Foundationalism postulates foundations for knowledge. Here agree its two branches — the rationalist and the empiricist — even if they disagree in their respective foundations, and disagree on how to erect a superstructure.

For the rationalist, only rational intuition can give a secure foundation, and only deduction can build further knowledge on that foundation. Here the model of knowledge is the axiomatic system, with its self-evident axioms and its theorems derived through logical deduction. Rationalists, therefore, were the logicists, who tried to reduce all mathematics to self-evident axioms.

More ambitious yet, Descartes sketched in his Meditations a strategy for rationally founding all knowledge, not only mathematical knowledge. But his strategy required substantive commitments that turned out to be less than axiomatic — commitments of natural theology, for example.

The failure of rationalism is evident both in Descartes and in logicism.

For their part, empiricists accept not only foundations by rational intuition but also foundations by sensory experience. Equally unsuccessful, however, was their project of reducing all physical reality to sensory experience — whose apotheosis is Carnap's phenomenalism. Besides, as Hume showed, the future cannot be predicted deductively: The reasoning required outstrips logical deduction.

Empiricism thus becomes more liberal than rationalism in two respects: First, it accepts a broader foundation, provided not only by rational intuition but also by sensory experience; second, it admits not only deductive reasoning but also inductive reasoning.

KNOWLEDGE AND JUSTIFICATION

To know, you must at least be right; but you can be right without knowing. If you say that the dice will come up 7, you may be right without knowing. If you know you are most likely cheating, one die perhaps showing 4 on all six of its faces while the other shows 3. But if you are right on a roll of fair dice you must simply have guessed right.

What distinguishes the cheater who knows from someone who is right only by luck? For one thing, the cheater predicts with a good basis. He is justified in his belief. But not just any sort of justification is thus relevant to knowledge, as shown by the bearing of confidence on success. The athlete, for example, tries to be confident. The fact that it will help gives him practical justification for confidence. But even if bound to win and thus to be right in the object of his confidence, such practical justification moves him no closer to knowledge than is the honest player who simply guesses right. What would bring him thus closer to knowledge is not the instrumental value of his confidence, but information about the skills and abilities of the competitors.

In this chapter I will distinguish between several forms of introspective awareness of sensations and give an account of their most basic properties. My goal is to formulate a theory of introspection that can be integrated with the reductionist view of sensations that I have defended in earlier chapters.

I will also develop a skeletal theory of the epistemological status of introspective beliefs about sensations. It is often claimed in philosophy that all such beliefs are infallible, and that the scope of introspection is sufficiently comprehensive that we can be said to be omniscient with respect to the realm of sensations. I will argue that these claims are much too strong, but I will also maintain that they contain an element of truth. I will conclude with a brief discussion of the question of what we mean when we describe an introspective belief as justified.

Frequently, when we say that a subject is introspectively aware of a sensation of type φ, we are making a claim that has the following truth conditions: (1) S has a sensation of type φ; (2) S believes that he or she has a sensation of type φ; and (3) the belief cited in (2) is both caused and confirmed by the sensation cited in (1). When a claim of this sort is true, S has what I call basic awareness of one of his or her sensations.

Basic awareness is a form of consciousness.

In this chapter I will be concerned with a group of interrelated episte-mological problems.

The first of these problems is the traditional problem of other minds. Let us say that a qualitative characteristic of sensations is an immediate quale if one or more human beings have direct access to it. Now common sense encourages one to hold that other human beings have sensations that are more or less similar to one's own, in the sense of exemplifying immediate qualia that are more or less like the immediate qualia one's own sensations exemplify. Common sense also encourages one to hold that the same is true, though to a lesser degree, of the members of certain biological species that are not too distant from homo sapiens. Can these beliefs be justified? If so, how? These two questions constitute the traditional problem of other minds.

Assuming that this first problem admits of a positive solution, in the sense that it is possible to show that our intuitions about other human beings and members of neighboring biological species are justified, the question arises whether it is possible to justify ascriptions of sensations with immediate qualia (hereafter called immediate sensations) to beings who stand at some remove from ourselves.

In recent years, philosophers have been much concerned with questions about the contents or meanings of the sensation concepts that we employ in everyday life. They have presented and defended a wide variety of positions. At one end of the spectrum is the view that our sensation concepts acquire their contents from internal ostensive definitions, and that these concepts are therefore purely phenomenological, in the sense that the question of whether a given concept applies to a given sensation depends entirely on the immediate phenomenological nature of the sensation. At the other end of the spectrum we find the view that our concepts of sensations are similar in content to theoretical concepts such as the concept of gravity and the concept of electric charge. According to this second view, the contents of our sensation concepts derive from the roles that the concepts play in our commonsense theory of mental activity. This view denies that the question of whether a concept applies to a given sensation has anything to do with the immediate phenomenological nature of the sensation. To determine whether the concept applies to a sensation, the view maintains, it is necessary and sufficient to consider the causal role that the sensation plays in the internal economy of the being in whom it occurs. If the sensation bears the right causal and counterfactual relations to other things (namely, to external stimuli, to behavior, and to other internal states), it falls under the concept.

If the arguments up to this point have been successful, we have reason to hope that type materialism is correct. We have encountered arguments that purport to show that type materialism is superior to dualism and the double-aspect theory, and also some arguments that purport to show that functionalism is badly flawed. In view of these arguments, it looks as though the only alternatives to accepting type materialism are (1) to deny the existence of our subject matter (that is, to reject the realist attitude toward sensations that we have assumed from the start), and (2) to accept the nihilistic view that sensations exist but are too elusive to be caught in the net of any metaphysical theory. Although there are people who favor (1) or (2), most of us would be deeply distressed if we were forced to embrace one of them.

Still, it would be premature to infer that type materialism is true; for there are a number of objections against it, and some of these objections have a certain amount of prima facie appeal. Indeed, there are objections against it that are widely thought to be conclusive.

The objections are of three kinds. First, there are arguments that dualists have constructed with a view to showing that their position is to be preferred to token materialism. Type materialism implies token materialism, and it is therefore at risk from all of the arguments that are directed against the latter.

We are concerned here to understand the ultimate metaphysical nature of both sensations and qualitative characteristics. Apart from the behaviorist doctrine that sensations can be reduced to congeries of behavioral dispositions, and the eliminativist thesis that the existence of sensations is in some sense fictional, both of which seem to me to be wildly implausible, there are just four views that one can take of these matters – type materialism, dualism, the double-aspect theory, and functionalism. We have reached the conclusion that, under certain assumptions, type materialism is to be preferred to dualism and the double-aspect theory. It is time now to look at functionalism.

Functionalism is broadly relevant to most of the main concerns of the philosophy of mind. It gives us a unified perspective from which to view sensations, emotions, the will, the nature of the self, and propositional attitudes (belief, desire, intention, and the other mental states that seem to involve relations to propositions). Here I wish to focus on those aspects of functionalism that are concerned with the sensory realm. Because of this, my exposition of functionalist doctrines will sidestep some technical questions that would have to be faced if we were considering the functionalist account of propositional attitudes. Further, my criticisms of functionalism will be directed only against its claims about sensations. They are not intended to call other functionalist doctrines into question.

In this chapter I will present and defend a theory of unity of consciousness. The theory that I will recommend has two features that set it apart from other theories about the same topic. First, it asserts that it is a mistake to think of unity of consciousness as having just one source. There are a number of different relations (hereafter called unity relations) that count as sources or forms of unity of consciousness. Second, my theory denies that all of the sources or forms of unity are compositional. Let us say that a relation R is compositional if it meets two conditions: First, the relata of R are sensations; second, there exists some other relation R′ such that if x and y stand in the relation R to one another, they do so because x and y both bear R′ to some other thing that is not itself a sensation. For example, having the same owner as is compositional: When two sensations bear this relation to one another, they do so because each of them stands in the relation being owned by to the same subject of experience. A relation is purely sensory if its relata are sensations and it is not compositional. I will defend the view that some unity relations are purely sensory.

The claim that it is possible to obtain knowledge by sense perception has been a target for skepticism since the dawn of philosophy. In contrast, there have been no sustained skeptical challenges to our claim to be able to obtain knowledge by introspection. What accounts for the difference? Is it impossible to extend the arguments that have been formulated by skeptics so as to obtain new arguments that apply to introspective beliefs? If so, why? If one were to try to construct a skeptical argument concerning introspective beliefs, at exactly what points would one encounter problems?

It will be useful to begin by considering a line of thought that is frequently used to justify skepticism about sense perception. Let PB be the set of propositions that represent the perceptual beliefs of a certain normal subject, S. Further, let SH be the hypothesis that is obtained by conjoining the following five propositions: S is part of an elaborate psychology experiment that is being conducted in a laboratory on a remote planet; S is a brain in a vat; S is connected to a computer that monitors all of S's thoughts; all of the sense experiences S has had up to now have been caused by events inside the computer; and in the future the computer will provide experiences like the ones that S has had in the past (that is, experiences that confirm the members of PB and that lead S to adopt new beliefs that are consistent with the members of PB).

In these pages we will be concerned with sensations themselves (that is, with concrete sensory events) and also with certain of the characteristics that sensations exemplify. Although we will consider characteristics of other kinds as well, we will be primarily concerned with characteristics that are qualitative. Qualitative characteristics include being a pain and being an itch. They also include the sensory characteristic that is exemplified by the gustatory sensations one has when one tastes orange juice, and the sensory characteristic that is exemplified by the olfactory sensations one has when one smells gasoline.

A terminological point. Ordinary language does not contain many names for qualitative characteristics of sensations (or qualia, as I shall sometimes call them). In addition to “being a pain” there is “being an itch” and “being a case of pleasure.” But there are not many others. In most cases we pick out qualitative characteristics of sensations by resorting to descriptions, and this is what I have done in giving the last two examples in the first paragraph. However, it can seem that descriptions like “the sensory characteristic that is exemplified by the gustatory sensations one has when one tastes orange juice” suffer from crippling ambiguities. If, for example, one tastes orange juice right after brushing one's teeth with mint flavored toothpaste, one experiences a sensation – a most unpleasant sensation – that is quite different than the one that is normally associated with tasting orange juice.