more to do with a sense of the other than a sense of the self. And, indeed, this makes sense in the case of out-of-body experiences. Such experiences typically involve seeing one’s real body (or sometimes the imaginary body) as if it were another person. So the temporoparietal junction may not be the seat of body ownership after all. To find the mechanism, we must consult other research.
The most promising line of investigation involves a class of perceptual illusions that are, in a way, the inverse of out-of-body experiences. When people have out-of-body experiences, they can see their real bodies without a sense of ownership. But there are also cases in which persons experience artificial bodies or body parts as their own. Under certain conditions, we can be made to feel like we are located in the body of a mannequin, a virtual avatar, or even a block of wood. The most studied demonstration of this phenomenon is the rubber hand illusion (Botvinick and Cohen 1998). Here, a rubber hand is placed in front of a person while that person’s real hand is hidden from view. Then both the real and the rubber hand are stroked using a paintbrush at the same time. After a short time, many people begin to experience the rubber hand as if it were their own. They even misremember the location of their real hand after the procedure is finished, locating it closer to the rubber hand than it actually was. The rubber hand illusion is well suited to study the sense of body ownership because it involves both a feeling of the body in space and an impression that the body belongs to the self.
Tsakiris et al. (2007) set out to identify the neural correlates of body ownership by inducing the rubber hand illusion inside an fMRI scanner. In comparison to a control condition in which the rubber and real hand are stroked out of sync, the illusory sense of ownership was associated with activity in the right posterior insula and the right frontal operculum. The frontal operculum has been associated with motor planning in a TMS study (Tunik et al. 2008), and the insula is associated with body perception and control (Critchley et al. 2004). Indeed, the right posterior insula has been directly implicated in other research as a basis of ownership. Farrer et al. (2003) found that activation in this region correlates with a sense of control in a task involving a virtual hand. Karnath et al. (2005) found that injuries here are associated with anosagnosia for paralyzed limbs, suggesting that an inability to monitor limb ownership results in a failure to recognize when an owned limb is not responsive.
To understand the role of the right posterior insula, it is helpful to understand how the rubber hand illusion works. We can experience our bodies using a number of senses: we can see a limb, feel something touching a limb, and experience position through proprioception. Normally such experiences converge on a single location in space, but, in illusory cases, that fails to happen. Seeing a rubber hand while feeling a paintbrush shifts the proprioceptive sense of location into the region occupied by the rubber hand. The visual representation of location dominates the proprioceptive, causing the shift. Against this background, it is reasonable to speculate that the right posterior insula plays a role in sensory integration. It brings the different sources of information together and facilitates the proprioceptive shift. The feeling of ownership may arise when senses are brought together. I can feel a hand without it feeling like mine (as in alien hand syndrome), but if the hand I feel is integrated with the hand I see in the right way then it feels like mine.
The results of the Tsakiris et al. (2007) experiment may seem somewhat puzzling at first. The authors found more activation in the right posterior insula when subjects experience the rubber hand illusion than in the control condition when there is no illusion. But, in both conditions, subjects presumably have a sense of ownership; the difference concerns only the accuracy of that sense. Moreover, in the control condition, there is greater activation in somatosensory cortex than in the illusion condition, even though both conditions involve a tactile feeling on the hand. Perhaps what is going on here is a shift in attention. In the control condition, subjects may be focused on the sensation in the hand, and then in the illusion condition they focus on the sense of ownership because they are surprised to feel the rubber hand as their own. In addition, the increased insula activation may reflect the extra neural work that is involved in forcing visual and proprioceptive information to come into alignment, despite the initial mismatch.
In summary, then, it does seem that the right posterior insula is playing a role in sensory integration and that integration of bodily representations supports judgments of ownership. Tsakiris et al. regard this as a kind of self-consciousness, and surely, in some sense, they are right. When I experience a body (real or otherwise) as mine, I am experiencing something as relating to myself. But things get a little bit thorny when we try to unpack that relation. Ownership is normally thought of as a two-place relation. Body ownership is a relation between a body and a self. It cannot be, however, that the phenomenal experience of body ownership involves two relata: a feeling of the body and a feeling of a self. Those interested in the feeling of ownership want to explain the latter in terms of the former. Thus, the feeling of ownership has to be understood in a non-relational way, if that makes any sense. Or rather, it is experienced in the integration of multiple experiences as of a body. Thus, there is an n-place relation here but no component is privileged as owner of the other components.
Once we see this, we can see a problem for any attempt to explain the phenomenal I in terms of body ownership. For the question arises, where is the I in experience? The I cannot be identified with any part, because each bodily perception on its own can be had without a sense of it belonging to the self. So the I must be experienced in the relation or, more plausibly, in the various components as they are related to one another. But what is it to experience two bodily perceptions as related in this way? It will not do to say we experience them as belonging to the same self, because then we need to experience a separate self to which they are related. Instead, we must just experience them as bound together like two features of an object. Perhaps we achieve this using a map strategy – perhaps visual, tactile, and proprioceptive experiences can be linked to a shared unconscious body map. This analysis treats bodily unity in the same way unity is explained more broadly. Body perceptions are bound together the way color is bound to shape. If this is right, it may cast doubt on the view that the experience of body ownership can be accurately described as a phenomenal experience of the self. Instead, it seems to be an experience of a bound whole, with no more claim to being me than my polysensory experience of your body when I simultaneously see you and shake your hand. In other words, what began as a theory of ownership turns out to be no such thing. It is just another case of polysensory binding and does nothing to bring anything especially self-like into experience.
Now hold on, you might object, surely when you unify the sensory components of your own body it feels more like yourself than when you unify the sensory components of someone else’s body. After all, I do not confuse myself for you. True enough, but this objection only serves to highlight the deeper concern. The question is, why don’t I confuse myself for you? One might think that the answer has to do with the specific qualities that get bound together. Proprioceptive and tactile sensations, in particular, seem to be uniquely self-involving. But this possibility has already been rejected. I can experience these qualities without experiencing them as me. Indeed, it may well be that when I form an integrated representation of someone else’s body, I do so by generating an integrated mental image using my own proprioceptive and tactile mechanisms. I may simulate what it is like to be the other. If so, when I encounter another I may have two tactile-proprioceptive unities in mind, and there may be nothing in those polysensory bundles as such that determines which one is me. This problem comes out all too vividly in the case of out-of-body experiences. The person who drifts outside of her own body will experience two polysensory somatic bundles. The fact that they are unified is not what makes her identify with one of these bundles as opposed to the other. She might feel both bundles simultaneously, but only identify with one. Thus, identification cannot simply be a matter of sensory integration. Something else must be going on.
Moreover, even if an integrated bundle of bodily perceptions were sufficient for a sense of ownership, it is far from obvious that it would be sufficient for having a sense of one’s self as a subject, as opposed to an object. If I feel a pain in the foot, I might be able to perceive that the affected foot is mine, but it does not follow that I experience the foot as a subject of thought, experience, and action. The foot does not correspond to the “I” in the ascription “I have a pain in my foot.” There is a gulf between “my” and “I.” The thesis of this chapter is that there is no phenomenal I, and integrated bodily representations do not allay that skepticism. If there were a phenomenal sense of ownership (which I doubt), it would not qualify as a phenomenal experience of being a subject.
In summary, I see several serious hurdles for those who want to find the phenomenal I in the experience of body ownership. First, a sense of ownership would not be sufficient for a sense of self as subject. Second, the leading theory of body ownership – integration of bodily perceptions – is not sufficient for explaining how we judge that a given bundle of bodily experiences are mine. Third, integration of bodily perceptions is not necessary for making such judgments. Fourth, there may not even be such a thing as the phenomenology of ownership.

3.3 Authorship

This proposal overcomes the four problems adduced in response to the suggestion that the phenomenal I derives from a sense of ownership. First, I said that ownership is not sufficient for a sense of oneself as subject, but, as just intimated, authorship is a form of subjectivity par excellence; to experience oneself as author is to experience oneself as the subject of an act – as the I, as opposed to the me. Second, I said the leading theory of ownership could not explain how a bundle of experiences feel like they are mine. Authorship may provide the solution. If I experience two separate bundles of bodily perceptions, I can figure out which one is mine by figuring out which one I control. A sense of control can provide a greater sense of mineness than mere sensory integration because it brings in a motor element that provides for a robust sense of possession; my body is the one that obeys my intentions in predictable ways. The element of control may help explain why the frontal operculum, a motor control structure, is active in the rubber hand studies. We imagine owning the rubber hand, in part, by preparing to issue action commands. Third, I said that sensory integration is too demanding, because people with just one bodily sense have no difficulty judging which body is theirs. Authorship handles this easily, because one can recognize one’s control over a single dimension of bodily perception, without reference to any others. Ian Waterman can confirm his bodily control using vision to see that his limbs obey his movements. Finally, I raised doubts about whether there even can be a phenomenology of ownership, because ownership requires an owner. With authorship, things are easier, because there is a functional and anatomical separation between the intentions or commands we issue and their subsequent effects.
One might try to argue against this approach by maintaining that authorship is never directly experienced but only inferred. One such argument can be distilled from Wegner’s (2002) critique of the conscious will. Wegner argues that we derive a sense of control when we experience a conscious intention that meets three conditions: it occurs just before an observed effect (priority), it is consistent with the observed effect (consistency), and it is the only salient event that meets these first two conditions (exclusivity). Wegner’s main point is that the sense of control can be an illusion, because these three conditions can occur even when there is no causal link between intention and effect. Suppose, for example, that my limbs are controlled by unconscious motor commands, but those commands also generate conscious intentions to move – intentions that dangle idly without having any impact on my behavior. The intentions occur before the movement, are consistent with it, and are the only conscious precursor, so I conclude that my intentions are efficacious. Wegner’s theory can also be adapted to raise a skeptical worry about the phenomenology of authorship. If he is right, control might not be something we experience; it may be something that we simply infer from conscious intentions and conscious bodily movements. This would spell trouble for defenders of the phenomenal I, because one can surely experience an intention or a movement without feeling any quality that corresponds to being a subject. Indeed, as Wegner points out, in some cases we experience intentions and actions without inferring control because alternative explanations become available. In one experiment, he asks subjects to press True or False on a keyboard in response to questions, but subjects are instructed to provide answers on behalf of another person by feeling minute muscle movements in that person’s hand, rather than providing their own answers. In actuality, the other person cannot even hear the questions, so the answers derive entirely from the subjects, but they nevertheless believe that the other person is selecting the answers. Intention and action are both experienced but subjects infer a lack of control (Wegner and Wheatley 1999). This experiment shows that control is not part of the phenomenology, but is only inferred by an argument to the best explanation.
The Wegner story gives us reason to be cautious in assuming that there is a phenomenal experience corresponding to authorship. It is possible that, in many cases, authorship is merely inferred. But Wegner’s theory is not the only game in town. In cognitive neuroscience there has been an active campaign to find the neural correlates of authorship, and the people conducting this research think that authorship can be phenomenologically experienced and they even have an explanation of how this occurs. The leading story stems from research on motor control. According to prevailing theories of motor control, intentional action begins when an intention is converted into a motor command by a mechanism called a “controller” (or sometimes an “inverse model”). The controller outputs motor commands, and it also sends “efference copies” of those commands to a mechanism called a “forward model” which uses the copy of a command to generate a prediction of how the body will move when the command is carried out. Meanwhile, the body carries out the command, and sensory transducers register the resultant changes, which are called reafferent signals. The reafferent signals are compared to the predicted outcomes by means of a comparator – a mechanism that determines whether predicted and actual movements are alike. This account has been adapted to explain the phenomenology of authorship (Wolpert et al. 1995). We experience actions as our own when a predicted action and a perceived action match. If our bodies move in unpredictable ways, we lose the sense of agency. This happens, for example, when something pushes into us, causing the body to move without a prior motor command.
The comparator model explains a number of interesting phenomena. It explains why it is so hard to tickle yourself (Blakemore et al. 2000). Tickling seems to depend crucially on unpredictability. If you try to tickle yourself, your motor system will accurately predict the hand movements, and you will lose the crucial element of surprise. The comparator model also explains some striking pathologies in the phenomenology of agency. There are cases of alien and anarchic hand syndrome where patients feel that an appendage is not in their control even though they can feel its movements (Biran and Chatterjee 2004). These patients characteristically describe the disobedient hand in the third person, as if it did not belong to them. The syndrome is typically associated with damage to the corpus callosum, which allows for communication between the hemispheres. That damage may prevent one hemisphere from predicting actions controlled by the other, giving rise to the experience of alien control.
So far, the cases I have been discussing involve motor behavior. A nice feature of the comparator model is that it can be adapted to explain the experience of authorship in cognitive processes as well (Frith et al. 2000; Hohwy 2007; though see Vosgerau and Newen 2007). One can explain the sense that a thought is my own by supposing that prior to generating a conscious experience of the thought, we anticipate what that experience will be like. The two are then compared, and if the thought is as predicted, it seems to belong to us. If not, it seems to come unbidden. Frith et al. (2000) use this model to explain the experience of thought insertion in schizophrenia. A thought that comes without being predicted may seem to come from some external source, as if it has been planted in the head. People with schizophrenia may experience their own thoughts this way because of a malfunction in their comparator mechanisms.
The extension of the comparator account to thought is a big advantage. One problem with the ownership theory discussed in the last section, as well as Damasio’s account of self-awareness in terms of feelings, is they place too much emphasis on the body. If there is a phenomenal I, it would be extremely surprising that it could manifest itself only through the body (see also Strawson 1999). After all, we are subjects of thought just as much as we are subjects of physical actions. There may even be conditions under which all bodily phenomenology is lost (sensory deprivation tanks? locked-in syndrome?), but there is no reason to think that such conditions essentially involve any less sense of selfhood than we find in the case of physical actions. For those who think the body is inessential to the sense of selfhood, the comparator model is surely attractive.
In sum, the comparator model provides an elegant and unified account of authorship. It locates the experience of authorship in a process that is well motivated by theories of motor control, and it extends nicely to cases of cognitive ownership as well. It also offers a tidy explanation of cases in which authorship seems to be lost. All this is very good news for those who want a reductive account of the phenomenal I. In authorship, unlike mere ownership, the self functions as the subject of a thought or act. Thus, authorship gives us an I rather than a mere me. It is, therefore, tempting to equate the phenomenal I with the phenomenology of authorship, which, in turn, can be equated with the experience of a successful matching process in an output/input comparator.
This account of agency differs from Wegner’s in that it treats authorship as an experience, rather than as a mere inference. But it can be easily reconciled with Wegner’s experimental findings and his skeptical hypothesis about free will. The comparator model does not presuppose the causal efficacy of conscious intentions, because it is consistent with the view that motor commands are initiated unconsciously. Wegner may be right that conscious intentions lack efficacy and are linked to actions through an inferential, rather than phenomenological process. Wegner’s experiments on erroneous lack of control may appear to conflict with the comparator model, but this appearance can be explained away by a more careful look at the findings. In Wegner’s true-or-false study, subjects judge that they did not choose which answer to type into the keyboard, but, if asked, they probably would not deny feeling a sense of control when pressing the keys. They feel like the initial choice of which key to press has an external source, but once that choice is registered, the subjects press the keys themselves and might experience a sense of authorship in doing so. In other words, these studies do not show that authorship is inferential, rather than phenomenal; they merely show that we can have false beliefs about the factors that lead us to carry out the controlled actions that we experience.
It may look like we have found the phenomenal I at last. Appearances, however, can be deceiving. I think the comparator model does offer a plausible theory of the processes by which we recognize agency and disruptions to agency, but I do not think it gives us the coveted I-qualia. I have two reasons for this disheartening assessment. First, I think that the comparator model cannot, in the end, give us a phenomenology of subjectivity in the required sense; it continues to present the self as object rather than subject. To see this, we need to consider the neural correlates of the comparison process (focusing here on the case of physical actions, as opposed to thoughts). One might think that comparison involves a conscious experience of a motor command being matched against a perceived movement. This would be something like an experience of authorship, because one would be experiencing the command that created an action together with the action caused. The command could stand proxy for the self, in an act of agentic control. But, in reality, the experience cannot be an experience of a motor command. For one thing, there is no evidence that we have motor phenomenology. The best theories of phenomenology are theories of sensory phenomenology, and every experience associated with motor response can be chalked up to a sensory experience of the corresponding movements. For another thing, the motor command cannot be directly matched against a perceived action, because motor command and action perceptions are couched in different neural codes; to make a match, we need to generate an anticipatory image of what the action will feel like perceptually and then compare that to the actual perception. And finally, neuroscientific research on comparator models using fMRI and cell recordings in animals suggests that the comparison takes place in the posterior parietal cortex – a somatosensory area – rather than motor cortex. If this is right, the feelings associated with authorship are really not experiences of an author and an authored act; rather, they are experiences of an anticipated act and a real act. To experience the body acting, whether through perception or anticipatory imagination, is to experience the self as object rather than as subject. Such an experience is not a phenomenal correlate of the “I” in first-person reports. It is a correlate of the me. When I say, “I feel myself act,” what I report is an action, not an actor. Put more pointedly, there is no experience of authorship. To suppose otherwise is to mistake a feeling that occurs when we are the authors of our actions for a feeling of authorship itself. It is to mistake the effect for the cause. When we use the experience generated by the comparator to judge that we are in control, the judgment is inferential, rather than a direct report of the phenomenology.
The second complaint I want to advance against the comparator theory of phenomenal authorship is that its defenders give the erroneous impression that the comparison process itself has a phenomenal character. If authorship had a phenomenology and the comparator theory of that phenomenology were true, then we might expect to find, in controlled action, a phenomenal experience of an anticipatory image being compared with a perceptual input from the body. But when we act, we find no such thing. We experience only our bodies in action. There is one experience, not two. It is a single unified perceptual experience of action, not a comparison between image and percept. It is the same for the phenomenology of thought. When we experience subvocal speech, for example, we do not experience an anticipation of this silent narrative, followed by the narrative itself – much less a matching between two verbal streams. We just experience one stream of speech. This suggests either that the comparison process is entirely unconscious, or that the anticipated act and the experienced act blend into each other seamlessly in experience. Either way, there is no hook to hang the selfy hat on. There is no item or process in experience that we can label the I. There is an experience of the action, but not the actor.
Against this deflationary reading, one might object that we must be able to experience ourselves as actors because, otherwise, we would not be able to consciously distinguish cases of control from cases where others act upon us. It does seem that we can tell the difference between lifting an arm and having an arm lifted by another, but, if my arguments succeed both should be alike; in both we just experience the movement. It will not help here to say that our own actions are accompanied by feelings of effort while externally caused actions come accompanied by resistance, for some self-controlled actions are effortless and some other-controlled actions meet with no resistance. The puzzle can be put like this: the distinction between internally and externally generated actions seems to be phenomenologically available, and the comparator model offers a good explanation, but on my account of the phenomenology of the comparator model, internally and externally generated actions should be phenomenologically alike because the matching process is unconscious.

4. CONCLUDING REMARKS: IS EXPERIENCE SELFLESS?

The foregoing has been an extended defense of Hume’s thesis against recent developments in cognitive neuroscience. Hume says that when he looks for a self in the flow of experience, he does not find one. He can find perceptions of the world and the body, but no subject. On a weak reading, Hume thinks there is no subject above and beyond the flow of perceptions, and, on a strong reading, he thinks none of our perceptions qualifies as an experience of one’s self as a subject. Both of these assertions have been challenged, and recent work in cognitive neuroscience has offered several theories, both reductive and non-reductive of the phenomenal I. It has been my contention that these theories fail. Some of them may point to ways we can experience ourselves as objects, but that we can do so has never been in dispute. Some may also provide some of the mechanisms and experiences on the basis of which we can infer that we are acting as subjects of our actions and thoughts, but they do not bring the self-qua-subject into phenomenal view. Thus, Hume seems to be right, even on a strong reading. There is no phenomenal I. If I wait for myself to appear in experience, I will never arrive. I might believe that I exist as a subject through inference and philosophical speculation, but I have never been acquainted with myself. I see indirect signs or, more strikingly, recognize my own agency at just those moments when agency is lost.
Fortunately, we do not need to end on this gloomy existential note. Perhaps the search for the self, like the wait for Godot, is a pointless exercise, but there may be a sense in which the I is present in virtue of its absence. To see this, consider three ways in which the self enters implicitly into experience. First, the self is present in the fact that we always experience the world from a perspective. It is a striking feature of consciousness that conscious states are always from a point of view. After all, sensory systems do encode information in vantage-point invariant ways as well, but these representations are unconscious. Consciousness is perspectival. Second, consciousness cannot present what lies beyond our senses; objects too far in the periphery or distance escape our field of view. Consciousness may fill in missing details, but it always presents the world as seriously bounded – bounded by the I. Third, the qualities of experience are dependent on our sensory apparatus. The colors and even the shapes we see are, in some sense, not given to us by the world. Shapes are out there, to be sure, but only at one level of analysis. The ladders that feel so solid under our feet are really swarms of atoms. So the senses do not simply pick up the world as it is; they select from a vast range of possible joints and impose order that would be invisible if we were different kinds of selves. Our goals, interests, and histories can contribute to this process of selection and construction, and, in that way, the contents of experience are not a transparent window onto the mind-external world, but are always instead filtered through the self. In these three ways, the self is always present in the perceptions given to us in consciousness. It is present, not as an item of experience, but as a kind of constraint. To echo Wittgenstein, echoing Schopenhauer, the self is the limit of my world. I cannot see the eye that sees, but seeing is constrained by the eye, and hence the I. For this reason, consciousness is not selfless; it is thoroughly permeated by the self. The self is absent if we look for it, but is always already there in each act of looking.1
1 My heartfelt thanks to JeeLoo Liu and Kelsey Fernandez for their extensive and helpful comments.