Reference And Representation Research Paper

1. Representation: The Most General Notion

What is common to all cases of representation is a correspondence between various states of what does the representing and various states of what is represented. Thus, the number of tree rings represents the age of a tree by virtue of the fact of the possibility of matching different tree ring states with the different ages. In this case, the correspondence is a very simple one: n tree rings corresponds to an age of n years. Likewise, there is a simple correspondence between various states of petrol gages and various levels of gas in the tank to which the gage is connected. But there are highly complex cases as well. Inscriptions in secure codes of credit card numbers represent those numbers, but via extraordinarily complex correspondences. When there are pairwise correspondences between differing states, there is the possibility of carrying information. This is how the number of tree rings can give information about the age of a tree. But, of course, the nature of the correspondence needs to be known for this possibility to be realized.

In many cases, the relevant correspondence relation has a causal underpinning that runs from the various states being represented to the various states doing the representing. For example, the level of gas in the tank causes the reading on the gage, and the appearance of the subject causes the array of paint on the canvas (via the actions of the painter). But structures can represent how things will be; a drawing may represent where a playground will be located but is, obviously, not caused by where it will be. There is representation in the wide sense when there are nonaccidental correspondences between states of affairs. There will be a causal explanation of these nonaccidental correspondences but its precise nature varies from case to case. (See Dretske 1995, Chap. 1 for more on the variety, but note that he ties representation to purpose.)

1.1 Representational Content And Reference In General

Things that represent are said to have representational content, that content being how things are being represented as being. The usual way of capturing representational content is set theoretically: content is given by a division of all the complete ways things might be—the event space in the language of statistics—into those that accord with how things are being represented as being and those that do not. Complete ways things might be are ways things might be in which everything is settled one way or the other. They are often called possible worlds. The content is then the set of possible worlds that accords with how things are being represented as being. For example, if a petrol gage represents a gas tank as half-full, its being half-full is the content, and that is identified with the set of possible worlds where the tank is half-full. There are, though, considerable differences of opinion about the nature of the possible worlds appealed to in this style of account. This research paper touches briefly on these in the discussion of representation in mind and language. (For an accessible discussion of the main views, see Bigelow and Pargetter 1990, Chap. 4.)

One advantage of the set-theoretic approach to content is that it allows a simple account of when things are as they are being represented as being. Things are as they are being represented as being—the representation is true or correct or accurate—if and only if the way things actually are is a member of the content. Or, to say it in terms of possible worlds, the representation is accurate if and only if the actual world is a member of the set of worlds which is the content of the representation. In still other words, states that represent say that things are a certain way in the sense of delimiting the possibilities for how they might be, and count as accurate when how things actually are is inside the relevant region of ‘logical space.’

A second advantage is that it allows a simple account of the widest sense in which states that represent are about, or refer to, this, that, or the next thing. Suppose that a barometer’s reading such and such represents that the surrounding air pressure is so and so. Intuitively, the reading is about the air pressure being so and so. This can be spelt out as follows. In every possible world where the reading is correct, in every world where things are as the barometer represents them, the air pressure is so and so. States which represent are about or refer to what is the case in all the worlds where things are as they are being represented as being. (For more on ‘aboutness’ and the difference between being about an object and a property, see Dretske 1995, Chap. 1, Sect. 5.)

1.2 Mental Representation

A footprint can be thought of as representing the shape of the foot that made it an hour ago. But equally it can be thought of as representing the weight of the person who made it, the shape of the footprint itself half an hour ago, the shape of the foot that made it one day before the foot print was made, and so on. Almost any state that represents will have many representational contents, and it is a nonsense question to ask which is its real content, to ask what the state really represents.

Among the representational structures of especial interest to humans are their brains, and, especially, those parts of our brains that constitute our powers of mental representation. In particular, humans belie e that things are thus and so, and do so by virtue of the processes in their brains which constitute their belief representing that things are thus and so. These processes qua states of brains will themselves have many contents including, for example, that the temperature of the surrounding brain fluid is thus and so, but only one will be the content of what is believed. To suppose otherwise would make what is believed absurdly indeterminate. Therefore, a major question in the philosophy of mind is what determines, out of all the contents of the brain structures that are beliefs, the one which is the content of those beliefs.

One popular answer is: the functional role the structure plays, where this is understood as the role the structure plays in mediating between: inputs to subjects, typically via their sense organs, outputs understood in terms of behavioral responses, and other internal states. However, there are many different ways of fleshing out this answer. (For a recent survey, see Braddon-Mitchell and Jackson 1996, Chaps. 3, 5.)

2. Linguistic Representation And Reference

2.1 Introduction

Language is an especially powerful representational device. With it, highly detailed claims are possible concerning how things are that far outrun, in terms of detail and precision, what can be done with barometers, petrol gages, etc. There is a syntactical distinction between sentences and parts of sentences that roughly corresponds to an important semantic or representational distinction.

A sentence makes a claim about how things are. This is how it gets to be true or false: true when the way the sentence represents things is the way they are, false otherwise. This explains the self-evident status of the biconditional ‘Snow is white’ is true if and only if snow is white. The left-hand side is true if and only if things are as ‘Snow is white’ represents them as being. Moreover, as a good way of stating how ‘Snow is white’ represents things as being is to use the sentence itself, the biconditional comes out true.

By contrast, parts of sentences like ‘Snow’ and ‘is white’ contribute in a systematic way to how whole sentences in which they appear represent things as being, but do not themselves represent how things are. This is why they are not themselves either true or false. They can, however, be thought of as referring to, respectively, snow and white things. This is because the contribution that the presence of the word ‘snow’ makes to the conditions under which ‘Snow is white’ is true, that is, to how the sentence represents things as being, is such that the sentence cannot be true unless there is some snow. Similarly, the contribution that the presence of ‘is white’ makes is such that the sentence cannot be true unless there are some white things.

Many recent discussions have focused on the referential properties of names, and many authors restrict the use of the term ‘refer’ to names. They prefer to talk of the relation between predicates (and, sometimes, common nouns) and things in terms of whether the predicates are true of the things or whether the things satisfy the predicates.

The next section considers the three main theories of reference for names.

2.2 Theories Of Reference For Names

A word like ‘London’ is used for something taken to have certain properties. This is why it is not used for water or Paris—it is known that they do not have the right properties. On the other hand, the large city that is the capital of England does have the right properties. In short, certain properties or descriptions are associated with the word ‘London’ and this guides its use. This suggests a simple theory of reference for names in general: ‘N’ refers to that which satisfies the descriptions associated with ‘N.’ If we use ‘D’ for those descriptions, another way of stating this theory is by saying that ‘N’ is short for the definite description ‘the D.’

A famous advantage of the description theory of reference is that it can explain why certain identity sentences are informative. Everything is, trivially, self-identical. How then can it be informative to be told that Hesperus is identical to Phosphorus? The answer is that the first name is associated with being the star first seen in the evening and the second name is associated with being the last star to disappear from the night sky as morning approaches, and it is informative to be told that one thing has these two properties. (The classic source for claims of this kind is Frege 1952.)

Recent discussions in the philosophy of language have raised a number of objections to the description theory of reference. (A good account of the major objections can be found in Devitt and Sterelny 1987. A classic source is Kripke 1980.)

First, if ‘N’ is an abbreviation of ‘the D,’ it should make no sense to suppose that N might have failed to be D. But, in fact, it typically makes perfect sense. For example, the descriptions associated with ‘Shakespeare’ include as a major part that he was the author of certain plays. (Surprisingly little is known about Shakespeare’s life.) But it makes perfect sense to suppose that Shakespeare did not write the plays associated with his name. This, indeed, is what those who think that Bacon wrote them do suppose. A similar point can be made for names of natural kinds. The description associated with ‘water’ is something like potable, clear liquid that fills lakes and falls as rain from the sky—this is roughly what is in a dictionary. However, if on some remote planet—traditionally called ‘Twin Earth’ by philosophers—there is found clear potable stuff in its lakes which falls from its sky, etc. that is not H O, it seems that the right thing to say would be that this stuff—XYZ, as it is often dubbed— is not water. As it would naturally be said, ‘It has been discovered that water is not the only clear potable liquid that fills lakes, falls from skies etc. in the universe.’ Conversely, if H O on Twin Earth were black and rather tar-like, it would still be water. As it would naturally be said, ‘It has been discovered that water comes in a quite different form on Twin Earth from its form at room temperature on Earth.’ (The classic source for Twin Earth is Putnam 1975.)

Second, if names are abbreviated definite descriptions, then their reference should in general be variable. By changing what is supposed to be the case, it should be possible to shift what they refer to. This is because the reference of definite descriptions is in general shift-able. ‘The President of the USA in 1998’ refers to Bill Clinton, but had he lost to George Bush, it would refer to Bush and not Clinton. But it seems that the reference of names cannot be shifted; they are, as it is said, rigid designators. ‘Clinton’ always refers to the very same person.

Description theorists have proposed a number of replies to these (and other) objections. For example, they have urged that it is wrong to conflate the description that settles the reference of a name with the description (s) most famously associated with a name. Although Shakespeare is most famous for the plays, the key description that settles the reference of the name is something like ‘the person who is a certain kind of causal origin of the use of the term ‘‘Shakespeare.’’ ’ The debate over whether or not Shakespeare wrote the plays is the debate over whether the person who is a certain kind of causal origin of the occurrences of the name on the covers of plays etc. did indeed write the plays. (For a recent catalogue of replies that the description theorists might make to various objections, see Jackson 1998.)

The causal theory of reference is the best known alternative to the description theory. The fundamental idea is that ‘Aristotle’ in our mouths refers to the person it does because of an initial act of baptism that tags that person with a name which then got passed on from speaker to speaker as the name to use for that tagged person until it ended up in our mouths. Of course, the name itself underwent various changes as it passed from the ancient Greek to Modern English. A similar account is suggested to explain why ‘water’ refers to H O. H O was baptized with the word ‘water’ (in ignorance of the fact that it was H O that was being baptized) and then the word was passed on as the word to use for that which had been baptized. This account explains why the clear potable stuff that fills the lakes etc. on Twin Earth is not water, is not what is referred to when the word ‘water’ is used, by appealing to the fact that it was not the natural kind baptized.

There are a number of questions to ask about the causal theory. Is it superior to a description theory that gives a prominent role to a description like ‘was baptized … and subsequently used for what had been so baptized’? Is it supposed to apply to the naming of as yet unborn children? Remembering that causal interactions are invariably with a very large number of factors, should a descriptive element be added to ensure the ‘right’ reference? (Some of these questions are addressed in Devitt and Sterelny 1987.)

Finally, there is the direct reference theory of reference. On this view, names like ‘Aristotle’ refer directly to Aristotle. The only semantic property of a name is its bearer. It is hard to discuss this view because it steadfastly refuses to address what seems a perfectly legitimate question, namely, what makes it true that ‘Aristotle’ refers to Aristotle and not to Plato or my dog? It is also hard to see what it can say about empty names. It has sometimes been speculated that Socrates never existed. ‘He’ was merely an expository device of Plato’s. Although this is almost certainly false, it makes sense to suppose that it is true. But is this possible if ‘Socrates’ has no meaning apart from its bearer? (For references and more discussion and dissent, see Devitt 1996, Chap. 4.)

3. Mental Representation And Linguistic Representation

3.1 How Are They Connected?

As noted above, our minds represent; in particular, to believe is to represent that things are thus and so. Usually, the easiest way for humans to express how they believe things are is to use a sentence. For example, the sentence ‘Clinton will not be impeached’ was used in February 1999 by many commentators on American politics to express what they believed. It is obvious that there is an intimate connection between mental and linguistic representation. What is controversial is the extent to which mental representation requires linguistic representation.

One view is that the phenomenon of mental representation in general, and belief in particular, is independent of language; dogs, for example, really do have thoughts and beliefs. Language is simply a device for telling about how things are taken to be. This view can though allow that there are some beliefs one could not have without having a language. It is, for instance, hard to make sense of a creature believing that there are a prime number of apples in the bowl in front of it in the absence of that creature having mastery of a complex coding system, i.e., a language.

However, some hold that belief in general requires language, where language means something substantially more sophisticated than the kind of signaling systems that, e.g., dogs have. On this view, although contentful utterance requires mental representation— telephone answering machines do not mean anything by the sentences they produce—in order to be a believer, you must have a language. Sometimes, this seems to be, at bottom, a verbal stipulation about the use of the term ‘belief’: dogs have quasibelief which is very like belief except that it is not expressed in a public language. But sometimes it is clear that the view is that dogs do not have anything like belief at all. A major question for this view is the very close connection between perception and belief. Perceptual experience seems to be shot through with belief-like representational content: the experience of it’s raining seems to be a state that by its very nature points to believing that it is raining, though it may be overridden by collateral information, for example, about the operation of a nearby sprinkler system. It is, accordingly, hard for advocates of the more radical view to avoid the implausible view that animals like dogs do not see or hear. (For a defense of the view that belief requires language, see Davidson 1984, Chap. 11, see also McDowell 1994.)

3.2 Capturing The Content Of Mental And Linguistic Representation

It was noted earlier that the standard way of capturing representational content is in terms of divisions among complete ways things might be, possible worlds. The problem of the content of logically equivalent sentences and beliefs raises an important question for how to think of these possible worlds. Intuitively, the sentences ‘X is an equilateral triangle’ and ‘X is an equiangular triangle’ differ in meaning. Likewise, intuitively, the belief that X is an equilateral triangle and the belief that X is an equiangular triangle are different beliefs. In both cases, the intuition rests on the fact that it was a Discovery that, necessarily, every equilateral triangle was equiangular, and conversely. But, on standard conceptions of possible worlds, ‘X is an equilateral triangle’ and ‘X is an equiangular triangle’ are true in exactly the same possible worlds precisely because they are logically equivalent.

There have been, broadly speaking, two reactions to this problem. One has been to bite the bullet and insist that the contents are the same. (The most worked out version of this response can be found in Stalnaker 1984.) The other has been to argue that, for the purposes of capturing content, a more fine-grained conception of the entities among which the division is being made is needed. For example, possible worlds might be thought of as constructions out of certain concepts so that, for example, the constructions at which ‘X is an equilateral triangle’ is true differ from those at which ‘X is an equiangular triangle’ is true by virtue of the former having the concept side, where the latter have the concept angle. (See Peacocke 1983, for an overview of the issues see Braddon-Mitchell and Jackson 1996.)

3.3 Mental Representation And The Language Of Thought

The functionalist answer to how a belief gets to have the belief content that it does—and, more generally, how a mental state gets to be the mental state that it is—was described earlier, and it was noted that there are many views about how to spell out the key functional roles. However, a popular answer holds that an important part of any functionalist account is the way belief and desire work together to produce behavior: roughly, subjects behave in a way such that what they desire is satisfied if what they believe is true (Stalnaker 1984, Braddon-Mitchell and Jackson 1996).

An alternative approach—the language of thought hypothesis—appeals to the idea that creatures capable of representational thought, creatures capable of having intentional mental states, have a language of thought. When subjects believe that snow is white, they have in their heads a sentence in ‘mentalese’ that says that snow is white. (This approach allows that dogs may have beliefs, because it is possible to have an internal language of thought, while lacking a public language in which to express thoughts.) This approach gives functional role a place in determining the difference, for example, between belief and desire, but content is treated via a ‘semantics for mentalese.’ Consider, for example, subjects who believe that snow is white. They have in their heads a token that stands for snow (the mentalese word for snow), a token that stands for whiteness (the mentalese predicate for whiteness), put together in a structure that makes up the mentalese sentence snow is white. What makes it true that they believe rather than desire that snow is white is the functional role this structure plays: it serves more of a ‘responding to the world’ role and less of a ‘changing the world’ role. (A classic source is Fodor 1988.)

This approach raises many important issues. One is how we should spell out the crucial relation between the words of mentalese and what they stand for? What makes it true that a structure in the brain is the mentalese for snow and not water, for example? One proffered answer is some kind of causal cum informational link between brain token and what is tokened, but the details are controversial. A second issue is whether neurological investigations might refute the language of thought hypothesis. Some cognitive scientists have argued that the way the brain represents the environment is more like the way maps represent than the way sentences represents (see Braddon-Mitchell and Jackson 1996).

4. Representation And Context

Representational content is often thought of as context dependent. For example, what a petrol gage represents about the level of gas in the tank depends on the nature of the connection between gage and tank. Connected in one of the more usual ways, the gage having its pointer in the far left position indicates that the tank is empty, but connected differently that pointer position might represent that the tank is full. That is to say, identical gage configurations can have very different representational contents depending on the gage’s ‘context.’ Or, as it is often put, representational content with respect to the amount of gas in the tank does not super ene on a petrol gage’s configuration. As it is said, the content is broad or wide or context dependent.

It is equally possible, however, to think of any given reading as saying: either the gage is connected to the tank in way W and the gas level is L , or the gage is connected in way W and the level is L , or … . In this way of thinking, a way naturally adopted when the manner of connection between gage and tank is unknown, the content is not dependent on the manner of connection.

In other words, and speaking more generally, there are two essentially equivalent ways of thinking of the interplay of representational content and context. In the first, it looks like this:

In C₁, S represents that P₁

…

In C_i, S represents that P_i

In the second, it looks like this:

S represents that if C₁ then P₁

…

S represents that if C_i then P_i

In the first way of thinking, content comes out as context dependent or broad; in the second, it comes out context independent or, as it is said, narrow.

A pressing, currently much debated question about both mental and linguistic content is whether we should think of it in the context dependent or in the context independent way. Or, equivalently, is what a person believes and what a person claims about how things are when they speak and write a broad or narrow property of them?

The traditional view is that mental and linguistic content is narrow. A principal source for this view is a powerful intuition often exploited by science fiction writers. The idea is that if the stimuli actually impinging on subjects via their environments is duplicated exactly, say, by a computer, the subjects will not be able to tell the difference. This is the core idea behind flight simulators and virtual reality machines, of course. Although these devices fall short of perfectly duplicating the impacts of their target environments on their subjects, their makers take it for granted that were perfect duplication achieved, subjects would be unable to tell whether, for example, they were really flying or were in a simulator. This suggests that subjects’ mental lives, including how their mental states represent things as being, supervene on their nature from the skin in (or, perhaps, from their central processors in). How their mental states represent things as being is, therefore, a narrow property of subjects: two subjects which are identical from the skin in are identical in how they take, desire etc. their worlds to be. Moreover, if the content of sentences is given by the content of the mental states they serve to express, it will follow that linguistic content is narrow.

However, many philosophers, often influenced by the Twin Earth example discussed earlier, repudiate this intuition. Consider, runs their argument, Mary and Mary’s duplicate from the skin in, Twin Mary, on Twin Earth. They both use the word ‘water’ for the clear, potable etc. stuff around them, but the conditions under which what is said is true, and equally those under which what is thought is true, differ. What Mary says and thinks when she uses the sentence ‘There is water nearby’ is true if and only if there is H O nearby, but what Twin Mary says and believes will be true if and only if there is XYZ nearby. Hence, runs the argument, the contents of the words and thoughts of Mary and Twin Mary differ; ergo, content is context dependent.

The issue is complicated, however, by the fact that it is plausible that there is an indexical element to the use of a great many words including ‘water.’ It is plausible that the word ‘water’ is used in part to mean the stuff that the users of the word have come across which is clear, potable etc. If this is right, a plausible explanation of why Mary and Twin Mary refer to different stuff in thought and talk when they use the word ‘water’ is that they and their language communities interact with different stuff when the word ‘water’ is used. That is, Mary refers to the clear, potable stuff that she and her community interact with, whereas Twin Mary refers to the stuff she and her (different) community interact with, and the reason they refer differently is that the stuff they interact with is different. This means that there is a sense in which what they say and think is the same when they say and think ‘There is water nearby.’ They both say and think that some of the clear potable etc. stuff that she herself interacts with is nearby. This does not affect the point that the truth conditions of what they say and think differ, but they differ in the same way that the truth conditions of ‘I am bald’ differ depending on who is speaking. There is a good sense in which each person who says or thinks ‘I am bald’ says or thinks the same thing, namely, that I myself am bald. (There is a large literature on this important question; see Pessin and Goldberg 1996 and Braddon-Mitchell and Jackson 1996.)

Bibliography:

1. Bigelow J, Pargetter R 1990 Science and Necessity. Cambridge University Press, Cambridge, UK
2. Braddon-Mitchell D, Jackson F 1996 The Philosophy of Mind and Cognition. Blackwell, Oxford, UK
3. Davidson D 1984 Inquiries into Truth and Interpretation. Oxford University Press, Oxford, UK
4. Devitt M 1996 Coming to our Senses: A Naturalistic Program for Semantic Localism. Cambridge University Press, Cambridge, UK
5. Devitt M, Sterelny K 1987 Language and Reality. MIT Press, Cambridge, MA
6. Dretske F 1995 Naturalizing the Mind. MIT Press, Cambridge, MA
7. Fodor J 1988 Psychosemantics. MIT Press, Cambridge, MA
8. Frege G 1952 Translations from the philosphical writing of
9. Gottlob Frege. In: Geach P, Black M (eds.) Translations from the Philosophical Writings of Gottlob Frege. Philosophical Library, New York
10. Jackson F 1998 Reference and description revisited. Philosophical Perspectives 12: 201–18
11. Kripke S 1980 Naming and Necessity. Harvard University Press, Cambridge, MA
12. McDowell J 1994 Mind and World. Harvard University Press, Cambridge MA
13. Peacocke C 1983 Sense and Content. Oxford University Press, Oxford, UK
14. Pessin A, Goldberg S (eds.) 1996 The Twin Earth Chronicles. M E Sharpe, Armonk, NY
15. Putnam H 1975 The meaning of ‘meaning’. In: Putnam H (ed.) Mind, Language and Reality. Cambridge University Press, Cambridge, UK
16. Stalnaker R 1984 Inquiry. MIT Press, Cambridge, MA