Reference and Definite Descriptions
Keith S. Donnellan
First appeared in The Philosophical Review 77 (1966): 281-304. In the public domain.
Reprinted in Ludlow, ed., Readings in the Philosophy of Language. Page numbers from this edition.
Definite descriptions, I shall argue, have two possible functions.  They are used to refer to what a speaker wishes to talk about, but they are also used quite differently. Moreover, a definite description occurring in one and the same sentence may, on different occasions of its use, function in either way. The failure to deal with this duality of function obscures the genuine referring use of definite descriptions. The best-known theories of definite descriptions, those of Russell and Strawson, I shall suggest, are both guilty of this. Before discussing this distinction in use, I will mention some features of these theories to which it is especially relevant.
On Russell's view a definite description may denote an entity: "if 'C' is a denoting phrase [as definite descriptions are by definition], it may happen that there is one entity x (there cannot be more than one) for which the proposition "x is identical with C" is true ... We may then say that the entity x is the denotation of the phrase 'C.' "  In using a definite description, then, a speaker may use an expression which denotes some entity, but this is the only relationship between that entity and the use of the definite description recognized by Russell. I shall argue, however, that there are two uses of definite descriptions. The definition of denotation given by Russell is applicable to both, but in one of these the definite description serves to do something more. I shall say that in this use the speaker uses the definite description to refer to something, and call this use the "referential use" of a definite description. Thus, if I am right, referring is not the same as denoting and the referential use of definite descriptions is not recognized on Russell's view.
 I should like to thank my colleagues, John Canfield, Sydney Shoemaker, and Timothy Smiley, who read an earlier draft and gave me helpful suggestions. I also had the benefit of the valuable and detailed comments of the referee for the paper, to whom I wish to express my gratitude.
 "On Denoting," reprinted in Logic and Knowledge, ed. by Robert C. Marsh (London, 1956), p. 51.
Furthermore, on Russell's view the type of expression that comes closest to performing the function of the referential use of definite descriptions turns out, as one might suspect, to be a proper name (in "the narrow logical sense"). Many of the things said about proper names by Russell can, I think, be said about the referential use of definite descriptions without straining senses unduly. Thus the gulf Russell thought he saw between names and definite descriptions is narrower than he thought.
Strawson, on the other hand, certainly does recognize a referential use of definite definitions. But what I think he did not see is that a definite description may have a quite different role -- may be used non-referentially, even as it occurs in one and the same sentence. Strawson, it is true, points out nonreferential uses of definite descriptions, but which use a definite description has seems to be for him a function of the kind of sentence in which it occurs; whereas, if I am right, there can be two possible uses of a definite description in the same sentence. Thus, in "On Referring," he says, speaking of expressions used to refer, "Any expression of any of these classes [one being that of definite descriptions] can occur as the subject of what would traditionally be regarded as a singular subject-predicate sentence; and would, so occurring, exemplify the use I wish to discuss."  So the definite description in, say, the sentence "The Republican candidate for president in 1968 will be a conservative" presumably exemplifies the referential use. But if I am right, we could not say this of the sentence in isolation from some particular occasion on which it is used to state something; and then it might or might not turn out that the definite description has a referential use.
Strawson and Russell seem to me to make a common assumption here about the question of how definite descriptions function: that we can ask how a definite description functions in some sentence independently of a particular occasion upon which it is used. This assumption is not really rejected in Strawson's arguments against Russell. Although he can sum up his position by saying, " 'Mentioning' or 'referring' is not something an expression does; it is something that someone can use an expression to do,"  he means by this to deny the radical view that a "genuine" referring expression has a referent, functions to refer, independent of the context of some use of the expression. The denial of this view, however, does not entail that definite descriptions cannot be identified as referring expressions in a sentence unless the sentence is being used. Just as we can speak of a function of a tool that is not at the moment performing its function, Strawson's view, I believe, allows us to speak of the referential function of
 "On Referring," reprinted in Philosophy and Ordinary Language, ed. by Charles C. Caton (Urbana, 1963), pp. 162-163. [Chap. 16 in this volume.]
 Ibid., p. 162.
 Ibid., p. 130.
a definite description in a sentence even when it is not being used. This, I hope to show, is a mistake.
A second assumption shared by Russell's and Strawson's account of definite descriptions is this. In many cases a person who uses a definite description can be said (in some sense) to presuppose or imply that something fits the description.  If I state that the king is on his throne, I presuppose or imply that there is a king. (At any rate, this would be a natural thing to say for anyone who doubted that there is a king.) Both Russell and Strawson assume that where the presupposition or implication is false, the truth value of what the speaker says is affected. For Russell the statement made is false; for Strawson it has no truth value. Now if there are two uses of definite descriptions, it may be that the truth value is affected differently in each case by the falsity of the presupposition or implication. This is what I shall in fact argue. It will turn out, I believe, that one or the other of the two views, Russell's or Strawson's, may be correct about the nonreferential use of definite descriptions, but neither fits the referential use. This is not so surprising about Russell's view, since he did not recognize this use in any case, but it is surprising about Strawson's since the referential use is what he tries to explain and defend. Furthermore, on Strawson's account, the result of there being nothing which fits the description is a failure of reference. This too, I believe, turns out not to be true about the referential use of definite descriptions.
There are some uses of definite descriptions which carry neither any hint of a referential use nor any presupposition or implication that something fits the description. In general, it seems, these are recognizable from the sentence frame in which the description occurs. These uses will not interest us, but it is necessary to point them out if only to set them aside.
An obvious example would be the sentence "The present king of France does not exist," used, say, to correct someone's mistaken impression that de Gaulle is the king of France.
A more interesting example is this. Suppose someone were to ask, "Is de Gaulle the king of France?" This is the natural form of words for a person to use who is in doubt as to whether de Gaulle is king or president of France. Given this background to the question, there seems to be no presupposition or implication that someone is the king of France. Nor is
 Here and elsewhere I use the disjunction "presuppose or imply" to avoid taking a stand that would side me with Russell or Strawson on the issue of what the relationship involved is. To take a stand here would be beside my main point as well as being misleading, since later on I shall argue that the presupposition or implication arises in a different way depending upon the use to which the definite description is put. This last also accounts for my use of the vagueness indicator, "in some sense.~
 In a footnote added to the original version of "On Referring" (op. cit., p. 181) Strawson seems to imply that where the presupposition is false, we still succeed in referring in a "secondary" way, which seems to mean "as we could be said to refer to fictional or make-believe things." But his view is still that we cannot refer in such a case in the "primary" way. This is, I believe, wrong. For a discussion of this modification of Strawson's view see Charles C. Caton, "Strawson on Referring," Mind, LXVIII (1959), 539-544.
the person attempting to refer to someone by using the definite description. On the other hand, reverse the name and description in the question and the speaker probably would be thought to presuppose or imply this. "Is the king of France de Gaulle?" is the natural question for one to ask who wonders whether it is de Gaulle rather than someone else who occupies the throne of France. 
Many times, however, the use of a definite description does carry a presupposition or implication that something fits the description. If definite descriptions do have a referring role, it will be here. But it is a mistake, I think, to try, as I believe both Russell and Strawson do, to settle this matter without further ado. What is needed, I believe, is the distinction I will now discuss.
I will call the two uses of definite descriptions I have in mind the attributive use and the referential use. A speaker who uses a definite description attributively in an assertion states something about whoever or whatever is the so-and-so. A speaker who uses a definite description referentially in an assertion, on the other hand, uses the description to enable his audience to pick out whom or what he is talking about and states something about that person or thing. In the first case the definite description might be said to occur essentially, for the speaker wishes to assert something about whatever or whoever fits that description; but in the referential use the definite description is merely one tool for doing a certain job -- calling attention to a person or thing -- and in general any other device for doing the same job, another description or a name, would do as well. In the attributive use, the attribute of being the so-and-so is all important, while it is not in the referential use.
To illustrate this distinction, in the case of a single sentence, consider the sentence, "Smith's murderer is insane." Suppose first that we come upon poor Smith foully murdered. From the brutal manner of the killing and the fact that Smith was the most lovable person in the world, we might exclaim, "Smith's murderer is insane." I will assume, to make it a simpler case, that in a quite ordinary sense we do not know who murdered Smith (though this is not in the end essential to the case). This, I shall say, is an attributive use of the definite description.
The contrast with such a use of the sentence is one of those situations in which we expect and intend our audience to realize whom we have in
 This is an adaptation of an example (used for a somewhat different purpose) given by Leonard Linsky in "Reference and Referents," in Philosophy and Ordinary Language, p. 80.
mind when we speak of Smith's murderer and, most importantly, to know that it is this person about whom we are going to say something.
For example, suppose that Jones has been charged with Smith's murder and has been placed on trial. Imagine that there is a discussion of Jones's odd behavior at his trial. We might sum up our impression of his behavior by saying, "Smith's murderer is insane." If someone asks to whom we are referring, by using this description, the answer here is "Jones." This, I shall say, is a referential use of the definite description.
That these two uses of the definite description in the same sentence are really quite different can perhaps best be brought out by considering the consequences of the assumption that Smith had no murderer (for example, he in fact committed suicide). In both situations, in using the definite description "Smith's murderer," the speaker in some sense presupposes or implies that there is a murderer. But when we hypothesize that the presupposition or implication is false, there are different results for the two uses. In both cases we have used the predicate "is insane," but in the first case, if there is no murderer, there is no person of whom it could be correctly said that we attributed insanity to him. Such a person could be identified (correctly) only in case someone fitted the description used. But in the second case, where the definite description is simply a means of identifying the person we want to talk about, it is quite possible for the correct identification to be made even though no one fits the description we used.  We were speaking about Jones even though he is not in fact Smith's murderer and, in the circumstances imagined, it was his behavior we were commenting upon. Jones might, for example, accuse us of saying false things of him in calling him insane and it would be no defense, I should think, that our description, "the murderer of Smith," failed to fit him.
It is, moreover, perfectly possible for our audience to know to whom we refer, in the second situation, even though they do not share our presupposition. A person hearing our comment in the context imagined might know we are talking about Jones even though he does not think Jones guilty.
Generalizing from this case, we can say, I think, that there are two uses of sentences of the form, "The j is y ." In the first, if nothing is the j then nothing has been said to be y . In the second, the fact that nothing is the j does not have this consequence.
With suitable changes the same difference in use can be formulated for uses of language other than assertions. Suppose one is at a party and,
 In "Reference and Referents" (pp. 74-75, 80), Linsky correctly points out that one does not fail to refer simply because the description used does not in fact fit anything (or fits more than one thing). Thus he pinpoints one of the difficulties in Strawsons view. Here, however, I use this fact about referring to make a distinction I believe he does not draw, between two uses of definite descriptions. I later discuss the second passage from Linskys paper.
seeing an interesting-looking person holding a martini glass, one asks, "Who is the man drinking a martini?" If it should turn out that there is only water in the glass, one has nevertheless asked a question about a particular person, a question that it is possible for someone to answer. Contrast this with the use of the same question by the chairman of the local Teetotalers Union. He has just been informed that a man is drinking a martini at their annual party. He responds by asking his informant, "Who is the man drinking a martini?" In asking the question the chairman does mot have some particular person in mind about whom he asks the question; if no one is drinking a martini, if the information is wrong, no person can be singled out as the person about whom the question was asked. Unlike the first case, the attribute of being the man drinking a martini is all-important, because if it is the attribute of no one, the chairman's question has no straightforward answer.
This illustrates also another difference between the referential and the attributive use of definite descriptions. In the one case we have asked a question about a particular person or thing even though nothing fits the description we used; in the other this is not so. But also in the one case our question can be answered; in the other it cannot be. In the referential use of a definite description we may succeed in picking out a person or thing to ask a question about even though he or it does not really fit the description; but in the attributive use if nothing fits the description, no straightforward answer to the question can be given.
This further difference is also illustrated by commands or orders containing definite descriptions. Consider the order, "Bring me the book on the table." If "the book on the table" is being used referentially, it is possible to fulfill the order even though there is no book on the table. If, for example, there is a book beside the table, though there is none on it, one might bring that book back and ask the issuer of the order whether this is "the book you meant." And it may be. But imagine we are told that someone has laid a book on our prize antique table, where nothing should be put. The order, "Bring me the book on the table" cannot now be obeyed unless there is a book that has been placed on the table. There is no possibility of bringing back a book which was never on the table and having it be the one that was meant, because there is no book that in the sense was "meant." In the one case the definite description was a device for getting the other person to pick the right book; if he is able to pick the right book even though it does not satisfy the description, one still succeeds in his purpose. In the other case, there is, antecedently, no "right
book" except one which fits the description; the attribute of being the book on the table is essential. Not only is there no book about which an order was issued, if there is no book on the table, but the order itself cannot be obeyed. When a definite description is used attributively in a command or question and nothing fits the description, the command cannot be obeyed and the question cannot be answered. This suggests some analogous consequence for assertions containing definite descriptions used attributively. Perhaps the analogous result is that the assertion is neither true nor false: this is Strawson's view of what happens when the presupposition of the use of a definite description is false. But if so, Strawson's view works not for definite descriptions used referentially, but for the quite different use, which I have called the attributive use.
I have tried to bring out the two uses of definite descriptions by pointing out the different consequences of supposing that nothing fits the description used. There are still other differences. One is this: when a definite description is used referentially, not only is there in some sense a presupposition or implication that someone or something fits the description, as there is also in the attributive use, but there is a quite different presupposition; the speaker presupposes of some particular someone or something that he or it fits the description. In asking, for example, "Who is the man drinking a martini?" where we mean to ask a question about that man over there, we are presupposing that that man over there is drinking a martini -- not just that someone is a man drinking a martini. When we say, in a context where it is clear we are referring to Jones, "Smith's murderer is insane," we are presupposing that Jones is Smith's murderer. No such presupposition is present in the attributive use of definite descriptions. There is, of course, the presupposition that someone or other did the murder, but the speaker does not presuppose of someone in particular -- Jones or Robinson, say -- that he did it. What I mean by this second kind of presupposition that someone or something in particular fits the description -- which is present in a referential use but not in an attributive use -- can perhaps be seen more clearly by considering a member of the speaker's audience who believes that Smith was not murdered at all. Now in the case of the referential use of the description, "Smith's murderer," he could accuse the speaker of mistakenly presupposing both that someone or other is the murderer and that also Jones is the murderer, for even though he believes Jones not to have done the deed, he knows that the speaker was referring to Jones. But in the case of the attributive use, he can accuse the speaker of having only the first, less
specific presupposition; he cannot pick out some person and claim that the speaker is presupposing that that person is Smith's murderer. Now the more particular presuppositions that we find present in referential uses are clearly not ones we can assign to a definite description in some particular sentence in isolation from a context of use. In order to know that a person presupposes that Jones is Smith's murderer in using the sentence "Smith's murderer is insane," we have to know that he is using the description referentially and also to whom he is referring. The sentence by itself does not tell us any of this.
From the way in which I set up each of the previous examples it might be supposed that the important difference between the referential and the attributive use lies in the beliefs of the speaker. Does he believe of some particular person or thing that he or it fits the description used? In the Smith murder example, for instance, there was in the one case no belief as to who did the deed, whereas in the contrasting case it was believed that Jones did it. But this is, in fact, not an essential difference. It is possible for a definite description to be used attributively even though the speaker (and his audience) believes that a certain person or thing fits the description. And it is possible for a definite description to be used referentially where the speaker believes that nothing fits the description. It is true -- and this is why, for simplicity, I set up the examples the way I did -- that if a speaker does not believe that anything fits the description or does not believe that he is in a position to pick out what does fit the description, it is likely that he is not using it referentially. It is also true that if he and his audience would pick out some particular thing or person as fitting the description, then a use of the definite description is very likely referential. But these are only presumptions and not entailments.
To use the Smith murder case again, suppose that Jones is on trial for the murder and I and everyone else believe him guilty. Suppose that I comment that the murderer of Smith is insane, but instead of backing this up, as in the example previously used, by citing Jones's behavior in the dock, I go on to outline reasons for thinking that anyone who murdered poor Smith in that particularly horrible way must be insane. If now it turns out that Jones was not the murderer after all, but someone else was, I think I can claim to have been right if the true murderer is after
all insane. Here, I think, I would be using the definite description attributively, even though I believe that a particular person fits the description.
It is also possible to think of cases in which the speaker does not believe that what he means to refer to by using the definite description fits the description, or to imagine cases in which the definite description is used referentially even though the speaker believes nothing fits the description. Admittedly, these cases may be parasitic on a more normal use; nevertheless, they are sufficient to show that such beliefs of the speaker are not decisive as to which use is made of a definite description.
Suppose the throne is occupied by a man I firmly believe to be not the king, but a usurper. Imagine also that his followers as firmly believe that he is the king. Suppose I wish to see this man. I might say to his minions, "Is the king in his courtinghouse?" I succeed in referring to the man I wish to refer to without myself believing that he fits the description. It is not even necessary, moreover, to suppose that his followers believe him to be the king. If they are cynical about the whole thing, know he is not the king, I may still succeed in referring to the man I wish to refer to. Similarly, neither I nor the people I speak to may suppose that anyone is the king and, finally, each party may know that the other does not so suppose and yet the reference may go through.
Both the attributive and the referential use of definite descriptions seem to carry a presupposition or implication that there is something which fits the description. But the reasons for the existence of the presupposition or implication are different in the two cases.
There is a presumption that a person who uses a definite description referentially believes that what he wishes to refer to fits the description. Because the purpose of using the description is to get the audience to pick out or think of the right thing or person, one would normally choose a description that he believes the thing or person fits. Normally a misdescription of that to which one wants to refer would mislead the audience. Hence, there is a presumption that the speaker believes something fits the description -- namely, that to which he refers.
When a definite description is used attributively, however, there is not the same possibility of misdescription. In the example of "Smith's murderer" used attributively, there was not the possibility of misdescribing Jones or anyone else; we were not referring to Jones nor to anyone else by
using the description. The presumption that the speaker believes someone is Smith's murderer does not arise here from a more specific presumption that he believes Jones or Robinson or someone else whom he can name or identify is Smith's murderer.
The presupposition or implication is borne by a definite description used attributively because if nothing fits the description the linguistic purpose of the speech act will be thwarted. That is, the speaker will not succeed in saying something true, if he makes an assertion; he will not succeed in asking a question that can be answered, if he has asked a question; he will not succeed in issuing an order that can be obeyed, if he has issued an order. If one states that Smith's murderer is insane, when Smith has no murderer, and uses the definite description nonreferentially, then one fails to say anything true. If one issues the order "Bring me Smith's murderer" under similar circumstances, the order cannot be obeyed; nothing would count as obeying it.
When the definite description is used referentially, on the other hand, the presupposition or implication stems simply from the fact that normally a person tries to describe correctly what he wants to refer to because normally this is the best way to get his audience to recognize what he is referring to. As we have seen, it is possible for the linguistic purpose of the speech act to be accomplished in such a case even though nothing fits the description; it is possible to say something true or to ask a question that gets answered or to issue a command that gets obeyed. For when the definite description is used referentially, one's audience may succeed in seeing to what one refers even though neither it nor anything else fits the description.
The result of the last section shows something to be wrong with the theories of both Russell and Strawson; for though they give differing accounts of the implication or presupposition involved, each gives only one. Yet, as I have argued, the presupposition or implication is present for a quite different reason, depending upon whether the definite description is used attributively or referentially, and exactly what presuppositions or implications are involved is also different. Moreover, neither theory seems a correct characterization of the referential use. On Russell's there is a logical entailment: "The j is y " entails "There exists one and only one j ." Whether or not this is so for the attributive use, it does not seem true
of the referential use of the definite description. The "implication" that something is the j , as I have argued, does not amount to an entailment; it is more like a presumption based on what is usually true of the use of a definite description to refer. In any case, of course, Russell's theory does not show -- what is true of the referential use -- that the implication that something is thej comes from the more specific implication that what is being referred to is the j . Hence, as a theory of definite descriptions, Russell's view seems to apply, if at all, to the attributive use only.
Russell's definition of denoting (a definite description denotes an entity if that entity fits the description uniquely) is clearly applicable to either use of definite descriptions. Thus whether or not a definite description is used referentially or attributively, it may have a denotation. Hence, denoting and referring, as I have explicated the latter notion, are distinct and Russell's view recognizes only the former. It seems to me, moreover, that this is a welcome result, that denoting and referring should not be confused. If one tried to maintain that they are the same notion, one result would be that a speaker might be referring to something without knowing it. If someone said, for example, in 1960 before he had any idea that Mr. Goldwater would be the Republican nominee in 1964, "The Republican candidate for president in 1964 will be a conservative," (perhaps on the basis of an analysis of the views of party leaders) the definite description here would denote Mr. Goldwater. But would we wish to say that the speaker had referred to, mentioned, or talked about Mr. Goldwater? I feel these terms would be out of place. Yet if we identify referring and denoting, it ought to be possible for it to turn out (after the Republican Convention) that the speaker had, unknown to himself, referred in 1960 to Mr. Goldwater. On my view, however, while the definite description used did denote Mr. Goldwater (using Russell's definition), the speaker used it attributively and did not refer to Mr. Goldwater.
Turning to Strawson's theory, it was supposed to demonstrate how definite descriptions are referential. But it goes too far in this direction. For there are nonreferential uses of definite descriptions also, even as they occur in one and the same sentence. I believe that Strawson's theory involves the following propositions:
(1) If someone asserts that the j is y he has not made a true or false statement if there is no j . 
(2) If there is no j then the speaker has failed to refer to anything. 
 In "A Reply to Mr. Sellars," Philosophical Review, LXIII (1954), 216-231, Strawson admits that we do not always refuse to ascribe truth to what a person says when the definite description he uses fails to fit anything (or fits more than one thing). To cite one of his examples, a person who said, "The United States Chamber of Deputies contains representatives of two major parties," would be allowed to have said something true even though he had used the wrong title. Strawson thinks this does not constitute a genuine problem for his view. He thinks that what we do in such cases, "where the speakers intended reference is pretty clear, is simply to amend his statement in accordance with his guessed intentions and assess the amended statement for truth or falsity; we are not awarding a truth value at all to the original statement" (p. 230).
The notion of an "amended statement," however, will not do. We may note, first of all, that the sort of case Strawson has in mind could arise only when a definite description is used referentially. For the "amendment" is made by seeing the speaker's intended reference. But this could happen only if the speaker had an intended reference, a particular person or thing in mind, independent of the description he used. The cases Strawson has in mind are presumably not cases of slips of the tongue or the like; presumably they are cases in which a definite description is used because the speaker believes, though he is mistaken, that he is
describing correctly what he wants to refer to. We supposedly amend the statement by knowing to what he intends to refer. But what description is to be used in the amended statement? In the example, perhaps, we could use "the United States Congress." But this description might be one the speaker would not even accept as correctly describing what he wants to refer to, because he is misinformed about the correct title. Hence, this is not a case of deciding what the speaker meant to say as opposed to what he in fact said, for the speaker did not mean to say "the United States Congress." If this is so, then there is no bar to the "amended" statement containing any description that does correctly pick out what the speaker intended to refer to. It could be, e.g., "The lower house of the United States Congress." But this means that there is no one unique "amended" statement to be assessed for truth value. And, in fact, it should now be clear that the notion of the amended statement really plays no role anyway. For if we can arrive at the amended statement only by first knowing to what the speaker intended to refer, we can assess the truth of what he said simply by deciding whether what he intended to refer to has the properties he ascribed to it.
 As noted earlier (n. 7), Strawson may allow that one has possibly referred in a "secondary" way, but, if I am right, the fact that there is no j does not preclude one from having referred in the same way one does if there is a j .
(3) The reason he has said nothing true or false is that he has failed to refer.
Each of these propositions is either false or, at best, applies to only one of the two uses of definite descriptions.
Proposition (1) is possibly true of the attributive use. In the example in which "Smith's murderer is insane" was said when Smith's body was first discovered, an attributive use of the definite description, there was no person to whom the speaker referred. If Smith had no murderer, nothing true was said. It is quite tempting to conclude, following Strawson, that nothing true or false was said. But where the definite description is used referentially, something true may well have been said. It is possible that something true was said of the person or thing referred to. 
Proposition (2) is, as we have seen, simply false. Where a definite description is used referentially it is perfectly possible to refer to something though nothing fits the description used.
The situation with proposition (3) is a bit more complicated. It ties together, on Strawson's view, the two strands given in (1) and (2). As an account of why, when the presupposition is false, nothing true or false has been stated, it clearly cannot work for the attributive use of definite descriptions, for the reason it supplies is that reference has failed. It does not then give the reason why, if indeed this is so, a speaker using a definite description attributively fails to say anything true or false if nothing fits the description. It does, however, raise a question about the referential use. Can reference fail when a definite description is used referentially?
I do not fail to refer merely because my audience does not correctly pick out what I am referring to. I can be referring to a particular man when I use the description "the man drinking a martini," even though the people to whom I speak fail to pick out the right person or any person at all. Nor, as we have stressed, do I fail to refer when nothing fits the description. But perhaps I fail to refer in some extreme circumstances, when there is nothing that I am willing to pick out as that to which I referred.
Suppose that I think I see at some distance a man walking and ask, "Is the man carrying a walking stick the professor of history?" We should perhaps distinguish four cases at this point. (a) There is a man carrying a walking stick; I have then referred to a person and asked a question about him that can be answered if my audience has the information. (b) The man over there is not carrying a walking stick, but an umbrella; I have
 For a further discussion of the notion of saying something true of someone or something, see sec. VIII.
still referred to someone and asked a question that can be answered, though if my audience sees that it is an umbrella and not a walking stick, they may also correct my apparently mistaken impression. (c) It is not a man at all, but a rock that looks like one; in this case, I think I still have referred to something, to the thing over there that happens to be a rock but that I took to be a man. But in this case it is not clear that my question can be answered correctly. This, I think, is not because I have failed to refer, but rather because, given the true nature of what I referred to, my question is not appropriate. A simple "No, that is not the professor of history" is at least a bit misleading if said by someone who realizes that I mistook a rock for a person. It may, therefore, be plausible to conclude that in such a case I have not asked a question to which there is a straightforwardly correct answer. But if this is true, it is not because nothing fits the description I used, but rather because what I referred to is a rock and my question has no correct answer when asked of a rock. (d) There is finally the case in which there is nothing at all where I thought there was a man with a walking stick; and perhaps here we have a genuine failure to refer at all, even though the description was used for the purpose of referring. There is no rock, nor anything else, to which I meant to refer; it was, perhaps, a trick of light that made me think there was a man there. I cannot say of anything, "That is what I was referring to, though I now see that it's not a man carrying a walking stick." This failure of reference, however, requires circumstances much more radical than the mere nonexistence of anything fitting the description used. It requires that there be nothing of which it can be said, "That is what he was referring to." Now perhaps also in such cases, if the speaker has asserted something, he fails to state anything true or false if there is nothing that can be identified as that to which he referred. But if so, the failure of reference and truth value does not come about merely because nothing fits the description he used. So (3) may be true of some cases of the referential use of definite descriptions; it may be true that a failure of reference results in a lack of truth value. But these cases are of a much more extreme sort than Strawson's theory implies.
I conclude, then, that neither Russell's nor Strawson's theory represents a correct account of the use of definite descriptions -- Russell's because it ignores altogether the referential use, Strawson's because it fails to make the distinction between the referential and the attributive and mixes together truths about each (together with some things that are false).
It does not seem possible to say categorically of a definite description in a particular sentence that it is a referring expression (of course, one could say this if he meant that it might be used to refer). In general, whether or not a definite description is used referentially or attributively is a function of the speaker's intentions in a particular case. "The murderer of Smith" may be used either way in the sentence "The murderer of Smith is insane." It does not appear plausible to account for this, either, as an ambiguity in the sentence. The grammatical structure of the sentence seems to me to be the same whether the description is used referentially or attributively: that is, it is not syntactically ambiguous. Nor does it seem at all attractive to suppose an ambiguity in the meaning of the words; it does not appear to be semantically ambiguous. (Perhaps we could say that the sentence is pragmatically ambiguous: the distinction between roles that the description plays is a function of the speaker's intentions.) These, of course, are intuitions; I do not have an argument for these conclusions. Nevertheless, the burden of proof is surely on the other side.
This, I think, means that the view, for example, that sentences can be divided up into predicates, logical operators, and referring expressions is not generally true. In the case of definite descriptions one cannot always assign the referential function in isolation from a particular occasion on which it is used.
There may be sentences in which a definite description can be used only attributively or only referentially. A sentence in which it seems that the definite description could be used only attributively would be "Point out the man who is drinking my martini." I am not so certain that any can be found in which the definite description can be used only referentially. Even if there are such sentences, it does not spoil the point that there are many sentences, apparently not ambiguous either syntactically or semantically, containing definite descriptions that can be used either way.
If it could be shown that the dual use of definite descriptions can be accounted for by the presence of an ambiguity, there is still a point to be made against the theories of Strawson and Russell. For neither, so far as I can see, has anything to say about the possibility of such an ambiguity and, in fact, neither seems compatible with such a possibility. Russell's does not recognize the possibility of the referring use, and Strawson's, as I have tried to show in the last section, combines elements from each use
into one unitary account. Thus the view that there is an ambiguity in such sentences does not seem any more attractive to these positions.
Using a definite description referentially, a speaker may say something true even though the description correctly applies to nothing. The sense in which he may say something true is the sense in which he may say something true about someone or something. This sense is, I think, an interesting one that needs investigation. Isolating it is one of the by-products of the distinction between the attributive and referential uses of definite descriptions.
For one thing, it raises questions about the notion of a statement. This is brought out by considering a passage in a paper by Leonard Linsky in which he rightly makes the point that one can refer to someone although the definite description used does not correctly describe the person:
... said of a spinster that "Her husband is kind to her" is neither true nor false. But a speaker might very well be referring to someone using these words, for he may think that someone is the husband of the lady (who in fact is a spinster). Still, the statement is neither true nor false, for it presupposes that the lady has a husband, which she has not. This last refutes Strawson's thesis that if the presupposition of existence is not satisfied, the speaker has failed to refer. 
There is much that is right in this passage. But because Linsky does not make the distinction between the referential and the attributive uses of definite descriptions, it does not represent a wholly adequate account of the situation. A perhaps minor point about this passage is that Linsky apparently thinks it sufficient to establish that the speaker in his example is referring to someone by using the definite description "her husband," that he believe that someone is her husband. This will only approximate the truth provided that the "someone" in the description of the belief means "someone in particular" and is not merely the existential quantifier, "there is someone or other." For in both the attributive and the referential use the belief that someone or other is the husband of the lady is very likely to be present. If, for example, the speaker has just met the lady and, noticing her cheerfulness and radiant good health, makes his remark from his conviction that these attributes are always the result of having good husbands, he would be using the definite description attributively. Since she has no husband, there is no one to pick out as the person to
 "Reference and Referents," p. 80. It should be clear that I agree with Linsky in holding that a speaker may refer even though the "presupposition of existence" is not satisfied. And I agree in thinking this an objection to Strawson s view. I think, however, that this point, among others, can be used to define two distinct uses of definite descriptions which, in turn, yields a more general criticism of Strawson. So, while I develop here a point of difference, which grows out of the distinction I want to make, I find myself in agreement with much of Linsky's article.
whom he was referring. Nevertheless, the speaker believed that someone or other was her husband. On the other hand, if the use of "her husband" was simply a way of referring to a man the speaker has just met whom he assumed to be the lady's husband, he would have referred to that man even though neither he nor anyone else fits the description. I think it is likely that in this passage Linsky did mean by "someone," in his description of the belief, "someone in particular." But even then, as we have seen, we have neither a sufficient nor a necessary condition for a referential use of the definite description. A definite description can be used attributively even when the speaker believes that some particular thing or person fits the description, and it can be used referentially in the absence of this belief.
My main point, here, however, has to do with Linsky's view that because the presupposition is not satisfied, the statement is neither true nor false. This seems to me possibly correct if the definite description is thought of as being used attributively (depending upon whether we go with Strawson or Russell). But when we consider it as used referentially, this categorical assertion is no longer clearly correct. For the man the speaker referred to may indeed be kind to the spinster; the speaker may have said something true about that man. Now the difficulty is in the notion of "the statement." Suppose that we know that the lady is a spinster, but nevertheless know that the man referred to by the speaker is kind to her. It seems to me that we shall, on the one hand, want to hold that the speaker said something true, but be reluctant to express this by "It is true that her husband is kind to her."
This shows, I think, a difficulty in speaking simply about "the statement" when definite descriptions are used referentially. For the speaker stated something, in this example, about a particular person, and his statement, we may suppose, was true. Nevertheless, we should not like to agree with his statement by using the sentence he used; we should not like to identify the true statement via the speaker's words. The reason for this is not so hard to find. If we say, in this example, "It is true that her husband is kind to her," we are now using the definite description either attributively or referentially. But we should not be subscribing to what the original speaker truly said if we use the description attributively, for it was only in its function as referring to a particular person that the definite description yields the possibility of saying something true (since the lady has no husband). Our reluctance, however, to endorse the original speaker's statement by using the definite description referentially to
refer to the same person stems from quite a different consideration. For if we too were laboring under the mistaken belief that this man was the lady's husband, we could agree with the original speaker using his exact words. (Moreover, it is possible, as we have seen, deliberately to use a definite description to refer to someone we believe not to fit the description.) Hence, our reluctance to use the original speaker's words does not arise from the fact that if we did we should not succeed in stating anything true or false. It rather stems from the fact that when a definite description is used referentially there is a presumption that the speaker believes that what he refers to fits the description. Since we, who know the lady to be a spinster, would not normally want to give the impression that we believe otherwise, we would not like to use the original speaker's way of referring to the man in question.
How then would we express agreement with the original speaker without involving ourselves in unwanted impressions about our beliefs? The answer shows another difference between the referential and attributive uses of definite descriptions and brings out an important point about genuine referring.
When a speaker says, "Thej is y " where "the j " is used attributively, if there is no j , we cannot correctly report the speaker as having said of this or that person or thing that it is y . But if the definite description is used referentially we can report the speaker as having attributed y to something. And we may refer to what the speaker referred to, using whatever description or name suits our purpose. Thus, if a speaker says, "Her husband is kind to her," referring to the man he was just talking to, and if that man is Jones, we may report him as having said of Jones that he is kind to her. If Jones is also the president of the college, we may report the speaker as having said of the president of the college that he is kind to her. And finally, if we are talking to Jones, we may say, referring to the original speaker, "He said of you that you are kind to her." It does not matter here whether or not the woman has a husband or whether, if she does, Jones is her husband. If the original speaker referred to Jones, he said of him that he is kind to her. Thus where the definite description is used referentially, but does not fit what was referred to, we can report what a speaker said and agree with him by using a description or name which does fit. In doing so we need not, it is important to note, choose a description or name which the original speaker would agree fits what he was referring to. That is, we can report the speaker in the above case to have said truly of Jones that he is kind to her even if the original speaker
did not know that the man he was referring to is named Jones or even if he thinks he is not named Jones.
Returning to what Linsky said in the passage quoted, he claimed that, were someone to say "Her husband is kind to her," when she has no husband, the statement would be neither true nor false. As I have said, this is a likely view to hold if the definite description is being used attributively. But if it is being used referentially it is not clear what is meant by "the statement." If we think about what the speaker said about the person he referred to, then there is no reason to suppose he has not said something true or false about him, even though he is not the lady's husband. And Linsky's claim would be wrong. On the other hand, if we do not identify the statement in this way, what is the statement that the speaker made? To say that the statement he made was that her husband is kind to her lands us in difficulties. For we have to decide whether in using the definite description here in the identification of the statement, we are using it attributively or referentially. If the former, then we misrepresent the linguistic performance of the speaker; if the latter, then we are ourselves referring to someone and reporting the speaker to have said something of that person, in which case we are back to the possibility that he did say something true or false of that person.
I am thus drawn to the conclusion that when a speaker uses a definite description referentially he may have stated something true or false even if nothing fits the description, and that there is not a clear sense in which he has made a statement which is neither true nor false.
I want to end by a brief examination of a picture of what a genuine referring expression is that one might derive from Russell's views. I want to suggest that this picture is not so far wrong as one might suppose and that strange as this may seem, some of the things we have said about the referential use of definite descriptions are not foreign to this picture.
Genuine proper names, in Russell's sense, would refer to something without ascribing any properties to it. They would, one might say, refer to the thing itself, not simply the thing in so far as it falls under a certain description. Now this would seem to Russell something a definite description could not do, for he assumed that if definite descriptions were capable of referring at all, they would refer to something only in so far as that thing satisfied the description. Not only have we seen this assumption
 Cf. "The Philosophy of Logical Atomism," reprinted in Logic and Knowledge, p. 200.
to be false, however, but in the last section we saw something more. We saw that when a definite description is used referentially, a speaker can be reported as having said something of something. And in reporting what it was of which he said something we are not restricted to the description he used, or synonyms of it; we may ourselves refer to it using any descriptions, names, and so forth, that will do the job. Now this seems to give a sense in which we are concerned with the thing itself and not just the thing under a certain description, when we report the linguistic act of a speaker using a definite description referentially. That is, such a definite description comes closer to performing the function of Russell's proper names than certainly he supposed.
Secondly, Russell thought, I believe, that whenever we use descriptions, as opposed to proper names, we introduce an element of generality which ought to be absent if what we are doing is referring to some particular thing. This is clear from his analysis of sentences containing definite descriptions. One of the conclusions we are supposed to draw from that analysis is that such sentences express what are in reality completely general propositions: there is a f and only one such and any f is y . We might put this in a slightly different way. If there is anything which might be identified as reference here, it is reference in a very weak sense -- namely, reference to whatever is the one and only one f , if there is any such. Now this is something we might well say about the attributive use of definite descriptions, as should be evident from the previous discussion. But this lack of particularity is absent from the referential use of definite descriptions precisely because the description is here merely a device for getting one's audience to pick out or think of the thing to be spoken about, a device which may serve its function even if the description is incorrect. More importantly perhaps, in the referential use as opposed to the attributive, there is a right thing to be picked out by the audience and its being the right thing is not simply a function of its fitting the description.