Reference is one of the most basic aspects of meaning. It is the way that words we utter point to things in the outside world. When I say That dog chased my cat, part of what is said is a reference to some particular flesh-and-blood dog, perhaps brown with floppy ears, named Fido, and some particular cat. If I said it again a week later I might now be referring to a different dog, Chopper, a huge hulking grey brute with tiny ears. To refer to something is to identify a word with some real individual. Words are about things. The thing referred to is called the referent (or sometimes also the reference).

The contrast is with the part of meaning that deals with our own concepts, or to other words. If I'm talking about democracy or wisdom I'm much less likely to be able to identify clear things in the world separate from my ideas about them. If I'm translating to or from French or German I'm likely to be content most of the time with treating one word as meaning another word, and not worrying about what outside objects they point to: dog means Hund and chien.

One very natural approach to the meaning of words and sentences is to see it as essentially a matter of working out the reference. If you know what thing I was referring to when I said dog, you will understand (most of) what I meant by saying it. In the philosophy of language, about half the approaches to meaning are referential, and the other half aren't, because of the problems with a naïve referential theory. Let's now examine those problems. Then we'll look at pieces of theory that various philosophers of language have used to approach these.

Non-existent things

We quite often talk about things that don't actually exist. We might think they do; or we might know they don't but want to say things about them anyway. In the first class are The Tooth Fairy gave me sixpence for my tooth and Phlogiston causes combustion.

The second class includes various ways we might want to talk about non-existent things. First we might want to ignore their non-existence: Sherlock Holmes's landlady was called Mrs Hudson. We don't need to scrupulously mention that she wasn't really because it's all fictional. In fact, it's perfectly true that she was called Mrs Hudson, and false that she was called Mrs Moriarty. This hardly seems to stretch 'true' and 'false' unacceptably. Next we want to talk about what other people thought: In the eighteenth century chemists believed phlogiston caused combustion. This is quite true: we don't have to get into knots about trying to deny that we believe in phlogiston. Finally we might want to come straight to the point about non-existent things and make perfectly true statements such as The Tooth Fairy doesn't exist.

If meaning is essentially reference, or if (a different thing) reference is essential to meaning, these all suffer from gaping holes in the sentence. What are you talking about if one of the things you're trying to point to doesn't exist? How can the whole sentence have a meaning if part of the meaning is absent? You clearly can't be using an ordinary reference: See that thing over there? That Easter Bunny? Well it doesn't exist.

Different names for the same thing

Suppose what you're referring to does exist. Then it's natural to suppose that you're just talking about it, that very thing. Anything you say about it that's true is true if it's true about that thing. And whatever you say means just what it means as long as you're talking about that thing. These sound blindingly obvious until you find cases where they don't work. What you mean by something, and whether it's true, can both be different depending not only on what you're talking about, but on how you're talking about it. Let's unpack these claims.

George Orwell was a famous writer: you've definitely heard of him. You might or might not know that his real name was Eric Blair. If you didn't, you might disagree if someone said Eric Blair was a famous writer: someone you've never heard of can't be all that famous. So the two sentences I think X was a famous writer can be true or false depending on what name goes in X, and it can be true for some Y and false for some Z even if Y and Z refer to the same person. So truth depends not just on reference but on how the referent is referred to.

The problem here is with propositional attitude verbs, like think and believe and assume. We can believe false things. So the contents of our thoughts have one shape, the real world another. Propositional (or intensional, sic, not the same as 'intentional') contexts don't allow equivalents to be substituted inside them while preserving truth (salva veritate).

Another distinction needs to be drawn here. If you believe Orwell was a great writer, there is one sense in which you also believe Blair was, even if you don't know it, or wouldn't say so. This is called the de re ('about the thing') reading: it's about the referent. It contrasts with the de dicto ('about the saying') reading, where you would say it about Orwell but not about Blair.

Another thing you can do with different names is say that they mean the same thing. You can tell someone quite a significant fact by saying Eric Blair is George Orwell. But telling them Eric Blair is Eric Blair doesn't seem to mean nearly as much.

Things that aren't things

The notion of reference works best for goodish medium-sized namable objects. We can point to that and say that's a marble, that's my cat, that's Madonna, that's the Eiffel Tower, and so on. What about the rest of the sentence: what about red and run, not to mention if and and?

One possible answer is a distinction between singular items and classes. My cat and Madonna are individuals who can be picked out, but other words refer to whole classes: the class of marbles, the class of red things, the class of things that are running. A sentence is then a predication, saying of one thing that it belongs in a certain class. At least for some sentences a simple form of semantics can be worked out in which the meaning of a general term is a class.

We also get a grammatical distinction between a noun phrase such as cat or black cat, which denotes a class, and my cat or that black cat, where the determiner (my, that) is a word that has made the expression referential.

The details of all this have been handled differently by various philosophers.

Mill: denotation

John Stuart Mill made a fairly obvious distinction in the mid nineteenth century, between the denotation of a word, which is the thing out in the world that it points to, and the connotation, or meaning, or the property it states that allows us to ascertain that the thing fits the word.

Frege: sense

Gottlob Frege near the end of the nineteenth century made a distinction between Sinn and Bedeutung, which are nowadays always translated as sense and reference (though Russell used meaning and denotation). The reference is the thing out in the world picked out by the object, while the sense is the way we are aware of it. We can have multiple senses for the same reference. The classic example due to Frege is that of the Morning Star and the Evening Star. It is a non-trivial fact requiring some astronomical knowledge that the bright star we see low on the horizon in the morning is in fact the same object as the bright star on the evening horizon, and both of these are the planet Venus. We know about them under three senses, and can think different things about them, even though in fact (even if we don't know it) they refer to the same thing.

Frege's distinction remains to the present day, though others have built on it. He did not regard senses as mental objects, but modern cognitivist philosophers are likely to take a similar interpretation but in which the sense is an internal (mental) representation, and sense is almost like reference to this concept.

Russell: definite description

Bertrand Russell, building on Frege's work, developed a narrow view of what could genuinely refer. He thought the only bits of language that could refer directly to the world were a few indexicals like I and this. He called these 'logically proper names'.

Everything else that seems to refer, he preferred to regard as denoting. This is different from Mill's sense of denotation, and more like Mill's connotation. Most expressions that single out an object in the world do so by being a description, or an abbreviation of a hidden description. What we ordinarily call a proper name actually contains a description we're implicitly thinking of when we use it: so the name Napoleon Bonaparte might mean the Emperor of the French who was defeated at Waterloo.

This assimilates most proper names to what Russell called a definite description: a phrase beginning with 'the' and uniquely describing one individual, so that it picks out its referent in the world. He analyses 'the Emperor of the French was defeated at Waterloo' as a union of two kinds of claim: first that there is one and only one person who is Emperor of the French, and second that whoever is Emperor of the French was defeated at Waterloo.

Kripke: rigid designator

Saul Kripke in the 1960s and 1970s developed a highly influential theory of names, by which something got a name at some point in its early history (its 'baptism'), and this name is carried through its history by people who get the name from other people who know it. So individuals can be identified even if they would have quite different histories in other possible worlds. Some baby got given the name Richard Nixon, and whoever that child grew up into, whether he entered politics or became a used-car salesperson, he would still have been Richard Nixon. Kripke said proper names were 'rigid designators'. So you can say Richard Nixon might not have been president of the USA in 1970, whereas you can't say he might not have been Richard Nixon, or that the president might not have been the president. A reference to Nixon is always a reference to that person, not to some other person who might have had some similar properties if things had been different.

One big flaw with associating descriptions with individuals, as Russell did, is that you might be wrong. You might think of Henry Purcell as the man who wrote the Trumpet Voluntary, so if you subsequently read that actually Jeremiah Clarke did, did you all the time mean Clarke when you said Purcell? This seems absurd, and Kripke's solution is to show how 'Purcell' always refers to Purcell even if everything you knew about him was wrong.

Donnellan: attributive

Keith Donnellan is the most recent contributor I'll consider, though the nature of reference is still very much an actively debated idea today. Donnellan said that definite descriptions sometimes worked the way Russell said, and he called these attributive. In saying 'the X' you're attributing some property to something, and this can be used to find the referent, which is whatever in fact satisfies that property.

However, other uses of definite descriptions, or what in Russell's terms are logically equivalent, such as ordinary proper names, are directly referential. You're using the description to pick out the person or thing pragmatically, but what you actual mean is the very person or thing you're referring to, even if you got the description wrong. This sounds like Kripke's approach, but Kripke is interested in the long-term history of names, whereas Donnellan is looking at one-off uses.

If you see John's brother at a party, and he's sipping a drink from a martini glass, you can point him out to someone by saying The man drinking a martini is John's brother. By this you are referring to him, over there: you're not making an attributive claim that whoever is drinking a martini is John's brother. It might be that he's actually drinking water, and that moreover someone behind you that you haven't seen really is the only person in the room drinking a martini. Nevertheless, the man drinking a martini in this case doesn't actually denote the man drinking a martini, because you didn't use it for denoting, you used it for directly referring to him there.