“How do you know that you think it will rain tomorrow?”
“How do you know that you have a headache?”
When read as demands for justification, these questions seem absurd. We don’t normally ask people to substantiate assertions like “I think it will rain tomorrow” or “I have a headache”. There is, at the very least, a strong presumption that sincere self-attributions about one’s thoughts and feelings are true. In fact, some philosophers believe that such self-attributions are less susceptible to doubt than any other claims. Even those who reject that extreme view generally acknowledge that there is some salient epistemic difference between (a) one’s belief that she thinks it will rain tomorrow, or that she has a headache, and (b) her belief that it is raining, or that another person has a headache.
Philosophers are chiefly interested in the above questions not as challenges to self-knowledge, but as inquiries about how, precisely, self-knowledge is achieved. They are moved by the observation that each of us seems, at times, to know her own mental states in a way in which others cannot. Do we really have such special or “privileged” access to what we think and feel? If so, how do we account for it? If we lack privileged access, how do we explain the widespread sense that there is a sharp contrast between self-knowledge and other-knowledge? These questions are the focus of the present volume, which includes a range of answers to them.
In this introduction, I will first describe several ways to interpret the claim that subjects enjoy privileged access to their own mental states and outline key considerations that bear on each. On some of these interpretations, the claim is that we enjoy an epistemically special access to evidence about our own states; on others, the distinctive feature of self-knowledge does not involve any sort of access to evidence. I will then introduce each of the contributions in this volume.
I. What is “Privileged Access”?
Here are the principal claims that have been made under the banner of “privileged access”. In these statements, S ranges over rational, cognitively well-developed persons and M ranges over mental states. The effects of limiting the scope of these terms will be considered below.
1. Infallibility: If S believes that she is in M, then she is, in fact, in M.
2. Self-intimation: If S is in M, then he believes that he is in M.
3. Epistemic asymmetry: S may have a type of warrant for the belief that she is in M which is unavailable to others.
a. S alone can possess non-inferential justification for this belief.
b. S alone can justify this belief introspectively.
4. Epistemic privilege: S’s belief that he is in M can, in principle, achieve a higher degree of epistemic certainty than others’ beliefs that he is in M.
5. Incorrigibility (or “first-person authority”): S’s claim that she is in M cannot be justifiably/reasonably disputed by others, or cannot be shown by others to be mistaken.
Each of these claims is open to further interpretation. One factor, which influences the force of (1)-(5), is the scope of “M”. On one possible position, for instance, we enjoy privileged access only to phenomenal states like headaches, and not to propositional attitudes like the belief that it will rain tomorrow. Alternatively, it may be claimed that one enjoys privileged access to their phenomenal states and to propositional attitude contents (like “it will rain tomorrow”), but one does not enjoy such privilege regarding the nature of one’s attitudes toward a propositional attitude content, e.g., whether one’s attitude towards “it rains” is belief, hope, or fear.
Another factor which influences the force of (1)-(5) is the modality concerned. It may be that, as things now stand, subjects enjoy an epistemically privileged access to their own mental states. But some will see this privilege as deriving from current empirical ignorance—specifically, from ignorance about the physical basis of mentality. They will claim that developments in science will allow us, by observing others’ brains, to achieve other-knowledge that is on a par with self-knowledge; and they will accordingly view (1)-(5) as contingently true at best. Others will claim that our privileged access is necessary, in that having privileged access to one’s own mental states is essential for rationality or agency, or that being the object of S’s epistemic privilege is essential to being truly a mental state of S’s.
Let us consider each of these construals of privileged access in turn.
Very few current philosophers accept (1), even in weak versions (e.g., as a contingent truth about self-ascribing phenomenal states). The doctrine of infallibility is often associated with Descartes, but it is not obvious that Descartes was committed to infallibility. He surely held that we cannot go wrong in self-ascribing thoughts if such self-ascriptions are based on the right kind of introspective evidence; but this commits him only to the claim that appropriate introspective evidence (and not merely the presence of a self-ascribing belief) suffices for true self-ascriptions.
In any case, counter-examples to (1) are easy to find. My psychotherapist tells me that I desire a change in career; because I trust her, I become convinced that I desire a career change. But her belief is based on a misinterpretation of my dreams. I do not, in fact, desire a career change, so my self-ascribing belief is false. Examples of false beliefs about one’s own phenomenal states are also available. A neuroscientist believes that she is looking at an image of her own brain, which is currently being scanned. She has found that the state represented by the image is highly correlated with headaches, and concludes on this basis that she currently has a headache. However, determined to continue her work, she vows not to attend to the (putative) headache sensation. Now suppose that she doesn’t have a headache. She is, in fact, viewing the image of someone else’s brain. (A competitive colleague, wanting her to quit for the day, has rerouted the wires so that the imaging screen is connected to a migraine patient.) In that case, she believes that she has a headache when, in fact, she doesn’t has one.
Even if (1) is indefensible as stated, more qualified infallibility claims remain tenable. The cases I described were cases in which the operative evidence for the self-attributing belief was not introspective. (Arguably, no evidence of any kind was operative in the case of the sweet taste.) Perhaps one cannot go wrong if one self-ascribes a mental state on the basis of introspection alone, without relying on perceptual evidence. Obviously, this will depend on what introspection involves; on some definitions of introspection, it may be trivially true that introspective beliefs are infallible. The interesting issue about infallibility is this: are there any conditions, which, if built into the antecedent of (1), would yield a non-trivially true infallibility claim?
Self-intimation is the complement to infallibility. According to (2), to have a mental state is to believe that one has it. This is dubious in its unqualified version. Some of the counter-examples to infallibility will also undermine self-intimation: for instance, when I falsely believe that I desire a career change, I also fail to believe what is true, viz., that I feel satisfaction about my career. For I may be deceived about my attitude towards my career without thereby being irrational; and it would be irrational to simultaneously believe both that I wanted to change careers and that I am satisfied with my current career. The doctrine of self-intimation also faces a deeper problem. Beliefs are themselves mental states; so if (2) is correct and one has a belief corresponding to every mental state one is in, then anyone who has at least one mental state would, it seems, have an infinite number of mental states. The only way to avoid such a regress while endorsing (2) categorically would be to maintain, implausibly, that there is a special class of highest-order beliefs which simultaneously register the presence of lower-order states and, reflexively, register their own presence.
However, the doctrine of self-intimation can be rendered more palatable by strengthening the requirements for satisfying the antecedent, or by weakening the requirements for satisfying the consequent. To strengthen the antecedent, limit the scope of M to states that are conscious (or, perhaps, to phenomenal states). To weaken the consequent, construe beliefs as dispositional: e.g., one believes that one is in M so long as one is such that, if one considers whether one is in M, then one will have an occurrent, episodic belief that one is in M. This results in the following, plausible self-intimation thesis. “When one is in a conscious state M, then one is such that, if one considers whether one is in M, one will occurrently believe that one is in M.”
The doctrine of self-intimation is inspired by the idea that paradigmatically mental states are conscious states. The term “conscious” has epistemic connotations similar to those of “aware”. Consider: “he was conscious of a knock on the door”, “she was conscious of her bank balance”. These epistemic connotations do not mean that all conscious states are states one is conscious (aware) of; as Fred Dretske notes, we should be sensitive to “the distinction between the content of awareness and the awareness of content”. (xx1xx) Still, the fact that a mental content is a content of awareness may render it especially well-suited as an object of awareness, that is, as a target for a higher-order belief. In this spirit, Sydney Shoemaker endorses a qualified version of (2), according to which M ranges over propositional attitudes, second-order beliefs are defined dispositionally, and S ranges over rational, conceptually capable persons. “[S]econd-order beliefs, and the self-knowledge they constitute, are supervenient on first-order beliefs and desires plus human rationality and intelligence.” (xx48xx)
(3) Epistemic Asymmetry
Because epistemic asymmetry (3) is a very broad claim, it is not susceptible to easy counter-examples. The thesis must be filled out by a specific contrast between self-knowledge and other-knowledge. On one traditional view, self-knowledge differs from other-knowledge in being non-inferential. While I know your mental states only by inferring them from your behavior, including linguistic behavior, I can know my own states directly. That is, self-knowledge need not involve any mediating observations of my behavior—e.g., my reaching for a glass of water—or empirical generalizations—e.g., to the effect that one typically reaches for a glass of water only when one is thirsty.
The issue about inference is linked to a larger methodological issue. It appears that you sometimes use a distinct method to determine what you think and feel, a method known as “introspection”, which will never yield knowledge of what someone else thinks or feels. Some philosophers are suspicious of the term “introspection”, with its connotation that understanding one’s own states involves looking inward. This suspicion stems, in part, from a denial that mental states are internal to the subject in any principled sense. (I return to the issue of internality below.) A related source of the suspicion is the denial that there is a distinct faculty of introspection. Accounts of self-knowledge which assimilate the faculty responsible for self-attributing beliefs to other rational belief-forming faculties would seem preferable, on parsimony grounds, to accounts, which posit a special introspective faculty. So-called “inner sense” accounts mitigate this difficulty by modeling the introspective faculty on more familiar, perceptual faculties.
(4) Epistemic Privilege
Interestingly, even if we grant that there is an epistemic asymmetry between self-knowledge and other-knowledge, more needs to be done to show that we enjoy epistemic privilege (4) vis-à-vis our own mental states. For non-inferential beliefs can be false; and there is no guarantee that a uniquely first-person belief-forming method or introspective faculty will be more reliable than a third-person method or faculty. Nor does epistemic privilege (4) imply an asymmetry in type of belief-forming process or faculty (3). It is possible that I can achieve a higher level of certainty, regarding my own mental states, than anyone else can achieve, even if others use the same method to determine my beliefs that I do. Perhaps the shared method yields greater certainty when applied first-personally than when applied third-personally.
Although there are no direct entailment relations between epistemic asymmetry and epistemic privilege, each of these can partially support the other. One way to argue for epistemic privilege is to combine a defense of epistemic asymmetry with the claim that the method or faculty uniquely employed in self-knowledge yields greater certainty than those which issue in other-knowledge. And an argument in the opposite direction is also possible: if one can establish that we have epistemic privilege, then one can make a case for epistemic asymmetry by showing that a difference in process, method, or faculty best explains the disparity in levels of certainty.
The thesis of epistemic privilege concerns the degree of certainty which self-knowledge can attain, relative to other-knowledge. Because it does not have consequences for every case of self-knowledge, this thesis will not be undermined by counter-examples as (1) and (2) were. To evaluate it, try the following thought experiment. Reflect on a current sensation of yours—preferably, an intense sensation. Consider how confident you are that you are feeling that sensation, while you are reflectively attending to it. Can you conceive being equally confident that someone else has a particular sensation? If you cannot, this thought experiment lends some support to the claim of epistemic privilege. Of course, it doesn’t conclusively establish this claim. Psychological confidence is not epistemic certainty; and the inability to conceive believing with equal confidence that someone else has a sensation may be due to a failure of imagination. Still, this result furnishes some evidence for epistemic privilege. To argue that we lack epistemic privilege, one must claim that the thought experiment doesn’t have this result, or that psychological confidence doesn’t reflect the intuition of epistemic certainty, or that the inconceivability at issue is merely the product of a cognitive limitation.
Finally, let us turn to the incorrigibility thesis (5), which says that it is unreasonable or improper to dispute another person’s self-attribution. As with infallibility and self-intimation, incorrigibility is a viable doctrine only when the thesis is appropriately qualified. Even if my psychotherapist sometimes makes mistakes, as in the case above, she may at times correctly, and justifiedly, dispute my self-ascriptions. For example, she may have good reason to believe that I am self-deceived (in current parlance, in denial) about a certain attitude of mine, which is repugnant and hence painful for me to acknowledge. Now such corrections by others have their limits: if you pinch yourself, and carefully attend to the twinge you feel as a result, you will not likely be willing to accept another’s contention that, in fact, you are not currently feeling a twinge. This sort of case suggests that your evidence that you’re currently feeling a twinge trumps any competing evidence possessed by anyone else. It thus explains incorrigibility (5) by some sort of epistemic asymmetry (3) and/or epistemic privilege (4).
While some will account for incorrigibility in this way, the doctrine of incorrigibility is historically tied to the rejection of evidence-based views. It has been motivated by the denial that incorrigibility derives from the subject’s unique access to internal or “private” evidence about her own states. Ludwig Wittgenstein was deeply suspicious of the notion of privacy, which the “access to evidence” explanation requires. Approaches to self-knowledge inspired by Wittgenstein’s rejection of irreducibly private facts favor an alternative diagnosis of incorrigibility. In ordinary circumstances, to question another’s self-attribution is to violate the linguistic practices which ground meaningful language use. Just as one who moves his rook diagonally is no longer playing the game of chess, one who refuses to take a self-attribution at face value is no longer playing the “language game”.
Philosophical theories about the mind tend, unsurprisingly, to shape philosophical accounts of self-knowledge. Of course, some theories about the mind are themselves fueled by observations about self-knowledge; the relation of influence between these is not one-directional. Most currently influential theories of the mind are naturalistic. Naturalism about the mind is the view that mental states and processes are continuous with other states and processes in the natural world. According to naturalism, the sort of facts and laws that will explain mental phenomena will be the same as, or similar to, those which explain non-mental phenomena.
The claim that subjects enjoy privileged access to their own states poses a prima facie problem for mental naturalism. As explained above, robust forms of the privileged access thesis take self-knowledge to be epistemically special, in that it differs from perceptually-based knowledge (including knowledge of others’ states) in a deep, principled way. To preserve naturalism, the epistemic specialness of self-knowledge must be accommodated within a larger scientific theory. According to many, a naturalistic account of self-knowledge will construe the epistemology of self-knowledge as fundamentally similar to the epistemology of perception. Perception is believed to be relatively well-understood; more importantly, even while we don’t fully understand how perceptual faculties operate, perception seems to pose no deep mystery. Too stark a difference between self-knowledge and perceptual knowledge threatens to make self-knowledge appear “magical”, that is, outside of the naturalistic realm. (Why would a profound difference between these make the latter but not the former seem magical? Simply put, this is because perceptual knowledge of non-mental physical facts is considered less obscure than introspective knowledge of mental facts.)
An account of self-knowledge motivated by these considerations is the inner sense theory, according to which self-knowledge is the product of a quasi-perceptual faculty of introspection. Dretske (this volume), himself a devoted naturalist, targets the inner sense theory. He raises the issue by asking: “How do you know you are not a zombie?” In philosophical parlance, a zombie is a creature that is physically similar to ordinary humans, but which lacks sensations. Dretske trusts that we do know that we have sensations, but he raises puzzles about how we know this. In particular, Dretske is at pains to show that ordinary perceptual experiences, which involve the having of sensations, do not themselves constitute an awareness of the sensations—rather, they afford only an awareness of external, physical objects. On his view, this observation weighs against the inner sense theory, which construes introspection as a kind of inner perception. More generally, it raises doubts about the epistemic status often accorded to self-knowledge. “Maybe our conviction that we know, in a direct and authoritative way, that we are conscious is simply a confusion of what we are aware of with our awareness of it”. (xx00xx)
William Lycan is a leading proponent of the inner sense theory, and his contribution defends the theory from Dretske’s challenges. He contends that introspective awareness of sensations is similar to perceptual awareness of physical objects in salient ways: e.g., in introspection we become aware of experiences by becoming aware of properties, which we use to identify them. And while it also differs in some way from perceptual awareness, Lycan argues, the differences are open to principled explanations. In addition to defending his view from Dretske, Lycan raises doubts about Dretske’s own views about self-knowledge. He applauds recent changes in Dretske’s account of self-knowledge, but maintains that certain problems persist. Specifically, he questions whether Dretske’s view affords an adequate epistemology for self-knowledge, one that truly explains how self-attributing beliefs are justified, without making such justification trivial; and he doubts whether Dretske’s view can do justice to the wide range of information about one’s own states that introspective reflection provides.
Despite this divergence in views, Dretske and Lycan are both representationalists, and are thus in agreement about a crucial issue about the mind. Representationalists claim that the content of beliefs, desires, sensations, and other mental states is exhausted by the states’ representational content. Representationalism about the mind is a view that they share with Michael Tye. And Tye makes an observation similar to Dretske’s: When you introspectively reflect on a sensory experience, Tye says, you are aware of the qualities (blueness, say) as qualities of external objects, not as qualities of your experience itself. Tye uses this “transparency” of experience to support his representationalist view of phenomenal states, according to which “visual phenomenal character is representational content of a certain sort, content into which certain external qualities enter”. (xx4xx) Still, Tye maintains, in being aware of the qualities of external objects in this way, one is also aware that one is in a given phenomenal state. Introspective knowledge of sensations occurs through this process: one is aware of the qualities of external objects, and this awareness “triggers” an application of the relevant phenomenal concepts. Because this triggering is automatic, the awareness of the sensation is direct and non-inferential.
Not everyone agrees that representationalism can accommodate self-knowledge. Joseph Levine argues that representationalism of the type accepted by Dretske, Lycan, and Tye cannot accommodate first-person privilege about one’s own sensations. For their representationalist view identifies the content of a sensation with the property of representing certain features of the world, and those features of the world are external to the subject. In this sense, their representationalism is externalist. Since the subject lacks any sort of special access to external features of the world, it seems that she is unable to determine, on non-perceptual (introspective) grounds, her current thoughts or sensations. Levine distinguishes this objection, and its consequences for externalist representationalism, from the much-debated issue about whether externalist theories in general can accommodate privileged self-knowledge. For the view under consideration identifies sensations with their externalist representational content, and so excludes the possibility that a purely internal difference can allow us to grasp our current sensations. The prospects for other views are brighter, Levine argues, because they fall short of identifying thoughts with external factors. This means that internal factors can partially determine content, and hence—even if the subject may not know her own sensations precisely—she can detect basic relations of similarity and difference between them.
Murat Aydede challenges an account of self-knowledge, which is accepted by some representationalists, namely, the displaced perception model. According to this model, one knows one’s own sensory experience by inference from a perceptual belief, where the perceptual belief typically concerns an external object. This inference must be grounded in a connecting belief to the effect that an observed state of the world indicates the presence of a particular sensation. Aydede considers a variety of connecting beliefs, which could underwrite introspective inferences, and argues that none are typically available to the self-attributing subject. He concludes that truly introspective knowledge of one’s own phenomenal states is not inferential. This runs counter to a view (once) defended by Dretske; but Aydede argues that it also affects Tye’s view, for he disputes Tye’s claim that his representationalism allows for non-inferential self-knowledge. He closes the piece by suggesting that only a direct referential relation between introspective state and introspected content will adequately capture phenomenal self-knowledge.
The next contribution steps back from particular theories of the mind, to raise a more general difficulty for accounts of self-knowledge. Paul Boghossian notes that there seem to be only three possible ways to justify introspective beliefs: on the basis of inference, on the basis of “inner” observation, or on the basis of nothing. But none of these seems tenable. The first seems to lead to a regress: if we know our own states only on the basis of inference, that inference is presumably from something (perhaps another mental state) which is itself known only inferentially. The second founders on what he describes as the “apparently inevitable” view that mental content is relational, that is, that the content of a state is not an intrinsic property of the state. All externalist views of content define content relationally, but so do nearly all non-externalist views. For instance, functionalism identifies mental states by their role in a larger causal network. So even non-externalist versions of functionalism, which maintain that the relevant causal network is internal to the subject, construe content as relational. But so long as content is relationally construed, we cannot grasp it by observing the state’s intrinsic features. And non-inferential observation appears to yield only knowledge of intrinsic features. Finally, we could not know our own states “on the basis of nothing”: accounts that take higher-order beliefs to automatically inherit the lower-order content of the states they concern will not provide for cognitively substantial self-knowledge. Boghossian does not suggest that we deny first-person privilege; his skeptical argument aims to show that we simply don’t understand how it is that we know our own states.
Christopher Peacocke rejects Boghossian’s claim that we know our states by observation, by inference, or on the basis of nothing; according to Peacocke, these categories are not exhaustive. In rejecting this claim, Peacocke attempts to steer a course between access to evidence accounts of self-knowledge and more deflationist accounts. His “intermediate position” focuses on consciously held beliefs: Peacocke claims that such beliefs are often the result of a conscious judgment, that is, a judgment which “contributes to what it’s like” for the thinker. Such a conscious judgment can provide a reason for the subject to judge that she has the corresponding belief. But this process is neither inferential nor observational; nor is the self-attribution a brute causal consequence of the first-order belief. For part of possessing the concept of belief (as our subject does) is to be disposed, upon making such a conscious judgment, to self-attribute belief in the content of that judgment. In this way, a conscious first-order event of judging (forming a belief) can simultaneously rationalize and cause the self-attribution, without serving as an evidential basis for the self-attribution. The first-order event is not observed, and it does not ground an inference. Instead, “the mental event itself … is the thinker’s reason for making the [self-attributing] judgment.” (xx90xx)
Self-attributions can be considered rational in a variety of ways. Peacocke focuses on the role of conscious, occurrent events—conscious judgments in particular—as both causing and rationalizing a self-attributing belief. Other philosophers emphasize how awareness of one’s own dispositional beliefs and desires is implicated in rational action. For instance, some degree of knowledge of one’s own beliefs seems required for rational agency, for only a self-aware person can become aware of her sources of information and thereby deal effectively with conflicting evidence. And a subject oblivious to her desires will have difficulty developing emotionally; autonomy surely calls for understanding what it is that motivates one to act. The next few pieces spotlight the relation between rational action and self-knowledge.
Shoemaker makes a case for a very close tie between rationality and self-knowledge. He proceeds by arguing that there could not be a “self-blind” individual, that is, one who has beliefs and desires, and is rational and conceptually capable, yet could not come to know his own beliefs and desires in any distinctively first-person way. He argues for this by making two claims. The first is that anyone who was rationally and conceptually capable would be able to use language assertively; the second is that anyone who can use language assertively will avoid utterances that exhibit a particular kind of paradoxical quality, such as “It is raining but I don’t believe that it is”. (The problematic feature of such utterances is known as “Moore’s paradox”, after G.E. Moore. See Moore 1993.) Since one who avoids Moore paradoxes will be apt to self-attribute beliefs in appropriate circumstances, he will be indistinguishable, behaviorally, from one who enjoys a distinctively first-person method of self-knowledge. And, according to Shoemaker, there is good reason to think that anyone behaviorally indistinguishable from one who enjoys first-person privilege also enjoys such privilege.
Charles Siewert addresses Shoemaker’s claim that our status as rational agents entails that we enjoy first-person privilege regarding our own mental states. He accepts that self-knowledge and rationality are intricately bound together; but, he contends, Shoemaker’s arguments for (and precise picture of) this bond must be reworked. For the bond between self-knowledge and rationality will not likely provide the materials for a full account of self-knowledge. One reason Siewert gives is this: the fact that I’m rational can’t justify a belief that I’m an accurate introspector, unless I can know that I’m rational. But this latter knowledge seems to require introspective evidence. Siewert also points out that rationality seems to be an all-or-nothing matter, while first-person privilege comes in degrees. In some circumstances, third-person countervailing evidence can compete with first-person access. Still, he thinks, the strength of the bond between self-knowledge and rationality underscores the importance of a philosophical account of self-knowledge for a larger understanding of ourselves as rational agents.
Wittgenstein’s arguments against the coherence of private languages inspire Crispin Wright’s rejection of an evidence-based account of self-knowledge, which he terms the “Cartesian observational model”. If a belief about a current sensation is a substantial cognitive accomplishment, he says, then it requires that the subject apply a concept to the sensation. Such concept use requires the possibility of error. Because the Cartesian model denies that we can err in grasping our sensations, this error can only consist in one’s mistakenly applying a particular concept to the sensation. The putative privacy of the concept prevents this mistake from being a mistake about how others use the concept; it can only lie in a misunderstanding of how one intended (perhaps through an initial baptism) to use the concept. But such a misunderstanding is possible only if one’s prior intentions render the concept fully determinate. Wright argues that intentions cannot function in this way, and hence that the Cartesian model of introspecting sensations is inaccurate. His argument about the nature of intentions also aims to undermine “access to evidence” models of our knowledge of our own propositional attitudes. He closes by sketching an “austere view” of intentional states, according to which it is partially constitutive of intentional states that the subject is the default authority on her own intentional states.
Like Wright, Richard Moran believes that a grasp of one’s own states is required for truly rational agency. He explicitly rejects the claim that privileged self-knowledge is a matter of having special access to evidence. For on his view, even if we do have special access to putatively “private” evidence, this will not explain the deep difference between self-knowledge and knowledge of others’ states. This deep difference entails that a full account of self-knowledge must draw on the fact that my beliefs, desires, and intentions “express my point of view on the world” (xx), and thereby partially constitute my identity as a rational agent. To illustrate this point, Moran draws on Jean Paul Sartre’s claim that consulting internal evidence to determine one’s own attitudes is a mark of self-alienation. A grasp of one’s own attitudes is truly first-personal only if, through this grasp, one understands the force of those external factors salient to justifying (or undermining) the attitude.
Dorit Bar-On and Douglas Long agree (with Peacocke, Wright, Moran, and Shoemaker) that no evidence-based account of self-knowledge will succeed. They advocate a “Neo Expressive” alternative, according to which avowals—such as “I’d really like some water!”—play a dual role. They are reliably produced by a state (in this case, a desire); but unlike other such products of the state—like “Water please!”—they are truth-assessable self-ascriptions. On a suitably dispositional construal of belief, they reflect the subject’s belief that she is in the state; and they count as warranted because they are reliably produced. In formulating this account, Bar-On and Long consciously attempt to incorporate the best features of evidence-based accounts and more austere accounts, while avoiding the pitfalls of each. They regard Wright’s view as too deflationary to accommodate the robust epistemic status of self-knowledge. And while they agree (with Shoemaker and others) that self-knowledge is intricately involved in rational agency, they think that the success of self-attributions must partially explain, and so cannot consist in, one’s rationality.
The next two pieces address the relation between introspective self-knowledge and the introspector’s status as a physical being. José Luis Bermúdez discusses the widely-accepted claim that, in introspection, we are “immune to error through misidentification” of the subject. For instance, even though we may err about the particular content of an introspected belief, we will not mistakenly attribute an introspected belief to the wrong person. Bermúdez accepts this claim but argues, against a suggestion of Shoemaker’s, that our immunity to this sort of error doesn’t explain another widely-accepted phenomenon: that we are not introspectively aware of a self, as such. According to Bermúdez, proprioceptive bodily awareness exhibits immunity to error through misidentification while at the same time rendering us aware of an embodied self; the introspective “elusiveness” of the self does not, then, derive from immunity to error through misidentification. Because proprioception affords first-person privilege, the elusiveness of the self is not required for privileged access.
In my contribution, I consider an objection that Arnauld made to Descartes’ introspective argument for dualism. Arnauld claims that Descartes’ argument illicitly presupposes that Descartes’ introspective concept of the mental is exhaustive. While Arnauld draws an analogy between Descartes’ argument and a flawed argument about geometrical kinds known through rational intuition, contemporary materialists object that Cartesian arguments for dualism are analogous to invalid arguments about physical kinds, known empirically. The Cartesian premise which requires defending is this: my current introspective grasp of my sensations provides a sufficient basis for evaluating the possibility of disembodiment. To determine what is required to adequately defend this premise, I examine similarities and differences between uses of perceptual evidence, evidence from rational intuition, and introspective evidence. I then draw on these results to sketch a defense of this premise; my defense exploits Saul Kripke’s well-known theory of reference and modality. While I have not proven that the Cartesian argument is sound, I hope to have shown that inferences from introspective data to ontological conclusions are more plausible than many have believed.
Ernest Sosa’s paper addresses the epistemological consequences of several traditional views of self-knowledge. He surveys these accounts—including Ryle’s denial of substantially privileged access, as well as (from the authors represented here) the views of Peacocke, Moran, and Wright. He maintains that none of these views provides a complete account of our first-person privilege. Moreover, none of them will serve the purposes of the foundationalist, who takes knowledge of one’s own conscious states to possess basic (foundational) justification. Considering the nature of those beliefs which do seem to exemplify first-person privilege, Sosa suggests that the foundationalist should exploit a virtue-based epistemology. Foundational beliefs, like paradigmatic cases of self-knowledge, are those which are non-accidentally true, and which stem from the exercise of a faculty that tends to yield non-accidentally true beliefs.
Alston, W. 1971. “Varieties of Privileged Access”. American Philosophical Quarterly 8: 223-241.
Moore, G.E. 1993. “Moore’s Paradox”. Chap. 12 of G.E. Moore: Selected Writings, T. Baldwin, ed. London: Routledge. (from an unfinished manuscript, circa 1944.)
 This list includes the principal interpretations of ‘privileged access’ but it is not exhaustive. To appreciate how much variation this term allows, see Alston (1971).
 Cf. Peacocke, on the task for theories of self-knowledge: “The general challenge in this area is to find anything intermediate between the unexceptionable but uninformative, on the one hand, and the absolutely unbelievable on the other.” (Peacocke, this volume, xx63xx)
 For more on whether self-knowledge is inferential, see Aydede (this volume).