Skip to content
printicon

Thought and experience: An investigation into concept empiricism

Research project

According to a classic concept empiricist view, defended by, e.g., Locke (1632-1704), all thought materials derive from experience. We can think what we have experienced and combinations of what we have experienced, but nothing more. In this general form, concept empiricism has always been controversial, and today it has relatively few defenders. However, there is a limited version of concept empiricism that even critics of Locke's general thesis often accept. This limited concept empiricism (henceforth LCE) says that there are concepts of simple sensory qualities – like red and pain – that one can acquire only through experience. LCE is an intuitively appealing view. But – perhaps partly as a consequence of this – it has more often been asserted than carefully defended. The purpose of this project is to examine whether there are good reasons to believe LCE. The project will distinguish and examine different versions of the view, different theories of concepts and their individuation, and various arguments for and against LCE. The working hypothesis is that the reasons for believing the view are considerably weaker than is often supposed. In addition to its intrinsic interest, the investigation bears on a number of other issues. It touches on issues like: (i) whether all thoughts can be expressed in language, (ii) what is required for linguistic understanding, and (iii) to what extent the mind is "private". The investigation also bears on the debate about the so-called "knowledge argument" for mind-body dualism, in which LCE has played a prominent role.

Head of project

Project overview

Project period

2007-01-01 2010-12-31

Funding

Finansår , 2007, 2008, 2009

huvudman: Riksbankens jubileumsfond, finansiar: , y2007: 360, y2008: 350, y2009: 350,

Research subject

Philosophy and philosophy of science

Project description

1 Introduction

Locke (1690) asked: "How comes [the mind] to be furnished?" He meant by this: How does one acquire the ability to think whatever one can think? How, for example, am I able to think about fictional creatures or places I have never visited?

Locke's answer, later shared by, e.g., Hume (1739) and Russell (1912), was that the mind gets furnished through experience. We can think what we have experienced and permutations of what we have experienced, but nothing more. When I think about fictional creatures or places I have never visited, I combine in my mind simple ideas all of which I have derived from experience. This view, that all materials of thinking derives from experience, is known as concept empiricism.

In its general form concept empiricism has always been controversial. Kant (1781), for example, held that we have concepts – among them *cause* and *substance* – which we have not derived from experience. And, even though there has been a renaissance for a general form of concept empiricism lately (Barsalou 1999, Prinz 2002), the position has had relatively few defenders in recent decades.

There is, however, a limited form of concept empiricism that even critics of Locke's general view often accept. This limited concept empiricism (henceforth LCE) says that there are concepts of sensory qualities – like red and pain – that one can acquire only through experience. Kant accepted such a thesis, as do many contemporary philosophers. One contemporary context in which the view is often expressed is the discussion about Frank Jackson's (1982) thought experiment concerning Mary. In Jackson's scenario, Mary is confined from birth to a black-and-white room in which she learns all that one can learn about colour and perception from books, lectures and black-and-white TV-shows. However, she never experiences a chromatic colour. Jackson originally used the thought experiment to argue for mind-body dualism. (The argument was that Mary plausibly knows all the physical facts, but there is one fact she doesn't know, namely what it's like to see red. Hence this must be a non-physical fact.) But in the debate about this argument, it is often claimed that there is a range of concepts of chromatic colour qualities that one can't possess unless one has experienced these qualities. The claim is made by, e.g., Tye (2000), Papineau (2002) and Thau (2002).

LCE is an intuitively appealing view. But – perhaps partly as a consequence of this – it has more often been asserted than carefully defended. The purpose of this project is to assess whether there are good enough reasons to believe LCE. The project will distinguish and examine different versions of the view, different views of concepts and their individuation, and various arguments for and against LCE. My main working hypothesis is that the reasons for believing one version or other of the view are considerably weaker than is often supposed.


2 Specification of the project

2.1 Versions and specifications of LCE

One can distinguish at least two versions of LCE:

Unqualified LCE: There is a concept *Q* for some sensory quality-type Q such that human subjects who have experienced Q occasionally employ *Q* and, for any possible thinker S: If S employs *Q* at a time t, then S experiences Q at t or has done so prior to t.

Anthropological LCE: There is a concept *Q* for some sensory quality-type Q such that human subjects who have experienced Q occasionally employ *Q* and for any subject S that conforms to the laws of human psychology: If S employs *Q* at a time t, then S experiences Q at t or has done so prior to t.

(To save space, I shall sometimes disregard the differences between the two theses and discuss them in one breath.)

Some comments to the theses:

(i) The psychological law invoked in the anthropological claim will here be taken with a "ceteris-paribus clause". Stating a law, Anthropological LCE makes a claim not only about what's true of all human beings, but also what would be true of us in counter-factual circumstances. However, the claim allows for extraordinary exceptions. To illustrate the claim, consider the thesis that: if a human subject S can pass for a competent Japanese speaker at a time t, then S has had some contact with the Japanese language at some earlier time. This is plausibly true of all actual human beings, and the regularity seems robust enough to support counter-factuals. We can safely suppose that if a given speaker of Japanese hadn't ever been exposed to that language, he or she wouldn't have passed for a Japanese speaker. But, plausibly, this claim admits of some *possible* exception; it's plausibly *in principle* possible for a human being to develop into a passable Japanese speaker by some accident that restructures her brain in the right way.

(ii) Those familiar with the history of philosophy may think that, because of an observation due to Hume (1739), both theses need to be qualified to be even remotely plausible. The observation is that, if a human subject has experienced a wide range of colours, she can plausibly "provide" herself with a concept of, say, a shade of blue that she has never experienced. The observation raises, I believe, several issues. But here I shall just propose one way in which one can accept the observation and still endorse either or both of the two theses, without qualification.

LCE does not claim that there are experience-dependent concepts of *all* sensory qualities. We should not, for example, take the theses to imply that there is a concept of *red-and-blue striped* that one can employ only if one has experienced red-and-blue stripes. The claim is only that there are experience-dependent concepts of *some* sensory qualities. Which qualities are these? The natural answer is that they are the *simple* qualities. The reason there is no experience-dependent concept of red-and-blue striped is that red-and-blue striped is a complex quality. If, on the basis of experience, you have the right kind of concepts of its components – redness, blueness and stripedness – then you can combine these concepts in your mind and form the right kind of concept of red-and-blue striped. No additional experience is necessary.

But which qualities are ultimately simple? Hume seems to have supposed – and it's somewhat natural to do so – that each discriminable shade of colour constitutes a simple quality. If one so identifies the simple qualities, then the "missing-shade-of-blue" scenario does seem to require a qualification to the LCE-theses. However, it is not mandatory to identify simple qualities in this way. An alternative, drawn from standard contemporary colour models (like the Natural Colour System), is to take there to be six, and just six, simple colour qualities: red, blue, green, yellow, black and white. Each shade of colour can be characterised as either a pure version of one of these, or as a "perceptual mixture" of two or more than two of them. For example, shades of orange can be characterised as mixtures of red and yellow, and are thus, by this suggestion, complex qualities. If one identifies simple colour qualities in this way, then the missing-shade-of-blue scenario does not require a qualification to either of the two theses. By experiencing a wide range of colours, I may acquire the right kind of concepts of, say, blue and white, and I can then combine these in my mind to form the right kind of concept of a shade of light blue that I have not previously experienced.

(iii) The expressions "concept" and "employ a concept" are used in several different ways in philosophy and cognitive science. Here they shall be understood in terms of thinking a thought (or proposition). To say that a person employs a certain concept is, roughly, to say that the person thinks a certain type of thought (or proposition). Thus understood, the theses above say, roughly, that there is a certain type of thought or proposition concerning (e.g.) the quality red that one can think only if one has experienced red. I regard it as clear – though I can't take the space to justify it here – that defenders of (one version or other of) LCE have intended to defend a thesis of this kind.

To assess LCE it is crucial to consider the issue of the "individuation" of thoughts or concepts; i.e., under what conditions we think the same thought and under what conditions we think different thoughts. We can distinguish two views on this controversial issue.

According to a Referentialist view, thoughts and concepts are individuated by what objects and properties they are about, or refer to. To illustrate, suppose that I, in two acts of thinking, attribute one and the same property to one and the same object; for example, that I attribute to the city of Detroit the property of being an American city. The Referentialist view entails that I then think one and the same thought in these two cases. (For a recent defence, see Thau 2002.)

According to a Fregean view (after Frege 1892) on the other hand, I may think different thoughts in two acts of thinking even my two thinkings attribute one and the same property to one and the same object. According to this view, thoughts and concepts are typically individuated by some test or other for "cognitive significance". Suppose for example that I first think what I would express by the utterance "Motown is an American city" and then, a moment later, think what I would express by the utterance "Detroit is an American city". And suppose I have such information that I, without any apparent defect in my rationality, believe what I think in the first case but doubt what I think in the second. If so, what I think in the two cases (or, perhaps, my two acts of thinking) may be said to have different "cognitive significance" for me, even if we suppose that my thinkings attribute one and the same property to one and the same object. And according to the Fregean view, I then think different thoughts in the two cases.

One ambition of this project is to assess different theses about how thoughts and concept should be individuated. But in this presentation I shall assume that one version or other of the Fregean view is correct. The Fregean view is the more common of the two, and is embraced by most defenders of LCE, including Tye (2000) and Papineau (2002). (However, Thau 2002 combines LCE with a Referentialist view.) Arguably, LCE is also more plausible against the background of a Fregean view than against the background of a Referentialist view. The combination of LCE and Fregeanism allows that I can possess *some* concept of red even if I have not experienced red. In contrast, the combination of LCE and Referentialism does not allow that: if concepts are individuated by what they refer to, LCE entails that I can't think about red at all unless I have experienced it. To this extent, the latter combination of views makes a stronger and more controversial claim than the former.

2.2 The "remodelling argument" against LCE

It has been claimed that there is a simple, decisive argument against LCE, at least in its unqualified form. The argument goes roughly as follows. Consider Mary in her black-and-white room. It is in principle possible – without presenting her with any chromatic colours – to remodel her brain in such a way that it ends up in exactly the state it would have been in if she had experienced chromatic colours. Thus remodelled, Mary would be able to name colours on the basis of observation, report which objects are typically red, describe similarity relations between colours, and so on, just as well as you or me. It is therefore reasonable to suppose – says this argument – that Remodelled Mary would have all the concepts of colours and experiences that you and I have. But Remodelled Mary would have these concepts without ever having experienced any chromatic colour. Hence at least the unqualified version of LCE is false. (This kind of scenario is discussed in Dennett 2005. Unger 1966 argues against concept empiricism on the basis of similar scenarios.)

I think it is clear, however, that we shouldn't reject any version of LCE solely on the basis of this kind of argument. For suppose there is some good reason to believe (say) Unqualified LCE. *If* there is such a reason, then we should deny that Remodelled Mary has all the concepts that you and I have, and therefore deny that the "remodelling argument" is sound. (Even if we haven't formulated a good reason to believe in LCE, we can at an abstract level appreciate the possibility of rejecting the remodelling argument on this kind of ground. Thus consider Davidson's (1987) argument that a copy of him wouldn't be able to recognise or even have thoughts about Davidson's friends as long as the copy lacked the right kind of connections with these people. We can appreciate the possibility of arguing, analogously, that there is some good reason to suppose that there is a concept of red the possession of which requires that one has the right kind of "experiential connection" with this quality, and on this ground deny that Remodelled Mary has this concept.) Thus, for the remodelling argument to be convincing, it seems it must at least be complemented with a survey of potential reasons *for* believing in (Unqualified) LCE.

2.3 Four arguments for (one version or other of) LCE

It is possible to extract or reconstruct from the literature at least four different arguments for one version or other of LCE. A main task for the project is to assess these arguments, as well as look out for other potential arguments for the view.

The argument from imagism:

According to some philosophers, there is a familiar concept *Q* for some sensory quality Q such that: If a subject S employs *Q* at a time t, then S experiences Q at t. (Papineau 2002, Thau 2002 and Chalmers 2003 endorse versions of this view.) This is a limited version of Locke's and Hume's classic imagistic view that thinking is a sort of sensing. It is clear that if this limited imagism (in one version or other) is correct, then LCE (in one version or other) is also correct.

The cognitive significance argument:

Suppose a subject has never experienced a chromatic colour quality, but has heard about such qualities and thereby come to master terms like "red", "blue", etc. It is sometimes claimed that if such a subject were presented with a red sheet of paper, she could form a concept, *R*, for the experienced quality but would lack the means to determine which quality she experiences in her "old terms". Invoking a Fregean principle of concept individuation one can on the basis of these claims argue that the concept *R* is distinct from the subject's concept *red*. The subject can, without any apparent failure of rationality, believe that the sheet of paper "is R", but doubt that it "is red". In the same way, it seems one could argue that *R* is distinct from all other concepts the subject could have had before she experienced red. (Versions of this argument can be found in Papineau 2002 and Chalmers 2003.)

The argument from the best account of sensory deprivation:

It may seem plausible that every subject that lacks experience of some (simple) sensory quality-type Q suffers from some kind of "cognitive deficiency" compared with a typical human subject who has experienced Q. Consider for example Jackson's Mary. It seems plausible that, despite her talents and extensive education, she is in one respect cognitively disadvantaged compared with those of us who have experienced chromatic colours. And one can argue that the best account of what this cognitive disadvantage consists in includes the assumption that there is a certain concept that sensorily deprived subjects lack and that other subjects have. (Versions of this type of argument can be found in Loar 1990, Thau 2002 and Byrne 2002.)

The nothing-else argument:

It's somewhat natural to think that while there may be many different concepts of red, there is one concept that characterises this quality *as it appears in experience*. If LCE is false, there must be some way, besides having the relevant type of experience, of acquiring this concept. However – so says the present argument – there can be no way of acquiring the ability to think about red *as it appears in experience* other than by having the relevant type of experience. Hence LCE is true. (Campbell 1993 appears to embrace an argument of this kind.)

My working hypothesis is that – with one potential exception – none of these arguments offers a sufficiently good reason to believe either of our two versions of LCE. The potential exception concerns whether the cognitive significance argument may provide sufficient support for Anthropological LCE. Here there is room for only some brief remarks about the arguments.

Regarding the argument from Imagism:

It seems to me true that I sometimes think about red "imagistically" and sometimes "non-imagistically". In the former cases, I (in some sense) use an instance or a "faint inner copy" of red in my thinking (Papineau 2002); in the latter cases I don't. But the relevant question for present purposes is whether there is some type of thought or proposition that one can think only imagistically, given that thoughts or propositions are individuated according to some Fregean test for "cognitive significance". This is highly doubtful. Suppose I now think imagistically about red. The crucial observation is that, in each such case, it seems possible that (a) I think (say, a moment later) non-imagistically about the same quality and (b) that it is transparently and immediately clear to me that I think about the same quality in the two acts of thinking. This knowledge may be no more fallible or inferential than the best knowledge I ever have that two acts of thinking of mine concern the same thing. Given this observation, I believe it is hard to use any cognitive significance test to justify that there is a type of proposition that I sometimes think imagistically, but can't think non-imagistically.

Regarding the cognitive significance argument:

I think it's clear that this argument can't provide any support for Unqualified LCE. It is in principle possible to have the ability to determine, in one's "old terms", which colour quality one experiences even at first sight. For example, Remodelled Mary (2.2 above) has this ability. (Note also that Remodelled Mary can have just as much *reason* to be confident in her colour identifications as anybody else; she may know exactly what remodelling she has gone through, and that she is as good as the next person at naming colours on the basis of observation.)

It is harder to determine whether the cognitive significance argument provides sufficient support for Anthropological LCE. One important observation in this context is that there is information about chromatic colours that could help a person – like Mary – to identify, in her "old terms", which quality she experiences at first sight, or at least to rule out certain alternatives regarding its identity. Thus consider the following two facts about red:

(i) At dusk, it is harder to distinguish red from black than from white.

(ii) In the dark, a red light is as easy to detect as a white light.

Note that what (i) and (ii) say about red is not equally true of all chromatic colours. For example, it is not true that *yellow* is harder to distinguish from black than from white at dusk, and it is not true that *blue* is as easy to detect in the dark as a white light (blue light scatters more than white and red light). It seems plausible that Mary, on the basis of knowing such facts, would at least be able to rule out some alternatives concerning what she experiences, if she were presented with a red sheet of paper. For example, she could plausibly determine that she does not experience yellow by the following reasoning: "Yellow is harder to distinguish from white than from black at dusk. But that colour looks like it would be harder to distinguish from black than from white at dusk. Hence I suppose this is not yellow".

Whether the cognitive significance argument ultimately provides support for Anthropological LCE is, as far as I can see, a complicated question, which may not have a clear answer. The issue seems to depend on at least three things:

First, whether there are facts – like (i) and (ii) – that can help a person who has not previously experienced a (simple) colour quality Q to, not just rule out *some* alternatives concerning which quality they experience at first sight, but rule out *all* alternatives other than the right one.

Second, how LCE is understood in the light of the-missing-shade-of-blue scenario. According to the suggestion offered above (2.1, point (ii)), Anthropological LCE rules out that a human subject can provide herself with a concept of red even if she has experienced pure versions of white, black, green, blue, and yellow, as well as perceptual mixtures of these. For the cognitive significance argument to support the thesis thus specified, it must be the case that a subject can't tell that she experiences red even if she both has had all these experiences and knows facts such as (i) and (ii).

Third, exactly how concepts and thoughts are individuated. Even Fregean tests for cognitive significance can individuate thoughts more or less finely. Frege himself (1892) wanted – at least in some cases – to distinguish propositions even if it would be impossible for an ideally rational subject to believe one and doubt the other (like the propositions *4=4* and *2+2=4*). The more fine-grained a test one assumes for distinguishing thoughts, the more hope there is to support Anthropological LCE with some version of the cognitive significance argument, but at the same time one incurs the obligation to both specify and motivate such a test.

Regarding the argument from the best account of sensory deprivation:

It seems again clear that this argument provides no support for Unqualified LCE. Even if we can't regard it as a datum that, e.g., Remodelled Mary is cognitively as well-equipped as you or me (2.2 above), we also can't regard it as a datum that she is in any respect cognitively inferior to us. There seems to be no way to settle that issue without first determining whether Unqualified LCE is true or not. But if so, the present argument does not support Unqualified LCE.

It is less obvious whether the argument can support Anthropological LCE. However, my working hypothesis is that – at least provided that the preceding two arguments can be answered – the present argument does not support the view.

Granted, there is reason to believe that, as a matter of psychological law, any human subject who lacks experience of some (simple) sensory quality Q will be cognitively deficient in some respect compared with typical human subjects who have (often) experienced Q. The evidence for this comes from actual cases where we lack experience of some observable object or property, like a taste, smell, person, or place. In these cases, we are typically in some respect cognitively deficient compared with those who have (often) experienced that object or property. And, typically, only experience can bring us out of our cognitive disadvantage (cf. Lewis 1988). We tend to remain disadvantaged even if we are told about the observable characteristics of the object or property, for example.

However, I hypothesise that, if a human subject S, who lacks experience of red, has done all that is humanly possible to learn about red, the best account of (what we have reason to believe would be) S's cognitive situation does not include the assumption that S lacks some concept that those who have experienced red typically possess.

In the first place, there is *some* reason to think that, even if S has done all that's humanly possible to learn about red, she would fail to know some fact that those who have (often) experienced red typically do know in some sense and in some way. To illustrate this possibility, consider again (i) and (ii) above (in the discussion about the cognitive significance argument). It seems that we who are acquainted with red know these facts, at least in some semi-tacit, practical way. While we may not be good at articulating them, it's plausible that we would typically be above chance if asked whether such a fact obtains or not. Moreover, it appears that we have a sort of practical mastery of such facts, which is expressed, for example, in various search tasks (such as finding a red object in the dark). Now, there may well be a very large number of facts like these concerning red. And if so, it may be psychologically impossible for S to learn them all through lessons. (Notice, though, that no reason has so far been given to suppose that there is any proposition about red that S could not think. The suggestion is that there is a set of propositions that is so large S can't come to *know*, through lessons, that all of them are true. But this is compatible with S being able to *think* each one of them.)

Further, whether or not there are facts that S wouldn't know, there is reason to believe that she would be deficient (compared with those of us who have experienced red) with regard to the following three abilities. First, she would be less adept at visualising red than we are. Second, she would be slower and less reliable at recalling whatever facts she knows about red, and especially so when it comes to visually accessible facts like (i) and (ii). Third, she would be less skilled at putting whatever facts she knows into practical use in search and identification tasks, and so she would execute such tasks less swiftly and less reliably than we do. The evidence for all of this comes, again, from actual cases. When we lack experience of some taste, smell, person, or place, we tend to be deficient in these respects compared with those who have had the relevant experiences.

It seems to me plausible that the above is, or approximates, an exhaustive account of how (we have reason to believe that) S would be cognitively deficient compared with those of us who have experienced red, in the circumstances described. Now, there is clearly no *explicit* mention, in this account, of the assumption that there is some concept of red that we possess and S lacks. One could still argue that there is some concept S would lack *in virtue of* some of the cognitive deficiencies mentioned, and that hence the relevant assumption is entailed by some aspect of the account. However, my working hypothesis is that, provided that the arguments from Imagism and cognitive significance can be answered, there is no basis for so arguing.

Regarding the nothing-else argument:

If the preceding arguments for LCE can be met, I think it is doubtful that this argument offers support for either version of LCE. At least, this much seems right: If the above arguments are satisfactorily answered by the outlined considerations, there is hardly any special reason to suppose that there are concepts of *sensory qualities* that one can acquire only through experience. For these considerations suggest that a person who has learned a lot about some (simple) sensory quality but has never experienced it is in the same kind of cognitive situation that somebody is in who has learned a lot about but never experienced some other observable object or property, like a person or place. The next question would then be if there is reason to accept some more general concept empiricism, for example, a thesis to the effect that there are concepts of observable objects and properties generally that one can acquire only through experience. This issue will be explored in the project. In this context, it will be examined what ways there are, in general, of acquiring various concepts of objects and properties.


3 The significance of the project

The question whether LCE (in one version or other) is correct is part of the more general issue about the relation between experience and thought. I take these issues to be intrinsically interesting. But the investigation has implications for a variety of other issues as well.

As already mentioned, LCE has played a prominent role in discussions about the mind-body problem, and in particular Jackson's "knowledge argument" for dualism. One ambition of the project is to investigate the role LCE has played in these discussions, and specifically whether LCE has played any essential role – and if so which – for the arguments in the context of which the view has been expressed.

LCE also has implications for the issue about the "privacy" of the mind. If LCE is correct, there are propositions concerning the conscious states of others that one can't as much as think – let alone know to be true – unless one has oneself been in conscious states that are similar enough.

A related implication of LCE – and perhaps the theoretically most significant one – concerns the conditions of language and communication. It is natural to suppose (i) that what I can think I can in principle also express in some language. Several influential 20th century philosophers have further argued (ii) that to understand a speaker of a language it is never required that one be similar to the speaker in any conscious, or otherwise "inner", respect; agreement in overt dispositions (and perhaps natural and social habitat) is always sufficient for understanding (Wittgenstein 1953, Quine 1960, Dummett 1973). However, LCE seems to imply that these two views – (i) that I can express in some language anything I can think, and (ii) that linguistic understanding is independent of "inner" similarity – can't both be true. Assuming – as seems reasonable – that one can understand an expression only if one can think what it says, it follows from LCE that: if it is possible to express in a language anything one can think, then it is possible for a speaker S to produce a linguistic expression that can be understood only by those who have had experiences similar enough to those of S. Among other reasons, it is because it has this implication that it is of some importance to determine whether we should believe LCE.

References

Barsalou, L. (1999) "Perceptual symbol systems", Behavioral and Brain Sciences, 22, 577-609.

Byrne, A. (2002) "Something about Mary", Grazer Philosophische Studien, 63, 123-40.

Campbell, J. (1993) "A simple view of colour", Reality: Representation and Projection, ed. J. Haldane and C. Wright, Oxford: Oxford UP.

Chalmers, D. (2003) "The content and epistemology of phenomenal belief", Consciousness: New Philosophical Perspectives, ed. Q. Smith and A. Jokic, Oxford: Oxford UP.

Davidson, D. (1987) "Knowing one's own mind", Proceedings and Addresses of the American Philosophical Association, 61, 441-58.

Dennett, D. (2005) Sweet Dreams: Philosophical Obstacles to a Science of Consciousness, Cambridge, Mass.: MIT Press.

Dummett, M. (1973) Frege: Philosophy of Language, London: Duckworth.

Frege, G. (1892/1997) "On Sinn and Bedeutung", The Frege Reader, ed. M. Beaney, Oxford: Blackwell.

Hume, D. (1739/1978) A Treatise of Human Nature, ed. P. Nidditch, Oxford: Oxford UP.

Jackson, F. (1982) "Epiphenomenal qualia", Philosophical Quarterly, 32, 127-36.

Kant, I. (1781/1929) Critique of Pure Reason, London: Macmillan.

Lewis, D. (1988) "What experience teaches", Proceeding of the Russellian Society, 13, 29-57.

Loar, B. (1990/1997) "Phenomenal states", The Nature of Consciousness: Philosophical Debates (1997), ed. N. Block et al., Cambridge, Mass.: MIT Press.

Locke, J. (1690/1979) An Essay Concerning Human Understanding, ed. P. Nidditch, Oxford: Oxford UP.

Nemirow, L. (1990) "Physicalism and the cognitive role of acquaintance", Mind and Cognition: A Reader, ed. W. Lycan, Oxford: Blackwell.

Papineau, D. (2002) Thinking about Consciousness, Oxford: Oxford UP.

Prinz, J. (2002) Furnishing the Mind: Concepts and Their Perceptual Basis, Cambridge, Mass.: MIT Press.

Quine, W.V.O. (1960) Word and Object, Cambridge, Mass.: MIT Press.

Russell, B. (1912) The Problems of Philosophy, Oxford: Oxford UP.

Thau, M. (2002) Consciousness and Cognition, Oxford: Oxford UP.

Tye, M. (2000) Consciousness, Color and Content, Cambridge, Mass.: MIT Press.

Unger, P. (1966) "On experience and the development of the understanding", American Philosophical Quarterly, 3, 48-56.

Wittgenstein, L. (1953) Philosophical Investigations, ed. G.E.M. Anscombe and R. Rhees, Oxford: Blackwell.