Page 1
INFALLIBILITY OF THE REFLECTIVE BELIEF ABOUT THE CONTENT
OF ONE’S OWN SENSORY EXPERIENCE
VERSION 10.1 (MARCH 19, 2005)
TOMOJI SHOGENJI
ABSTRACT
This paper defends the view that reflective beliefs about the contents of one’s own sensory experience are infallible. The main challenge in defending this view is to explain the relation between one’s own sensory experience and the belief about its content. I propose that their relation is not observation, inference, or conceptual articulation, but semantic ascent similar to the transition from a sentence in the object language to a sentence in the meta-language that affirms the truth of the original sentence. This account allows us to understand why the reflective belief about the content of one’s own sensory experience is infallible.
This paper defends the view that the reflective belief about the content of one’s own (and current—hereafter this is assumed) sensory experience is infallible. The view is a key component of the traditional form of foundationalism in the epistemology of empirical beliefs, and has received some forceful criticisms—both theoretical and empirical—from the opponents of traditional foundationalism.1 However, the view has a strong intuitive appeal independently of the project of traditional foundationalism, so that many Page 2 philosophers with no interest in traditional foundationalism are in favor of it. The recent dispute over the compatibility of self-knowledge and semantic externalism is a good indication of the view’s appeal.2 The suspected incompatibility of self-knowledge and semantic externalism would not have attracted much attention if the participants of the dispute were ready to renounce the claim of self-knowledge, or a privileged access to one’s own mental states. As we will see shortly, the notion of a privileged “access” is not helpful and even misleading, but the underlying view that reflective beliefs about one’s own mental states have a special epistemic status seems sound. An analysis of this special epistemic status should be of interest to those who share this view, beyond the context of traditional foundationalism. My focus in this paper is the epistemic status of the belief about the content of one’s own sensory experience, so that the discussion will be directly relevant to the traditional form of foundationalism, but the assumptions I make about one’s own sensory experience are less controversial about one’s own occurrent belief. Therefore, one can use essentially the same reasoning to show that the reflective belief about the content of one’s own occurrent belief is infallible.3
1. PRELIMINARIES
This section delineates the framework of discussion and establishes terminology. First, I take an experience to be conscious by definition. I sometimes speak redundantly of
1 See DePaul (2001), and BonJour and Sosa (2003) for some recent defenses and criticisms of the traditional form of foundationalism. 2 Many important contributions to this dispute are collected in Ludlow and Martin (1998), and Nuccetelli (2003). I have reservations about semantic externalism with regard to those contents of mental states that are relevant to epistemic deliberation, but I do not take a stance on the dispute in this paper beyond the defense of self-knowledge. 3 There is one problem about self knowledge of occurrent beliefs that does not arise about self-knowledge of sensory experience, which is the determination of the propositional attitude. In the case of sensory experience, it is clear from the phenomenology of the experience that one is having sensory experience, but it may not be as clear in the case of occurrent beliefs whether one is really having a belief as distinguished from suspicion, assumption, etc. toward the propositional content. This problem does not affect the claim that a reflective belief about the content of one’s occurrent belief is infallible, but it casts doubt on the part of the meta-belief that one is having a belief with that content. One way of side-stepping this issue is to regard thinking as generic propositional attitude that comprises beliefs, suspicions, assumptions, etc. (cf. Page 3
“conscious experience” but that is only for emphasis. I do not mean to imply that there are also unconscious experiences. This stipulation is reasonable in the context of epistemology since it is conscious mental states, and not those mental states buried in the unconscious, that are integrated into our inferential network for further epistemic deliberation. Second, I take a sensory experience to have a representational content4 (I often call it simply “content”). To use Chisholm’s “appeared to” locution (Chisholm 1966, Ch. 6), examples of sensory experiences are being appeared to as if it is raining, being appeared to as if a cat is on the mat, etc. These sensory experiences have the representational contents that it is raining, that a cat is on the mat, etc., and are therefore true or false, depending on the way the external world actually is. In short a sensory experience is both conscious and representational. I take this to mean that we are conscious of the content of our own sensory experience in the sense that when we are appeared to as if it is raining, we have the (fallible) awareness that it is raining.5 The next preliminary point concerns the state-content ambiguity of the term “sensory experience”. As in any discussion of representational states, one must guard against an equivocation of the term “sensory experience” between the mental state itself and its representational content. My policy in this paper is to use the term “sensory experience” only in reference to the mental state itself, and not its representational content. For example, when I say that being appeared to as if it is raining is a sensory experience, I am referring to a mental state (presumably realized by some neural state of the brain) whose representational content is that it is raining. The term does not directly refer to the representational content—viz., that it is raining. This means that a sensory experience that is true or false is a mental state itself, and not its representational content. It is similar to saying that a sentence—the bearer of a representational content—and not its representational content, is true or false. According to this terminology, what is often called a belief about one’s own sensory experience is actually a belief about the representational content of one’s own sensory experience. For example, the belief that I
Burge 1996, p. 240), and contend that reflective beliefs about the content of one’s own occurrent thinking are infallible. 4 See Harman (1990) for a clear exposition and defense of this position. 5 Laurence BonJour (2001) speaks of awareness of content that is non-apperceptive (i.e., not a higher-order awareness that one has a mental state with a certain content) and yet infallible. It is unclear to me what it Page 4
am appeared to as if it is raining is the belief that the sensory experience I am having has the representational content that it is raining. Some people may find this locution cumbersome but I follow this terminological policy to avoid confusion that sometimes arises from the state-content ambiguity in the discussion of representational states. For the sake of brevity I will call the representational content of a sensory experience “appearance” and the belief that one’s own sensory experience has a certain representational content “appearance belief”. For example, when I am appeared to as if it is raining, that it is raining is an appearance, and the belief that I am appeared to as if it is raining is an appearance belief. I will call the relation between a sensory experience and the appearance belief about its content that makes the latter infallible “reflection” and an appearance belief formed from a sensory experience by the relation of reflection “reflective appearance belief.” The thesis of this paper, expressed succinctly, is therefore: Reflective appearance beliefs are infallible. The main task of the paper is to analyze the relation of reflection and explain how it makes reflective appearance beliefs infallible.6 One controversy over sensory experience deserves special attention. I have used as examples of appearances (contents of sensory experiences) that it is raining and that a cat is on the mat. Some people may object to these examples since these appearances are conceptually articulated,7 while according to some philosophers (e.g., Evans 1982), the content of a sensory experience is not conceptually articulated. Only a belief has a conceptually articulated content. I do not accept this view, but it is not my position either that all sensory experiences have conceptually articulated contents. I only defend the moderate form of conceptualism about sensory experience that some sensory experiences have conceptually articulated contents, and further even when the content of a sensory experience is conceptually articulated, it is usually partially conceptually articulated. For example, I may be appeared to partially conceptually as if it is raining, while further
means to have a non-apperceptive awareness of the content, e.g., that it is raining, other than having the awareness that it is raining, which is obviously fallible—i.e., it may not be raining in the external world. 6 It is not my claim that all appearance beliefs are infallible. In Section 5 I will describe some cases of fallible appearance beliefs that are not formed by reflection. 7 They are moreover propositionally articulated so that they are apt for alethic evaluation. The content of a sensory experience can be conceptually articulated without being propositionally articulated—e.g., I am appeared to redly—but my focus will be those with propositional articulation, which are of more interest in the context of foundationalist epistemology. Page 5
details of the way it is raining may constitute part of the appearance that receives no conceptual articulation. This moderate form of conceptualism is much more defensible than all-out conceptualism—i.e., the view that the contents of all sensory experiences are totally conceptually articulated. The standard objections to conceptualism about sensory experience only threaten all-out conceptualism, and do not affect moderate conceptualism. For example, it is difficult to deny phenomenologically that some sensory experiences have no conceptually articulated contents. Moderate conceptualism can easily accommodate this point since it does not deny the existence of such sensory experiences. It is also pointed out by the opponents of conceptualism that the content of a sensory experience is too fine-grained to be captured conceptually (Evans 1982, p. 229). There are, for example, more shades of blue we can distinguish visually than we can articulate in words. Moderate conceptualism can accommodate this point as well since it does not claim that the content of a sensory experience is totally conceptually articulated. Non-conceptualists, such as Evans, go all the way to the other extreme, claiming that only the contents of beliefs are conceptually articulated. But in my view having a sensory experience with a conceptually articulated content does not amount to forming a belief. I can be appeared to as if it is raining with a clear conceptual articulation but without believing that it is raining. This is most obvious in the case of an informed illusion. For example, if I have been informed that I am in an illusion-inducing experimental setting, I may not believe that it is raining even if I am appeared to as if it is raining. A belief carries a commitment to its truth, while a sensory experience qua sensory experience does not, even when its content is conceptually articulated. This is why it is absurd to say, “I believe it is raining, but it is not” (cf. Moore 1944, p. 204), but it is not absurd to say, “I am appeared to as if it is raining, but it is not.” We cannot hold a belief without a commitment to its truth, but we can have a sensory experience with no commitment to its truth. It may be suggested here that although the mental state in the informed illusion case is not a belief, it is not a sensory experience, either. It is, say, a conceptual experience, which must be distinguished from a sensory experience. That is to say, a sensory experience itself never has a conceptually articulated content—its non- Page 6
conceptual content provides raw materials for conceptual articulation. The two-layer view is tempting because it seems the same sensory experience can be conceptually articulated in different ways, but the view is actually either false or misleading. To see why, suppose I am originally appeared to in a conceptually unarticulated way, and then the content of the sensory experience becomes conceptually articulated—e.g., I become appeared to as if it is raining. This means that I now have a conceptual experience, according to the two-layer view. The difficult question for the advocate of this view is whether I still have a sensory experience, which must have a conceptually unarticulated content, according to the two layer view. If the answer is affirmative, the view is false because once I become appeared to as if it is raining with clear conceptual articulation, I no longer have the experience of being appeared to in a conceptually unarticulated way.8 The original experience is replaced by an experience with a conceptually articulated content. The negative answer that I no longer have a sensory experience is misleading because even after some aspects of its content are conceptually articulated, I still have a conscious mental state with the fine-grained phenomenology that is characteristic of a sensory experience. It is better to regard conceptual experience as a species of sensory experience, acknowledging that some sensory experiences have (partly) conceptually articulated content while others do not—and that is the position of moderate conceptualism about sensory experience.
2. OBSERVATION, INFERENCE, AND CONCEPTUAL ARTICULATION
We are now ready to address the main topic of this paper, namely the relation of reflection between a sensory experience and an appearance belief about its content that makes the latter infallible. This section examines three initially attractive views of the
8 The representational content is different in this regard from the external object of perception. The external object of perception retains its identity regardless of the way we perceive it, but the existence and characteristics of the representational content are dependent on the representation. Where three is no representation, there is no representational content. When the representation becomes conceptual, there is no longer non-conceptual representational content. Unlike the external object of perception the representational content does not survive a change in representation. Page 7
relation as (1) observation, (2) inference, and (3) conceptual articulation, and explains why they all fail. I will propose my own account in the next section.
2.1 OBSERVATION
Let us begin with observation as a candidate for the relation of reflection. According to this view, we form an appearance belief by observing our own mental state. For example, when I have the sensory experience of being appeared to as if it is raining, I observe this sensory experience and form the belief that I am appeared to as if it is raining. An obvious problem with this view is that it does not explain the infallibility of the appearance belief, for any observation is potentially mistaken. It is sometimes suggested that we have a privileged access to our own mental states, but that is not a real explanation. It only gives a name to the mysterious way the observation of one’s own mental state is infallible. In the absence of a real explanation of infallibility, the observation view is unattractive to the proponents of the infallibility thesis. Of course, the absence of a real explanation of infallibility is no reason to reject the observation view if one does not wish to defend the claim of infallibility. There is, however, a more fundamental problem with the observation view—namely, it equivocates the term “sensory experience” between the mental state itself and its representational content. In my terminology an appearance belief is not directly a belief about a sensory experience (which is a mental state itself) but a belief about the representational content of a sensory experience. This is an important distinction, for the observation of a sensory experience (the mental state itself) does not reveal its representational content in any immediate way. Observing a sensory experience that has the representational content that it is raining is not the same, obviously, as observing that it is raining. The observation of the sensory experience itself would reveal, presumably, some properties of the brain such as spiking frequencies in some neural pathways. But perceiving those properties does not allow us to recognize in any immediate way the Page 8
representational content of the sensory experience, or what the sensory experience represents—viz., that it is raining.9 Some people have suggested that what we observe in reflection is not the sensory experience itself, but the representational content of the sensory experience (Langsan 2002). It is hard, however, to make sense of this claim. We can observe concrete entities such as the rain falling from the sky or the mental state that has the representational content that it is raining, but we cannot observe in any ordinary sense the representational content that it is raining. In order to do so, we would have to “observe” an abstract entity such as a proposition. There is, of course, the expression “observe that it is raining,” but it only means that the observation leads to the belief that it is raining. It does not mean that one literally observes the abstract entity that it is raining. I conclude that the relation of reflection between a sensory experience and the appearance belief about its content that makes the latter infallible is not observation.
2.2 INFERENCE
The second candidate for the relation of reflection between a sensory experience and an appearance belief about its content is inference. The idea is that an appearance belief about the content of a sensory experience is formed by an inference from the sensory experience. In order to defend this view, we need to assume that the sensory experience has a conceptually articulated content since it is impossible to make an inference from a mental state with no conceptually articulated content.10 As explained in Section 1, I have no objection to this assumption. Moderate conceptualism about sensory experience acknowledges that some appearances are (partially) conceptually articulated. But as with the observation view, an obvious problem with this view is its failure to explain the infallibility of appearance beliefs. Sensory experiences are by no means infallible, and
9 This point is made by Dretske (1993) and Güzeldere (1995) against the higher-order representation theory (HOR) of consciousness. I will discuss HOR in Section 4. 10 Some foundationalists (e.g., Fales 1996, Ch. 6) speak of “inference” from a non-conceptual content to a conceptual content, but what they call “inference” is more appropriately characterized as conceptual articulation of the content, or a combination of conceptual articulation and inference proper. I will examine below the view that the relation of reflection is conceptual articulation. Page 9
thus appearance beliefs cannot be infallible, either, if they are derived by inference from the sensory experiences. This is, of course, not a problem for people with no interest in the claim of infallibility. There is, however, a more fundamental problem with the inference view that is decisive against it even if one has no intention to defend the claim of infallibility. Namely, the truth and falsity of an appearance belief—e.g., whether or not I am appeared to as if it is raining—does not depend on the truth and falsity of the sensory experience it is supposed to be inferred from—e.g., whether or not it is raining.11 This is incomprehensible if their relation is inference. To see the strangeness of the view, suppose I am appeared to as if it is raining but it is not actually raining, and hence my sensory experience falsely represents the external world. If the relation of reflection between a sensory experience and an appearance belief about its content were inference, the falsity of the sensory experience should negatively affect the appearance belief inferred from it, that I am appeared to as if it is raining—but it does not. Regardless of the weather in the external world, my appearance belief that I am appeared to as if it is raining is true as long as I am appeared to as if it is raining. A sensory experience represents the external world, and hence one can derive from it by inference a (fallible) perceptual belief that represents the external world. One cannot derive from it by inference an appearance belief whose truth and falsity has no bearing on the way the external world actually is. I conclude that the inference view is untenable.
2.3 CONCEPTUAL ARTICULATION
The third candidate for the relation of reflection is conceptual articulation. According to this view, we form an appearance belief by conceptually articulating the content of our sensory experience. Obviously, the view requires that the sensory experience in question originally have a conceptually unarticulated content. That is not a problem. Moderate conceptualism about sensory experience only claims that some appearances are (partially) conceptually articulated. The problem, however, is that conceptual articulation does not turn a sensory experience into an appearance belief. Suppose I am originally appeared to
11 This point has been made by Dretske (1995, Ch. 2) and Brueckner (1998). Page 10
in a conceptually unarticulated way, and then this appearance becomes conceptually articulated. What I obtain by this process is still a sensory experience—namely, a sensory experience with a conceptually articulated content, such as being appeared to as if it is raining. This is not an appearance belief. The underlying problem of the view is essentially the same as the one we found in the inference view. To recall, if we were forming an appearance belief by making an inference from a sensory experience that represents the external world, the truth and falsity of the sensory experience should be relevant to the truth and falsity of the appearance belief inferred from it. But the truth and falsity of an appearance belief has no bearing on the way the external world actually is, which determines the truth and falsity of the sensory experience. Similarly, if we were forming an appearance belief by conceptually articulating the content of a sensory experience that represents the external world, the truth and falsity of the formed belief should at least partly depend on the way the external world actually is. But the truth and falsity of an appearance belief has no bearing on the way the external world actually is. To conclude, conceptual articulation is not the relation of reflection between a sensory experience and the appearance belief about its content. Some people may think of combining conceptual articulation with either observation or inference to explain the relation of reflection, but any such attempt is doomed. This is because conceptual articulation generates a sensory experience with a conceptually articulated content, and it has been shown already that neither inference nor observation turns a sensory experience with a conceptually articulated content into an appearance belief about its content. Consequently, combining conceptual articulation with either observation or inference does not turn a sensory experience with a conceptually unarticulated content into an appearance belief about its content. We must look elsewhere for the relation of reflection between a sensory experience and an appearance belief about its content that makes the latter infallible.
3. REFLECTIVE SEMANTIC ASCENT Page 11
Let us take stock to see more clearly the problem we are facing. It is assumed in this paper that a sensory experience is a conscious mental state with a representational content, and that some contents are (partially) conceptually articulated—e.g., I may be appeared to as if a cat is on the mat. When the content is conceptually articulated in such a way, it seems quite obvious—even trivial—that on the basis of the sensory experience I can form the true appearance belief that I am appeared to as if a cat is on the mat since I have the conscious sensory experience that a cat is on the mat. Still none of the three initially attractive accounts of the relation of reflection between a sensory experience and an appearance belief about its content has turned out to be tenable. One reason for their failure, I propose, is that these accounts all rely on substantial relations—observation, inference, or conceptual articulation—to explain the transition from a sensory experience to an appearance belief about its content. In my view the relation is much less substantial, which is appropriate for understanding the (intuitively) trivial way appearance beliefs are true. My proposal is that an infallible appearance belief is formed by semantic ascent from a sensory experience, where semantic ascent is understood in the following way.
Semantic ascent is a transition from an n-th order representational entity to an n- plus-first order representational entity that ascribes a semantic property to the original n-th order representational entity based solely on the content of the original n-th order representational entity without any new empirical input.
We focus here on the transition from a first-order representational entity to a second order representational entity. One familiar type of semantic ascent is the transition from a sentence in the object language to a sentence in the meta-language that affirms the truth of the original sentence—for example, from ‘Snow is white’ to ‘‘Snow is white’ is true’. It is a transition from a first-order representational entity to the second-order representational entity, and the second-order representational entity ascribes the semantic property of truth to the original first-order representational entity. No new empirical input is needed for the transition. In a similar way the relation of reflection allows a transition Page 12
from a first-order representational entity (a sensory experience) to a second-order representational entity (an appearance belief), and the second-order representational entity ascribes to the first-order representational entity the semantic property of having a certain representational content. No new empirical input is needed for the transition. Comparison to the ascent to the truth claim gives us a first grip on the formation of an appearance belief from a sensory experience by the relation of reflection since both of these transitions are species of semantic ascent. There are, however, some obvious differences between the two cases. First, there is a difference between linguistic semantic ascent and mental semantic ascent. Linguistic semantic ascent is a transition from a sentence in the object language to a sentence in the meta-language, while a mental semantic ascent is a transition from a first-order mental state to a second-order mental state. However, this difference does not affect the structure of the ascent. We can easily construct a mental counterpart of linguistic semantic ascent. For example, instead of semantically ascending from the sentence ‘Snow is white’ in the object language to the sentence ‘‘Snow is white’ is true’ in the meta-language, we can semantically ascend from the first-order belief that snow is white to the second-order belief that the belief that snow is white is true. For closer structural parallelism, one can compare the linguistic semantic ascent from the sentence ‘Snow is white’ to the sentence ‘The sentence whose content is that snow is white is true’, and the mental semantic ascent from the belief that snow is white to the belief that the belief whose content is that snow is white is true. The standard form of linguistic semantic ascent with quotation marks is made simpler by the tacit assumption that the object language and the meta-language are homophonic—i.e., the content of the sentence ‘Snow is white’ is that snow is white. A more substantive difference between the truth claim and the appearance belief is the semantic properties ascribed to the original first-order representational entities. The sentence ‘‘Snow is white’ is true’ ascribes the semantic property of truth to the original first-order sentence, while the appearance belief that I am appeared to as if it is raining ascribes the semantic property of having a certain representational content to the original first-order mental state. I will call semantic ascent of the former type alethic semantic ascent while calling the latter reflective semantic ascent.
Page 13
Alethic semantic ascent is a transition from an n-th order representational entity to an n-plus-first order representational entity that ascribes the semantic property of truth to the original n-th order representational entity based solely on the content of the original n-th order representational entity without any new empirical input.
Reflective semantic ascent is a transition from an n-th order representational entity to the n-plus-first order representational entity that ascribes the semantic property of having a certain representational content to the original n-th order representational entity based solely on the content of the original n-th order representational entity without any new empirical input.
Alethic and reflective semantic ascents are both species of semantic ascent but with significant differences in the resulting second-order representational entities. To begin with, the second-order representational entity that results from alethic semantic ascent— such as the sentence ‘‘Snow is white’ is true’ or the meta-belief that the belief that snow is white is true—is fallible while an appearance belief—such as the belief that I am appeared to as if it is raining—that results from reflective semantic ascent is infallible, or so I contend. A related difference is that alethic semantic ascent can still be considered a form of inference while reflective semantic ascent is not an inference of any kind. Alethic semantic ascent preserves truth—e.g., if ‘Snow is white’ is true, then ‘‘Snow is white’ is true’ is also true. In fact there is no difference in substance between the contents of the first- and second-order sentences. If the original sentence is about the external world, the sentence that affirms its truth is also indirectly about the external world. The sentence that affirms the truth of the original sentence is fallible simply because the original sentence is fallible. An appearance belief formed by reflective semantic ascent, on the other hand, is infallible, and this is not because the original sensory experience is infallible. The original sensory experience, such as being appeared to as if it is raining, represents the external world and is fallible, but the appearance belief obtained from it by reflective semantic ascent is still infallible. As mentioned earlier, this would be Page 14
incomprehensible if the transition were a form of inference. There is no inference whose contingent conclusion is infallible no matter whether its key premise is true or false. Reflective semantic ascent that constitutes the relation of reflection is therefore not an inference. Here is then our puzzle. The content of an infallible appearance belief is obviously different in substance from the content of a fallible sensory experience. One can nevertheless turn a sensory experience into an appearance belief by reflective semantic ascent without any new empirical input. The reason, I submit, is that when I form an appearance belief, I only re-characterize what the external world appears like to me as have the sensory experience, as the representational content of my sensory experience without making any change in substance.12 Let me elaborate the point. First, when I have a sensory experience, the external world appears to me in a certain way. For example, when I have the sensory experience of being appeared to as if it is raining, I have the (fallible) awareness that it is raining. But what the external world appears like to me—that it is raining—is by the very fact that it is the way the world appears to me the content of my sensory experience. By re-characterizing what the external world appears like to me—namely that it is raining—as the representational content of my sensory experience, I obtain the true appearance belief about the content of my sensory experience. The key in this procedure is that what the external world appears like to me as I have the sensory experience is the content of the sensory experience, and thus I can simply ascribe what the external world appears like to me to my sensory experience as its representational content. There is therefore no need for new observation, inference, or conceptual articulation in the process of forming an appearance belief. We can now see why an appearance belief formed by reflective semantic ascent from a sensory experience is infallible. If I am appeared to as if p, where p is what the external world appears like to me, then by the very fact that it is the way the world
12 Some proposals have been made in the same direction. Tyler Burge (1996), for example, suggests that in certain cases of self-knowledge the intentional (representational) content of one’s own mental state is “thought and thought about in the same act” (p. 244), but Burge applies this account to a very limited class of self-knowledge, which he calls cogito-like cases. (See below for more on cogito-like cases.) With regard to a broader class of self-knowledge Burge speaks of “a constitutive relation […] between the judgments about one’s thoughts and the judgments’ being true” (p. 245), but he does not spell out the nature of this constitutive relation. Other authors with similar views, such as Dretske (1995, Ch. 2) on “displaced Page 15
appears to me, that p is the representational content of the sensory experience. Thus, if I form the appearance belief that I am appeared to as if p by adding the mental prefix I am appeared to as if to what the external world appears like to me, namely p,13 the resulting belief that I am appeared to as if p is trivially true. The procedure is similar to adding to the sentence ‘Snow is white’ quotation marks and the suffix ‘is true’ to form the sentence ‘‘Snow is white’ is true’ in the meta-language. In the case of alethic semantic ascent, the resulting sentence is trivially equivalent to the original sentence. In the case of reflective semantic ascent the resulting appearance belief is trivially true. Fastidious epistemologists may worry about a small temporal gap between the occurrence of the sensory experience and the formation of the appearance belief—by the time I complete the procedure of reflective semantic ascent, the sensory experience may no longer have the same content, which makes the resulting appearance belief fallible. But this worry is based on a misunderstanding of reflective semantic ascent. The mental prefix I am appeared to as if is added to what the world appears like to me as I am having the sensory experience. In fact I can form the partial thought I am appeared to as if first, and then add whatever the world appears like to me at the moment to complete the appearance belief. I should note that this understanding of a reflective appearance belief does not reduce reflective semantic ascent to what Tyler Burge (1996) calls a cogito-like case, where one makes an ascription of the content true by the very act of ascription. For example, when I judge that I am thinking that there are physical objects, I make this judgment true by the very act of judging that I am thinking that there are physical objects, for I cannot make this judgment without entertaining the thought that there are physical objects. Note here that in a cogito-like case no matter what content I ascribe to my thought, it comes out true by the very act of ascription. This is because judgment in a cogito-like case is an execution that makes the judgment true, and not merely a description. In this sense a cogito-like case is very similar to the speech act of assertion. The execution of the assertion that p makes it true that I assert that p. In contrast an
perception” and Peacocke (1996) on “conceptual redeployment” also provide a partial account of the relation of reflection. 13 In a similar spirit, Papineau (2002, Ch.4) proposes that the concept of experience “the experience:---” is formed by adding the experience operator “the experience:” to perceptual classification “---”. Page 16
appearance belief is purely descriptive—it describes the content of the sensory experience. Unlike a cogito-like case or a speech act of assertion, forming an appearance belief does not automatically make the appearance belief true. Indeed, we will see in Section 5 below that an appearance belief can be false if it is not formed by reflective semantic ascent.
4. HIGHER ORDER REPRESENTATION THEORY
I have explained the infallibility of reflective appearance beliefs by the notion of reflective semantic ascent from a first-order representational entity (a sensory experience) to a second-order representational entity (an appearance belief). By regarding a conscious sensory experience as a first-order representational entity, my analysis so far assumes the first-order representational theory of consciousness (FOR), viz. the view that a first-order mental state can be a conscious mental state without being itself represented by a higher- order mental state.14 My analysis may be questioned for this reason by those in favor of the higher-order representational theory of consciousness (HOR), viz. the view that having a conscious mental state requires that one have a higher-order mental state that represents a mental state that is itself representational.15 Further, by assigning conceptual contents to some sensory experiences, my analysis so far assumes what might be called the first-order representational theory of conceptual contents (FORc), viz. the view that a sensory experience can have a conceptually articulated content without being represented by a second-order mental state. This may be challenged by those who think that the content of a first-order sensory experience is non-conceptual unless we form an appearance belief about it (BonJour 2001).16 I call the view the higher-order
14 FOR advocates include BonJour (2001), Dretske (1993), and Güzeldere (1995). 15 HOR advocates include Armstrong (1968), Lycan (1996), and Rosenthal (1993). 16 My analysis of the infallibility of an appearance belief holds even if the sensory experience has a conceptually unarticulated content. For example, if I have the sensory experience of being appeared to in a conceptually unarticulated way, then reflective semantic ascent generates the appearance belief that I am appeared to in a conceptually unarticulated way, or simply that I am appeared to thusly. If my analysis in this section is correct, this appearance belief is infallible, though the result is not very interesting since an Page 17
representational theory of conceptual contents (HORc). In this section I want to show that my analysis is compatible not only with FOR and FORc but also with plausible versions of HOR and HORc. We need a brief overview of reasons for holding the different views on the subject. HOR’s major appeal comes from phenomenology. It is our common experience that an automatic pre-conscious activity (such as operating a car in a familiar setting) becomes a conscious activity when we start paying attention to the way our mind responds to the changes in the environment. Conversely the process of registering part of the environment in our consciousness is often accompanied by a reflection of our own mental state. HOR advocates take these observations to indicate that having a conscious experience requires a formation of a higher-order mental state. But the opponents of HOR ask a difficult theoretical question—namely, whether the content of the purported higher- order mental state is the same as the content of the first-order mental state. For example, when we have the conscious sensory experience of being appeared to as if it is raining, does the purported higher-order mental state about the sensory experience also have the content that it is raining? If it does, then it is not really a higher-order mental state about the first-order mental state, the HOR opponents point out. But if it doesn’t, then why is the content of our conscious sensory experience that it is raining, instead of (say) some neural conditions of the brain, which the higher-order mental state presumably represents? Representing the first-order mental sate does not allow us to comprehend its content in any immediate way. The impression that it does arises from an insufficient attention to the state-content ambiguity of representational terms. The effectiveness of this theoretical challenge to HOR depends on the way we interpret HOR. The challenge makes it hard for the HOR advocate to maintain that the higher-order mental state itself is the conscious mental state since one would then have to claim that the first- and second-order mental states have the same content. But the HOR advocate may only claim that having a conscious experience requires a formation of a higher-order mental state, as the phenomenology indicates, without insisting that the higher-order mental state itself is the conscious mental state. In other words, the HOR
appearance belief of this kind is of no use in the justification of further beliefs that have conceptually articulated contents. Page 18
advocate may concede that the conscious mental state itself is a first-order mental state, but it is a conscious mental state only when a higher-order mental state represents it. For example, my mental state of being appeared to as if it is raining becomes a conscious sensory experience, while still remaining a first-order mental state, only when this first- order mental state is represented by the appearance belief that I am appeared to as if it is raining. In other words, a conscious mental state is a first-order mental state that is accompanied by a higher-order mental state that represents it. This view accommodates the apparent phenomenology that makes HOR attractive without making HOR vulnerable to the theoretical challenge from its opponents. I call this version of HOR “plausible HOR”. We can also distinguish two versions of the higher-order representational theory of conceptual contents (HORc). According to one version, only the content of an appearance belief is conceptually articulated, while the content of a sensory experience itself is never conceptually articulated.17 The problem with this version (in fact any view that denies conceptual articulation to the contents of all sensory experiences) is its implication that no appearance belief that ascribes a conceptually articulated content to a sensory experience can be literally true, let alone infallible. It is incoherent to claim that an appearance belief that ascribes a conceptually articulated content to a sensory experience is literally true while claiming at the same time that no sensory experience has a conceptually articulated content. If I can only be appeared to in a conceptually unarticulated way, then the belief that I am appeared to as if it is raining must be false. It may be suggested here that the appearance belief is true in the sense that I am appeared to in such a way that if the content of the sensory experience were conceptually articulated, it would be that it is raining. This explanation is unsatisfactory because the same non- conceptual content may receive different conceptually articulations and some of them may even be mutually inconsistent, e.g., if the sensory experience has an ambiguous content such as the duck-rabbit figure. It is arbitrary to pick one of the potential conceptual articulations and claim that it makes the appearance belief true (or false). It is also hard to maintain, phenomenologically, that we retain a conscious first-order sensory experience with a conceptually unarticulated content while we already have the
17 This seems to be BonJour’s (2001) view. Page 19 appearance beliefs that assigns a conceptually articulated content to it such as being appeared to as if it is raining. A more plausible version of HORc is that the first-order sensory experience itself can have a conceptually articulated content, but it does so only when it is accompanied by an appearance belief about its content. In other words, having a conceptually articulated content still requires the existence of a higher-order representation, but it is still the first-order sensory experience that acquires a conceptually articulated content. This version of HORc has no implication that any appearance belief that assigns a conceptually articulated content to a sensory experience is false. I call this version of HORc “plausible HORc”. I now want to show that my analysis of the infallibility of reflective appearance beliefs by reflective semantic ascent is compatible with plausible HOR and plausible HORc. First, if plausible HOR is correct, we have a conscious sensory experience only when we form an appearance belief about its content. This means that prior to the formation of an appearance belief there is no conscious sensory experience to turn into an appearance belief. With no pre-existing conscious sensory experience, I start thinking I am appeared to as if. This thought fragment I am appeared to as if turns my originally unconscious sensory state into a conscious sensory experience. As a result, the external world comes to appear in a certain way to my consciousness. I then complete the formation of an appearance belief by combining the thought fragment I am appeared to as if with whatever the external world now appears like to me. This operation constitutes reflective semantic ascent, and turns the emerging conscious sensory experience into an appearance belief about its content. The completed appearance belief is infallible because what the external world appears like to me (as a result of forming an appearance belief) is the content of my sensory experience. Similarly, if plausible HORc is correct, a sensory experience has a conceptually articulated content only when we form an appearance belief about its content. This means that when I start thinking I am appeared to as if, the content of my sensory experience becomes conceptually articulated, and the external world comes to appear in a conceptually articulated way. At that point I can complete the formation of an appearance belief by combining the thought fragment I am appeared to as if with what the world now conceptually appears like to me. This operation constitutes reflective semantic ascent, Page 20 and turns the sensory experience with the emerging conceptually articulated content into an appearance belief about its content. The completed appearance belief is infallible because what the external world conceptually appears like to me (as a result of forming an appearance belief) is the content of my sensory experience. I want to stress that even plausible HOR and plausible HORc do not reduce reflective semantic ascent to what Tyler Burge calls a cogito-like case, where one makes an ascription of the content true by the very act of ascription. An appearance belief under plausible HOR and plausible HORc is still purely descriptive, and not an execution that makes the content ascription true. It is very different in this regard from the speech act of assertion. I may start uttering the sentence fragment “I hereby assert that” and then follow it up with an utterance of some sentential clause, such as “it is raining”, under the conditions appropriate for the speech act of assertion. When the whole utterance is produced, the very act of uttering “I hereby assert that it is raining” makes it true that I assert that it is raining. In fact I can combine the sentence fragment “I hereby assert that” with any sentential clause to make the assertion with that content true. There is no way that I utter, say, “I hereby assert that it is snowing in Paris” under the conditions appropriate for making an assertion, and yet I do not make the assertion that it is snowing in Paris. In contrast, in order to form an infallible appearance belief by reflective semantic ascent, I have to combine the thought fragment I am appeared to as if with the emerging content of my conscious sensory experience, and not with any content of choice. Forming the appearance belief that I am appeared to as if five cats are on the mat does not make it true that I am appeared to as if five cats are on the mat. Indeed, as we will see in the next section, an appearance belief can be false if we do not form it by reflective semantic ascent.
5. FALLIBLE APPEARANCE BELIEFS
This section addresses likely objections to the claim of infallibility for reflective appearance beliefs. I want to begin with a potential counterexample: I see some cats on Page 21
the mat and come to believe, perhaps hastily, that I am appeared to as if five cats are on the mat, while in fact I am appeared to as if six cats are on the mat.18 The issue is not the number of real cats on the mat but the number of cats as represented by my sensory experience. Can an appearance belief be mistaken in this way with regard to the content of one’s own sensory experience? To answer this question properly, we need to distinguish two kinds of cases. First, the appearance belief may not be formed by reflective semantic ascent from the sensory experience. For example, I may have the belief already that five cats are on the mat because someone has told me so. When I see the cats, I may make an inference, without paying much attention to what I see or even overruling what I see, that I must be appeared to as if five cats are on the mat since there are five cats, I have been told, on the mat. In some cases false ascription of the content may not involve any inference. For example, even if I am only appeared to as if some cats are on the mat with no further conceptual articulation as to the number of the represented cats, but I may still form the appearance belief that I am appeared to as if five cats are on the mat by arbitrarily filling in the conceptual detail without consulting the appearance, or what the world appears like to me as I have the sensory experience. In such cases I grant that the resulting appearance belief is fallible. Note, however, that my claim is that a reflective appearance belief—an appearance belief formed by adding the mental prefix I am appeared to as if to what the world appears like to me as I have the sensory experience—is infallible. I do not claim that an appearance belief formed in some other way is infallible.19 With regard to the appearance belief formed by reflective semantic ascent from the sensory experience, there is no possibility, I maintain, that it is false. It may still seem that I can hastily form a reflective appearance believe that I am appeared to as if five cats are on the mat, while I am actually appeared to as if six cats are on the mat. But that is a logical impossibility. A reflective appearance belief is formed by adding the mental
18 The example is a variation of the cluster of cases—such as speckles on a hen and stripes on a tiger—used in the dispute over the infallibility of the observation of sense data. (Ayer 1940, p. 124; Price 1941; Chisholm 1942) I made the number of cats in the example fairly small to make it more likely that the number is part of the conceptually articulated content of the sensory experience. 19 As mentioned earlier, this fallibility of an appearance belief makes it clear that we are not dealing here with what Tyler Burge (1996) calls a cogito-like case in which one makes an ascription of the content true by the very act of ascription, no matter what content one ascribes. Reflective semantic ascent, in contrast, Page 22
prefix I am appeared to as if to what the external world appears like to me as I have the sensory experience. If this operation generates the belief that I am appeared to as if five cats are on the mat, then that five cats are on the mat is the way the external world appears to me as I have the sensory experience. The appearance belief then cannot be mistaken because that five cats on the mat is indeed the content the appearance belief ascribes to the sensory experience. Some critics familiar with the recent psychological literature on false self-reports may question a seemingly innocent step in my reasoning—namely, what the external world appears like to me as I have a sensory experience is the representational content of the sensory experience. In some experiments on false self-reports an unsuspecting subject is presented with a figure that induces a perceptual illusion.20 The subject reports that she perceives the object in a certain way, which is different—as expected—from the true figure. However, her visually guided motor behavior is inconsistent with the report. The subject behaves as if she perceived the figure correctly. There is no reason to doubt that her verbal report is based on what the figure appears like to her as she has the sensory experience. It may seem in this case that what the figure appears like to her as she has the sensory experience is not truly the representational content of her sensory experience, which is revealed by her visually guided motor behavior. This is an intriguing finding, but when interpreted properly, it does not affect the claim that what the external world appears like to the subject as she has a sensory experience is the representational content of the sensory experience. Recall that by stipulation the term “experience” in this paper refers to a conscious mental state. This means that in our terminology the subject in the experiment is indeed appeared to in the way she verbally reports. It is irrelevant to the determination of the content of her conscious experience that she “sees” the object differently at some subconscious level. I grant that ascribing to the subject a subconscious mental state whose content is different from what she verbally reports helps explain the subject’s non-verbal behavior better, but that does not affect the view I defend. What mental state we assign to the subject depends
demands that one must assign to the sensory experience what the world appears like to me as I have the sensory experience. 20 See Aglioti et al. (1995), and Milner and Goodale (1995, Ch.6). See also Clark (2001) on the philosophical implications of these experimental findings. Page 23
on the nature and purpose of the inquiry. If we want to explain and predict the subject’s visually guided motor behavior in the experiment, we had better set aside her verbal report and focus on the subconscious mental state. However, that is not appropriate for the purpose of epistemology since it is conscious mental states, and not those buried in the subconscious, that are integrated into the subject’s inferential network for her further epistemic deliberation. There is therefore no good reason, for the purpose of epistemology, to think that the subject in the experiment reports the content of her sensory experience falsely. This completes my defense of the view that reflective appearance beliefs are infallible. There is, however, one issue that needs to be addressed. I have acknowledged that an appearance belief that is not formed by reflective semantic ascent from a sensory experience is fallible. I only claim that a reflective appearance belief—an appearance belief formed by reflective semantic ascent from a sensory experience—is infallible. This qualification gives rise to the question: Can we identify a reflective appearance belief without an error? The claim of infallibility is epistemologically less interesting if our identification of an infallible belief is grossly fallible, and it is hard to deny that in some moments of mental lapse we may mistakenly think that a certain appearance belief is formed by reflective semantic ascent from a sensory experience, while it is actually not. My position on this issue is that the identification of an infallible belief is indeed fallible, but an error in the identification is never empirical. To explain this position, it is helpful to compare the issue to the identification of a logical truth. A logical truth is infallible but our identification of a logical truth is not— we sometimes think some claim is logically true, while it is actually not. There are, of course, formal procedures we can follow to confirm the claim’s logical status, such as the truth tree method. But mistakes still arise in the application of the procedure. The identification of a logical truth is therefore fallible even when we employ a formal procedure to confirm the claim’s logical status. I think the same is true about the identification of a reflective appearance belief. As in the case of a logical truth, there is a procedure I can follow to ensure that the belief I have is a reflective appearance belief— namely, to add the mental prefix I am appeared to as if to whatever the external world appears like to me as I have a sensory experience. But just as an application of a formal Page 24
procedure in logic can go wrong, there can be conceptual confusion in the process of reflective semantic ascent—e.g., about the mental prefix I am appeared to as if—and I may make an error in the mental action. The error may result in an incorrect identification of a reflective appearance belief. The possibility of an incorrect identification may be disappointing to some traditional foundationalists, but the result we have is actually the best we can hope for in the epistemology of empirical beliefs. An error in the application of reflective semantic ascent is not an error of the empirical kind—i.e., not an error due to misleading or insufficient empirical evidence—but an error of the conceptual kind. This is because reflective semantic ascent is a purely conceptual operation whose success does not hinge on the correctness of any empirical evidence. In other words, the identification of an infallible appearance belief by the procedure of reflective semantic ascent is empirically infallible. Since there is no way, as far as I can see, to preclude the possibility of a mental slip in conceptual operation in any identification of a true belief, the best we can hope for in the epistemology of empirical beliefs is to eliminate all possibilities of an empirical error. The identification of an infallible appearance belief by the procedure of reflective semantic ascent achieves that.
6. CONCLUSION
In this paper I defended the view that the reflective appearance belief about the content of one’s own sensory experience is infallible. The explanation is that the relation of reflection between a sensory experience and an appearance belief about its content that makes the latter infallible is not observation, inference, or conceptual articulation, but reflective semantic ascent—i.e., a transition from a sensory experience to an appearance belief by an addition of the mental prefix I am appeared to as if to what the world appears like to me as I have the sensory experience. The ascent makes the resulting appearance belief infallible. Page 25
For those who are interested in the project of traditional foundationalism, I want to note that the sensory experience that is transformed into an appearance belief by reflective semantic ascent plays the role of an unjustified justifier. The notion of an unjustified justifier is often associated with contextualism in epistemology, but in my account it is not the context but reflective semantic ascent that makes a sensory experience an unjustified justifier. The sensory experience of being appeared to as if p is not epistemically justified at the foundational level—i.e., at the outset of the foundational project of building epistemic justification, there is no good reason to think that p is actually the case in the external world. Nevertheless, its representational content, p, or what the external world appears like to me as I have the sensory experience, makes it certain that the reflective appearance belief that I am appeared to as if p is true. It is this unjustified justifier in reflective semantic ascent that stops the regress of justification in the epistemology of empirical beliefs.
REFERENCE
Aglioti, S., DeSouza, J. and Goodale, M. 1995. “Size-Contrast Illusions Deceive the Eye but not the Hand,” Current Biology 5, pp. 679-685. Armstrong, D. M. 1968. A Materialist theory of the mind (London: Routledge & Kegan Paul). Ayer, A. J. 1940. The Foundation of Empirical Knowledge (London: Macmillan) BonJour, L. 2001. “Toward a Defense of Empirical Foundationalism” in DePaul (2001), pp. 21-38. BonJour, L. and Sosa, E. 2003. Epistemic Justification (Oxford: Blackwell) Burge, T. 1996. “Our Entitlement to Self-Knowledge,” Proceedings of the Aristotelian Society 96, pp. 91-116, reprinted in Ludlow and Martin (1998), pp. 239-263. Chisholm, R. 1942. “The Problem of the Speckled Hen,” Mind 51, pp. 368-373. Chisholm, R. 1966. Theory of Knowledge (Englewood Cliffs: Prentice-Hall). Page 26
Clark, A. 2001. “Visual Experience and Motor Action: Are the Bonds Too Tight?” Philosophical Review 110, pp. 495-519. DePaul, M. ed. 2001. Resurrecting Old-Fashioned Foundationalism (Lanham: Rowman & Littlefield). Dretske, F. 1993. “Conscious Experience,” Mind 102, pp. 262-283. Dretske, F. 1995. Naturalizing the Mind (Cambridge MA: MIT Press). Evans, G. 1982. Varieties of Reference (Oxford: Clarendon Press). Fales, E. 1996. A Defense of the Given (Lanham: Rowman & Littlefield). Güzeldere, G. 1995. “Is Consciousness the Perception of What Passes in One’s Own Mind?” in T. Metzinger (ed.), Conscious Experience (Paderborn: Schöningh), pp. 335-357. Harman, G. 1990. “Intrinsic Quality of Experience,” Philosophical Perspective 4, pp. 31-52. Langsam, H. 2002. “Externalism, Self-Knowledge, and Inner Observation,” Australasian Journal of Philosophy 80, pp. 42-61. Ludlow, P. and Martin, N. eds. 1998. Externalism and Self-Knowledge (Stanford: CSLI Publications). Lycan, W. G. 1996. Consciousness and Experience. (MIT Press) McDowell, J. 1994. Mind and World (Cambridge, MA: Harvard University Press). Milner, A. D. and Goodale, M. A. 1995. The Visual Brain in Action (Oxford: Oxford University Press). Moore, G. E. 1944. “Russell’s ‘Theory of Descriptions’” in P. A. Schlipp (ed.), The Philosophy of Bertrand Russell (Evanston IL: Northwestern University Press), pp. 175-225. Nuccetelli, S. ed. 2003. New Essays on Semantic Externalism and Self-Knowledge (Cambridge MA: MIT Press). Papineau, D. 2002. Thinking about Consciousness (Oxford: Clarendon Press). Peacocke, C. 1996. “Entitlement, Self-Knowledge and Conceptual Redeployment,” Proceedings of the Aristotelian Society 96, pp. 117-158, reprinted in Ludlow and Martin (1998), pp. 265-303. Page 27
Price, H. H. 1941. Review of Foundations of Empirical Knowledge by A. J. Ayer, in Mind 50, pp. 280-293. Rosenthal, D. 1993. “Thinking that one thinks” in M. Davies and G. Humphreys (eds.), Consciousness (Oxford: Blackwell, 1993), pp. 197-223.